<?xml version="1.0" encoding="UTF-8"?>
 <rdf:RDF xmlns="http://purl.org/rss/1.0/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:taxo="http://purl.org/rss/1.0/modules/taxonomy/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://web.resource.org/cc/" xmlns:syn="http://purl.org/rss/1.0/modules/syndication/" xmlns:admin="http://webns.net/mvcb/">
  <channel rdf:about="http://pinboard.in">
    <title>Pinboard (cshalizi)</title>
    <link>https://pinboard.in/u:cshalizi/public/</link>
    <description>recent bookmarks from cshalizi</description>
    <items>
      <rdf:Seq>	<rdf:li rdf:resource="https://arxiv.org/abs/2603.12277"/>
	<rdf:li rdf:resource="https://philpapers.org/archive/POIWDH.pdf"/>
	<rdf:li rdf:resource="https://www.annualreviews.org/content/journals/10.1146/annurev-polisci-032624-025000"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2606.13280"/>
	<rdf:li rdf:resource="https://hdl.handle.net/2027/heb01869.0001.001"/>
	<rdf:li rdf:resource="https://doi.org/10.1108/FTCIT-09-2025-0149"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2504.09762"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2604.19560"/>
	<rdf:li rdf:resource="https://direct.mit.edu/books/monograph/5848/AI-amp-IAn-Intellectual-History-of-Artificial"/>
	<rdf:li rdf:resource="https://direct.mit.edu/books/monograph/6116/The-Great-Energy-TransitionAmerica-from-1876-to"/>
	<rdf:li rdf:resource="https://www.cambridge.org/core/journals/philosophy-of-science/article/abs/carnapian-inductive-logic-for-exponential-smoothing/84DD31142459DFD0289CCE8915E79952?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles"/>
	<rdf:li rdf:resource="https://academic.oup.com/qje/article/141/2/1581/8435315?nbd=45497357721"/>
	<rdf:li rdf:resource="https://direct.mit.edu/books/oa-edited-volume/6112/Dennett-s-Real-Patterns-in-Science-and-Nature"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2601.10825"/>
	<rdf:li rdf:resource="https://www.cambridge.org/core/journals/philosophy-of-science/article/measured-inference-scales-statistics-and-scientific-inference/527F6793A1C954F01A321E72F780E931?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Issues"/>
	<rdf:li rdf:resource="https://www.sciencedirect.com/science/article/abs/pii/S0167268116301202?via%3Dihub"/>
	<rdf:li rdf:resource="https://jbgruber.github.io/rollama/"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2602.15902"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/1108.0799"/>
	<rdf:li rdf:resource="https://anadodik.github.io/publication/fever-dream/fever-dream.pdf"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2501.00663"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2602.04250"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2512.24999"/>
	<rdf:li rdf:resource="https://link.springer.com/article/10.1007/s10955-025-03555-1"/>
	<rdf:li rdf:resource="https://www.pnas.org/doi/10.1073/pnas.2411731121"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2512.00175"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2512.02208"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2512.02193"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2212.01987"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2510.15511"/>
	<rdf:li rdf:resource="https://osf.io/preprints/psyarxiv/ep3ub_v3"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2510.16250"/>
	<rdf:li rdf:resource="https://sociologicalscience.com/articles-v12-28-685/"/>
	<rdf:li rdf:resource="https://onlinelibrary.wiley.com/doi/full/10.1111/jtsa.70025?campaign=wolearlyview"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2403.02579"/>
	<rdf:li rdf:resource="https://proceedings.mlr.press/v195/bosch23a.html"/>
	<rdf:li rdf:resource="https://www.aeaweb.org/articles?id=10.1257/aer.20230458"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2509.04664"/>
	<rdf:li rdf:resource="https://link.springer.com/article/10.1007/BF02333197"/>
	<rdf:li rdf:resource="https://academic.oup.com/jeea/article-abstract/22/5/2261/7582277?redirectedFrom=PDF&amp;login=false"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2502.05148"/>
	<rdf:li rdf:resource="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5409144"/>
	<rdf:li rdf:resource="https://openreview.net/forum?id=SBE2q9qwZj"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2508.21055"/>
	<rdf:li rdf:resource="https://www.turing.ac.uk/news/publications/doing-ai-differently"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/1807.03819"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2508.05776"/>
	<rdf:li rdf:resource="https://www.nature.com/articles/s41586-025-09199-1"/>
	<rdf:li rdf:resource="https://martinuzzifrancesco.github.io/posts/a-brief-introduction-to-reservoir-computing/"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2412.13212"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/1808.04962"/>
	<rdf:li rdf:resource="https://www.nature.com/articles/s41467-024-45187-1"/>
	<rdf:li rdf:resource="https://academic.oup.com/sf/advance-article/doi/10.1093/sf/soaf079/8171706?login=false"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2406.01382"/>
	<rdf:li rdf:resource="https://philsci-archive.pitt.edu/26062/"/>
	<rdf:li rdf:resource="https://www.jair.org/index.php/jair/article/view/10640"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2507.06268"/>
	<rdf:li rdf:resource="https://www.sciencedirect.com/science/article/abs/pii/S0065245808604180"/>
	<rdf:li rdf:resource="https://www.cambridge.org/core/journals/american-political-science-review/article/social-media-social-control-and-the-politics-of-public-shaming/2BC3349DF48F25D83ADD3271FF2FCEB6"/>
	<rdf:li rdf:resource="https://www.stat.berkeley.edu/~brill/Papers/jtsa2012.pdf"/>
	<rdf:li rdf:resource="https://jid.journals.yorku.ca/index.php/jid/article/view/40591"/>
	<rdf:li rdf:resource="https://www.jstor.org/stable/4123261"/>
	<rdf:li rdf:resource="http://computationalculture.net/situating-bayesian-knowledge/"/>
	<rdf:li rdf:resource="https://aigi.ox.ac.uk/publications/chain-of-thought-is-not-explainability/"/>
	<rdf:li rdf:resource="https://www.tandfonline.com/doi/full/10.1080/26939169.2025.2497547"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2412.20292"/>
	<rdf:li rdf:resource="https://www.cambridge.org/core/journals/philosophy-of-science/article/demarcating-defining-and-diagnosing-pseudoscience/78BE2C62ABD782B7572154AAFF9F7D13?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2505.17120"/>
	<rdf:li rdf:resource="https://www.sciencedirect.com/science/article/pii/S1090513812001237"/>
	<rdf:li rdf:resource="https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00415/107615/PAQ-65-Million-Probably-Asked-Questions-and-What"/>
      </rdf:Seq>
    </items>
  </channel><item rdf:about="https://arxiv.org/abs/2603.12277">
    <title>[2603.12277] Prompt Injection as Role Confusion</title>
    <dc:date>2026-07-15T01:42:49+00:00</dc:date>
    <link>https://arxiv.org/abs/2603.12277</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["LLMs see the world as a single stream of text, partitioned into roles like <user> or <tool>. We trace prompt injection to role confusion: models perceive the source of text from how it sounds, not its labeled role. A command hidden in a webpage hijacks an agent simply because it sounds like <user> text, despite its <tool> label. We design role probes to measure how LLMs internally perceive "who is speaking," and find that injected text occupies the same representational space as the trusted role it imitates. We demonstrate this with CoT Forgery, a zero-shot attack that injects fabricated reasoning into user prompts and tool outputs. Models mistake the forgery for their own thoughts, yielding 60% attack success against frontier models with near-zero baselines. Strikingly, the degree of role confusion predicts attack success before a single token is generated. This mechanism generalizes beyond CoT Forgery to standard agent prompt injections, revealing prompt injection as a measurable consequence of role perception. To the model, sounding like a role is indistinguishable from being one. Project page and writeup: this https URL"

--- See [https://role-confusion.github.io/] for engaging plain-language write-up.]]></description>
<dc:subject>to:NB to_read large_language_models_(so_called)</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:b289438036ed/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://philpapers.org/archive/POIWDH.pdf">
    <title>What Do Historical Language Models Model?</title>
    <dc:date>2026-06-26T02:40:52+00:00</dc:date>
    <link>https://philpapers.org/archive/POIWDH.pdf</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Historical language models are increasingly used to infer attitudes, beliefs, or viewpoints from past societies. This paper argues that such uses rest on fragile epistemic assumptions. We show that historical language models do not simulate past minds or populations, but instead model the structure of surviving textual archives, which are shaped by systematic biases of literacy, genre, and preservation. By introducing a validity ladder that distinguishes textual, discursive, and population-level claims, we provide a framework for evaluating what kinds of historical inferences these models can legitimately support. This perspective clarifies how historical language models can contribute to research in the social sciences and humanities without encouraging over-interpretation."]]></description>
<dc:subject>to:NB to_read large_language_models_(so_called) historiography historiography_101 social_science_methodology social_measurement via:henry_farrell ginzburg!_thou_shouldst_be_living_at_this_hour</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:be678eb46e38/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:historiography"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:historiography_101"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_science_methodology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_measurement"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:henry_farrell"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:ginzburg!_thou_shouldst_be_living_at_this_hour"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.annualreviews.org/content/journals/10.1146/annurev-polisci-032624-025000">
    <title>Conceptualizing Academic Freedom | Annual Reviews</title>
    <dc:date>2026-06-18T14:27:59+00:00</dc:date>
    <link>https://www.annualreviews.org/content/journals/10.1146/annurev-polisci-032624-025000</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Academic freedom is an unusual and complex set of norms and practices. It arises out of the combination of the corporate self-governance of medieval universities and the spirit of disciplinary scientific inquiry in modern research universities. It combines a principle of antiorthodoxy as to conclusions with the robust associational self-governance of scholarly communities whose members evaluate one another as participants in that shared enterprise. It has never been easily or wholly embraced by wider societies; today it is under wholesale attack. This article combines conceptual, normative, and historical analyses of academic freedom as a general norm with attention to conflicts over it in the mid-to-late 2010s and early 2020s. Some genuinely hard cases and questions tested the meaning of academic freedom and university values well before the current crisis."]]></description>
<dc:subject>to:NB to_read academic_freedom via:henry_farrell</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:01dff0f0ccd5/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:academic_freedom"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:henry_farrell"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2606.13280">
    <title>[2606.13280] Generalization Bounds for Transformer-Based Next-Token Prediction in a Language Model</title>
    <dc:date>2026-06-17T16:26:17+00:00</dc:date>
    <link>https://arxiv.org/abs/2606.13280</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["A refined statistical understanding of LLM pre-training requires the analysis of the transformer architecture for data distributions that encapsulate key characteristics of text data. To address this, we propose a text data distribution based on an extension of the log-bilinear language model from the natural language processing literature. For this data generating process, we derive generalization bounds for deep transformer architectures, highlighting the dependence on the network architecture, the vocabulary size, the number of documents and the document length."]]></description>
<dc:subject>to:NB to_read learning_theory natural_language_processing large_language_models_(so_called) via:mraginsky</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:87e01de8c417/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:learning_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:natural_language_processing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://hdl.handle.net/2027/heb01869.0001.001">
    <title>Empires (Doyle, 1986)</title>
    <dc:date>2026-05-20T16:50:03+00:00</dc:date>
    <link>https://hdl.handle.net/2027/heb01869.0001.001</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- Tried explaining the paper to AEO and she immediately gave me more to read (in particular chapter 3).

CMU access link: https://www-fulcrum-org.cmu.idm.oclc.org/concern/monographs/5999n353w]]></description>
<dc:subject>to:NB books:noted to_read imperialism comparative_history re:do-institutions-evolve via:aeo</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e53f31fe26c3/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:books:noted"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:imperialism"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:comparative_history"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:do-institutions-evolve"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:aeo"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://doi.org/10.1108/FTCIT-09-2025-0149">
    <title>Volume 23 Issue 3-4 | Foundations and Trends in Communications and Information Theory | Emerald Publishing</title>
    <dc:date>2026-05-19T15:46:33+00:00</dc:date>
    <link>https://doi.org/10.1108/FTCIT-09-2025-0149</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Online learning is a foundational paradigm underlying applications from recommendation systems to the continual learning of modern AI models. Yet much of its theory centers on either fully adversarial or purely stochastic settings. However, real-world environments typically fall between these extremes, making classical models inadequate for describing practical behavior. This monograph develops a unified perspective for analyzing online learning under more nuanced and realistic environments. The authors approach the problem through the lens of universality from information theory and extend tools such as the Shtarkov sum, covering numbers and packing arguments to the online setting, revealing deeper structural connections between these two fields. Building on this viewpoint, they characterize minimax regret for logarithmic and Lipschitz losses, analyze expected regret under i.i.d. and more general stochastic processes and study hybrid adversarial–stochastic scenarios. The authors further develop constructive algorithms that achieve near-optimal regret guarantees, yielding a coherent and fine-grained information-theoretic framework of online universal learning."]]></description>
<dc:subject>to_read information_theory learning_theory learning_under_dependence in_NB low-regret_learning online_learning</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:eb2f88390f74/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:information_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:learning_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:learning_under_dependence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:low-regret_learning"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:online_learning"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2504.09762">
    <title>[2504.09762] Position: Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces!</title>
    <dc:date>2026-05-01T13:41:17+00:00</dc:date>
    <link>https://arxiv.org/abs/2504.09762</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Intermediate token generation (ITG), where a model produces output before the solution, has become a standard method to improve the performance of language models on reasoning tasks. These intermediate tokens have been called \say{reasoning traces} or even \say{thoughts} -- implicitly anthropomorphizing the traces, and implying that these traces resemble steps a human might take when solving a challenging problem, and as such can provide an interpretable window into the operation of the model's thinking process to the end user. In this position paper, we present evidence that this anthropomorphization isn't a harmless metaphor, and instead is quite dangerous -- it confuses the nature of these models and how to use them effectively, and leads to questionable research. We call on the community to avoid such anthropomorphization of intermediate tokens."]]></description>
<dc:subject>via:rvenkat to_read to:NB large_language_models_(so_called) chain-of-thought_(so_called)</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ade957324966/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:rvenkat"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:chain-of-thought_(so_called)"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2604.19560">
    <title>[2604.19560] Separating Geometry from Probability in the Analysis of Generalization</title>
    <dc:date>2026-04-22T20:22:34+00:00</dc:date>
    <link>https://arxiv.org/abs/2604.19560</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The goal of machine learning is to find models that minimize prediction error on data that has not yet been seen. Its operational paradigm assumes access to a dataset S and articulates a scheme for evaluating how well a given model performs on an arbitrary sample. The sample can be S (in which case we speak of ``in-sample'' performance) or some entirely new S′ (in which case we speak of ``out-of-sample'' performance). Traditional analysis of generalization assumes that both in- and out-of-sample data are i.i.d.\ draws from an infinite population. However, these probabilistic assumptions cannot be verified even in principle. This paper presents an alternative view of generalization through the lens of sensitivity analysis of solutions of optimization problems to perturbations in the problem data. Under this framework, generalization bounds are obtained by purely deterministic means and take the form of variational principles that relate in-sample and out-of-sample evaluations through an error term that quantifies how close out-of-sample data are to in-sample data. Statistical assumptions can then be used \textit{ex post} to characterize the situations when this error term is small (either on average or with high probability)."]]></description>
<dc:subject>to:NB to_read recht.benjamin raginsky.maxim learning_theory optimization via:mraginsky to_teach:childs_garden_of_statistical_learning_theory straight_into_my_veins interpolation</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:963cc7aaa897/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recht.benjamin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:raginsky.maxim"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:learning_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:optimization"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:childs_garden_of_statistical_learning_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:straight_into_my_veins"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:interpolation"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://direct.mit.edu/books/monograph/5848/AI-amp-IAn-Intellectual-History-of-Artificial">
    <title>AI &amp; I: An Intellectual History of Artificial Intelligence | Books Gateway | MIT Press</title>
    <dc:date>2026-04-16T16:45:59+00:00</dc:date>
    <link>https://direct.mit.edu/books/monograph/5848/AI-amp-IAn-Intellectual-History-of-Artificial</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["A concise and illuminating history of the field of artificial intelligence from one of its earliest and most respected pioneers.
"AI & I is an intellectual history of the field of artificial intelligence from the perspective of one of its first practitioners, Eugene Charniak. Charniak entered the field in 1967, roughly 12 years after AI's founding, and was involved in many of AI's formative milestones. In this book, he traces the trajectory of breakthroughs and disappointments of the discipline up to the current day, clearly and engagingly demystifying this oft revered and misunderstood technology. His argument is controversial but well supported: that classical AI has been almost uniformly unsuccessful and that the modern deep learning approach should be viewed as the foundation for all the exciting developments that are to come.
"Written for the scientifically educated layperson, this book chronicles the history of the field of AI, starting with its origin in 1956, as a topic for a small academic workshop held at Dartmouth University. From there, the author covers reasoning and knowledge representation, reasoning under uncertainty, chess, computer vision, speech recognition, language acquisition, deep learning, and learning writ large. Ultimately, Charniak takes issue with the controversy of AI—the fear that its invention means the end of jobs, creativity, and potentially even humans as a species—and explains why such concerns are unfounded. Instead, he believes that we should embrace the technology and all its potential to benefit society."

--- Charniak's statistical language processing book from the 1990s is great so I'm excited for this.]]></description>
<dc:subject>in_NB books:noted downloaded to_read artificial_intelligence machine_learning charniak.eugene</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ae501489213b/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:books:noted"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:downloaded"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:artificial_intelligence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:machine_learning"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:charniak.eugene"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://direct.mit.edu/books/monograph/6116/The-Great-Energy-TransitionAmerica-from-1876-to">
    <title>The Great Energy Transition: America from 1876 to 1929 | Books Gateway | MIT Press</title>
    <dc:date>2026-04-16T13:16:40+00:00</dc:date>
    <link>https://direct.mit.edu/books/monograph/6116/The-Great-Energy-TransitionAmerica-from-1876-to</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["How new forms of energy transformed every aspect of American life in a span of 50 years, from 1876 to 1929—and how it seeded our current polarization.

"The era of reform. The Gilded Age. The Progressive Era. What historians often divide into discrete eras was one period of profound change: a massive, multipronged energy transition. Oil, gas, and electricity were woven into a culture that had to heal sectional differences after the Civil War, absorb an enormous influx of immigrants, shift from a rural to an urban society, and adopt a scientific understanding of nature.
"Every job, business, house, and street underwent a transformation so rapid and radical that Americans simply could not grasp the larger pattern. The concepts of “technology” and an “energy transition” had yet to emerge, and observers struggled to understand their experiences using inadequate terms such as “kaleidoscopic change,” “applied science,” and “the machine age.” In The Great Energy Transition, David Nye documents this transformation—and explains our failure to see it for what it was.
"In this disorienting transformation, Nye locates the roots of today’s cultural polarization. The great energy transition accelerated demographic and economic trends, including higher wages, increasing longevity, the commodification of experience, engineering nature, corporatism, urbanization, resistance to science, and racial segregation. At the same time, the book points to the innovations and institutions that held the country together, from national parks and monuments to mass consumption and newly invented media events."]]></description>
<dc:subject>to:NB books:noted to_read american_history 19th_century_history 20th_century_history great_transformation the_present_before_it_was_widely_distributed re:the_singularity_in_our_past_light-cone books:suggest_to_library</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e461d24ea7cb/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:books:noted"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:american_history"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:19th_century_history"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:20th_century_history"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:great_transformation"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:the_present_before_it_was_widely_distributed"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:the_singularity_in_our_past_light-cone"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:books:suggest_to_library"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.cambridge.org/core/journals/philosophy-of-science/article/abs/carnapian-inductive-logic-for-exponential-smoothing/84DD31142459DFD0289CCE8915E79952?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles">
    <title>Carnapian Inductive Logic for Exponential Smoothing | Philosophy of Science | Cambridge Core</title>
    <dc:date>2026-04-15T13:16:02+00:00</dc:date>
    <link>https://www.cambridge.org/core/journals/philosophy-of-science/article/abs/carnapian-inductive-logic-for-exponential-smoothing/84DD31142459DFD0289CCE8915E79952?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["This paper explores the inductive logic associated with exponential smoothing, the most widely used predictive rule that manifests the idea that more recent observations have a stronger influence on predictive probabilities than more remote ones. The main result shows that exponential smoothing can be derived from a set of plausible qualitative invariance assumptions about conditional probabilities. I discuss various aspects of the resulting inductive logic, including its connections to exchangeable processes, to Bayesian predictive inference and kernel methods in machine learning, as well as the philosophy of probabilistic invariance conditions and symmetries."]]></description>
<dc:subject>to:NB prediction statistics non-stationarity induction to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:0b80819ed32c/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:prediction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:non-stationarity"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:induction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://academic.oup.com/qje/article/141/2/1581/8435315?nbd=45497357721">
    <title>How Do You Identify a Good Manager?* | The Quarterly Journal of Economics | Oxford Academic</title>
    <dc:date>2026-04-13T19:40:03+00:00</dc:date>
    <link>https://academic.oup.com/qje/article/141/2/1581/8435315?nbd=45497357721</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We introduce and validate a novel approach to identifying good managers. In a preregistered lab experiment, we causally identify managerial contributions by randomly assigning managers to teams and controlling for individual skill. We find that manager contributions are crucial for team success, and that people who self-select into management roles perform worse than randomly assigned managers. Managerial performance is strongly predicted by economic decision-making skill but not by demographic characteristics. Two validation studies support our experimental results. Participants who succeed in the lab receive more real-world promotions and, in a separate study of retail store managers, skill measures strongly predict store sales. A one standard deviation increase in manager quality increases annual per store sales by US$4.1 million (25% increase). Selecting managers on skills rather than demographic characteristics or the desire to lead could substantially improve organizational performance."

--- I want to believe, so this needs to be treated with skepticism.]]></description>
<dc:subject>to:NB experimental_economics management economics to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a8ebab345ac7/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:experimental_economics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:management"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:economics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://direct.mit.edu/books/oa-edited-volume/6112/Dennett-s-Real-Patterns-in-Science-and-Nature">
    <title>Dennett's Real Patterns in Science and Nature | Books Gateway | MIT Press</title>
    <dc:date>2026-04-08T17:36:05+00:00</dc:date>
    <link>https://direct.mit.edu/books/oa-edited-volume/6112/Dennett-s-Real-Patterns-in-Science-and-Nature</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["How the concept of a pattern, as understood in information science and applied in contemporary AI, can address deep questions in science and philosophy.
"The explosive growth of AI and machine learning in recent decades is predicated on the recognition and exploitation of patterns in data. Of course, scientists have engaged in their own—less automated—processes of pattern recognition since the birth of science itself, and biological organisms evolved their own neural networks for pattern recognition long before people and their technology came along.
"In his seminal work, “Real Patterns,” philosopher and cognitive scientist Daniel Dennett laid out a road map for connecting the idea of “patterns” as understood by information theory to the practices of scientists and to our own cognitive capacity to model and predict the world around us. In this book—the first dedicated to the topic of real patterns—Tyler Millhouse, Steve Petersen, and Don Ross follow this road map. They explore the relevance of patterns to important aspects of both science and nature, including the emergence of high-level structure in physics, the nature of biological species, the measurement of welfare in economics, the evaluation of causal models, and the possibility of understanding in large neural networks."]]></description>
<dc:subject>books:noted philosophy_of_science emergence dennett.daniel_c. to_read downloaded in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:3fd27bfc9cb4/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:books:noted"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:philosophy_of_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:emergence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:dennett.daniel_c."/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:downloaded"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2601.10825">
    <title>[2601.10825] Reasoning Models Generate Societies of Thought</title>
    <dc:date>2026-04-08T16:58:19+00:00</dc:date>
    <link>https://arxiv.org/abs/2601.10825</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Large language models have achieved remarkable capabilities across domains, yet mechanisms underlying sophisticated reasoning remain elusive. Recent reasoning models outperform comparable instruction-tuned models on complex cognitive tasks, attributed to extended computation through longer chains of thought. Here we show that enhanced reasoning emerges not from extended computation alone, but from simulating multi-agent-like interactions -- a society of thought -- which enables diversification and debate among internal cognitive perspectives characterized by distinct personality traits and domain expertise. Through quantitative analysis and mechanistic interpretability methods applied to reasoning traces, we find that reasoning models like DeepSeek-R1 and QwQ-32B exhibit much greater perspective diversity than instruction-tuned models, activating broader conflict between heterogeneous personality- and expertise-related features during reasoning. This multi-agent structure manifests in conversational behaviors, including question-answering, perspective shifts, and the reconciliation of conflicting views, and in socio-emotional roles that characterize sharp back-and-forth conversations, together accounting for the accuracy advantage in reasoning tasks. Controlled reinforcement learning experiments reveal that base models increase conversational behaviors when rewarded solely for reasoning accuracy, and fine-tuning models with conversational scaffolding accelerates reasoning improvement over base models. These findings indicate that the social organization of thought enables effective exploration of solution spaces. We suggest that reasoning models establish a computational parallel to collective intelligence in human groups, where diversity enables superior problem-solving when systematically structured, which suggests new opportunities for agent organization to harness the wisdom of crowds."]]></description>
<dc:subject>to:NB to_read artificial_intelligence large_language_models_(so_called) ensemble_methods evans.james kith_and_kin</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:f2cf63b3474b/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:artificial_intelligence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:ensemble_methods"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:evans.james"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kith_and_kin"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.cambridge.org/core/journals/philosophy-of-science/article/measured-inference-scales-statistics-and-scientific-inference/527F6793A1C954F01A321E72F780E931?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Issues">
    <title>Measured Inference: Scales, Statistics, and Scientific Inference | Philosophy of Science | Cambridge Core</title>
    <dc:date>2026-04-08T16:56:58+00:00</dc:date>
    <link>https://www.cambridge.org/core/journals/philosophy-of-science/article/measured-inference-scales-statistics-and-scientific-inference/527F6793A1C954F01A321E72F780E931?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Issues</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Despite the recent “epistemic turn” in the philosophy of measurement, philosophers have ignored a nearly 80-year controversy about the relationship between statistical inference and measurement theory. Some scholars maintain that measurement theory places no constraints on statistics, whereas others argue that the measurement scale (e.g., ordinal or interval) of one’s data determines which statistical methods are “permissible.” I defend an intermediate position: Even if existing measurement theory were irrelevant to statistical inference, it would be critical for scientific inference, which requires connecting statistical hypotheses to broader research hypotheses."]]></description>
<dc:subject>to:NB measurement philosophy_of_science statistics mayo-wilson.conor to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:cb7083aa4fd5/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:measurement"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:philosophy_of_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:mayo-wilson.conor"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.sciencedirect.com/science/article/abs/pii/S0167268116301202?via%3Dihub">
    <title>OPEC, the Seven Sisters, and oil market dominance: An evolutionary game theory and agent-based modeling approach - ScienceDirect</title>
    <dc:date>2026-03-27T19:56:07+00:00</dc:date>
    <link>https://www.sciencedirect.com/science/article/abs/pii/S0167268116301202?via%3Dihub</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["A methodological toolkit comprised of evolutionary game theory and agent-based modeling is used to study OPEC and the Seven Sisters as they struggled for control over global petroleum markets during the 1960s and 1970s. An evolutionary game theory model incorporates heterogeneous populations, energy-specific variables, and behavioral considerations to capture the fundamentals of the applied problem. An agent-based model is used to provide detailed results and demonstrate the importance of the natural resource to the outcome of the model."


]]></description>
<dc:subject>to:NB to_read economic_history agent-based_models evolutionary_game_theory via:aeo 20th_century_history</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ceba46bcf38f/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:economic_history"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:agent-based_models"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:evolutionary_game_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:aeo"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:20th_century_history"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://jbgruber.github.io/rollama/">
    <title>Communicate with Ollama to Run Large Language Models Locally • rollama</title>
    <dc:date>2026-03-26T16:16:04+00:00</dc:date>
    <link>https://jbgruber.github.io/rollama/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The goal of rollama is to wrap the Ollama API, which allows you to run different LLMs locally and create an experience similar to ChatGPT/OpenAI’s API. Ollama is very easy to deploy and handles a huge number of models. Checkout the project here: https://github.com/ollama/ollama.
"While there are several R packages for working with LLMs, rollama takes an opinionated approach centred on local, open-weight models: prioritising privacy, reproducibility, and ease of use for research tasks. The package and its learning materials are particularly focused on annotating text and images — making it a natural fit for (social) scientists who want to use LLMs without relying on proprietary APIs or sending sensitive data to third-party servers. It also offers deep integration with the Ollama ecosystem beyond just chat, including model management features like creating, copying, and pushing custom models."

--- Where "to_read" means "to install and play around with".  (Of course my 2017-vintage machine may groan and smoke...)]]></description>
<dc:subject>to:NB to_read large_language_models_(so_called) to_teach:statistics_and_generative_ai via:phnk</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:04a78f4bc198/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:statistics_and_generative_ai"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:phnk"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2602.15902">
    <title>[2602.15902] Doc-to-LoRA: Learning to Instantly Internalize Contexts</title>
    <dc:date>2026-03-11T14:08:14+00:00</dc:date>
    <link>https://arxiv.org/abs/2602.15902</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Long input sequences are central to in-context learning, document understanding, and multi-step reasoning of Large Language Models (LLMs). However, the quadratic attention cost of Transformers makes inference memory-intensive and slow. While context distillation (CD) can transfer information into model parameters, per-prompt distillation is impractical due to training costs and latency. To address these limitations, we propose Doc-to-LoRA (D2L), a lightweight hypernetwork that meta-learns to perform approximate CD within a single forward pass. Given an unseen prompt, D2L generates a LoRA adapter for a target LLM, enabling subsequent queries to be answered without re-consuming the original context, reducing latency and KV-cache memory consumption during inference of the target LLM. On a long-context needle-in-a-haystack task, D2L successfully learns to map contexts into adapters that store the needle information, achieving near-perfect zero-shot accuracy at sequence lengths exceeding the target LLM's native context window by more than 4x. On real-world QA datasets with limited compute, D2L outperforms standard CD while significantly reducing peak memory consumption and update latency. We envision that D2L can facilitate rapid adaptation of LLMs, opening up the possibility of frequent knowledge updates and personalized chat behavior."

--- Prompting is conditioning, and so localizing to a particular part of the state-space, so this is, what, changing the transition probabilities so we automatically stay in that region?  (Obvious answer is to read the paper and _then_ think about this...)]]></description>
<dc:subject>to:NB large_language_models_(so_called) to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:012f6f18ff71/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/1108.0799">
    <title>[1108.0799] Ito calculus without probability in idealized financial markets</title>
    <dc:date>2026-03-10T10:53:13+00:00</dc:date>
    <link>https://arxiv.org/abs/1108.0799</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We consider idealized financial markets in which price paths of the traded securities are cadlag functions, imposing mild restrictions on the allowed size of jumps. We prove the existence of quadratic variation for typical price paths, where the qualification "typical" means that there is a trading strategy that risks only one monetary unit and brings infinite capital if quadratic variation does not exist. This result allows one to apply numerous known results in pathwise Ito calculus to typical price paths; we give a brief overview of such results."]]></description>
<dc:subject>to:NB to_read stochastic_differential_equations mathematics vovk.vladimir finance via:mraginsky</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:d0fd5a421d09/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:stochastic_differential_equations"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:mathematics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:vovk.vladimir"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:finance"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://anadodik.github.io/publication/fever-dream/fever-dream.pdf">
    <title>The American Fever Dream: How Generative Models Privatize the Social Fabric</title>
    <dc:date>2026-02-26T17:37:29+00:00</dc:date>
    <link>https://anadodik.github.io/publication/fever-dream/fever-dream.pdf</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We put forth a critical theoretical framework for analyzing generative models both descriptively and normatively. Our thesis is that
generative models automate the production not only of intellectual labor or intelligence, but of a broader set of human social capacities
we name “social doing.” We do this by historicizing the commodification of sociality in the digital economy, leading to the availability
of social data as the precondition for generative models. We elaborate our definition of “social doing” by drawing a distinction between
“use” and “exchange” sociality and further differentiate between the ways that generative models either substitute for or mediate
existing social relations and processes. We then turn to existing empirical research on how people use generative model-based products
and the effects that their use has upon them. In this, we introduce the concept of the American Fever Dream, a social reality in part
fabricated by Silicon Valley’s privately owned and undemocratically governed generative models. Lastly, we offer a normative analysis
based on our findings and framework, and discuss future design opportunities."

--- Because obviously the best way to understand a new technology is to draw detailed analogies to the categories, and even the rhetoric, of an 1867 treatise on political economy.]]></description>
<dc:subject>to:NB to_read large_language_models_(so_called) re:ai_as_social_technology</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:f8ca0b2d610b/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:ai_as_social_technology"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2501.00663">
    <title>[2501.00663] Titans: Learning to Memorize at Test Time</title>
    <dc:date>2026-02-19T06:14:58+00:00</dc:date>
    <link>https://arxiv.org/abs/2501.00663</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Over more than a decade there has been an extensive research effort on how to effectively utilize recurrent models and attention. While recurrent models aim to compress the data into a fixed-size memory (called hidden state), attention allows attending to the entire context window, capturing the direct dependencies of all tokens. This more accurate modeling of dependencies, however, comes with a quadratic cost, limiting the model to a fixed-length context. We present a new neural long-term memory module that learns to memorize historical context and helps attention to attend to the current context while utilizing long past information. We show that this neural memory has the advantage of fast parallelizable training while maintaining a fast inference. From a memory perspective, we argue that attention due to its limited context but accurate dependency modeling performs as a short-term memory, while neural memory due to its ability to memorize the data, acts as a long-term, more persistent, memory. Based on these two modules, we introduce a new family of architectures, called Titans, and present three variants to address how one can effectively incorporate memory into this architecture. Our experimental results on language modeling, common-sense reasoning, genomics, and time series tasks show that Titans are more effective than Transformers and recent modern linear recurrent models. They further can effectively scale to larger than 2M context window size with higher accuracy in needle-in-haystack tasks compared to baselines."]]></description>
<dc:subject>to_read large_language_models_(so_called) neural_networks via:absfac in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:9af070ea390c/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neural_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:absfac"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2602.04250">
    <title>[2602.04250] A Note on Physical Dependence and Mixing Conditions for Triangular Arrays</title>
    <dc:date>2026-02-05T14:10:03+00:00</dc:date>
    <link>https://arxiv.org/abs/2602.04250</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Under mild structural assumptions and regularity conditions on the marginal and conditional densities, an explicit bound on the β-mixing coefficients in terms of the physical dependence measure is provided. Consequently, weak physical dependence implies β-mixing and strong mixing for triangular arrays, complementing Hill (2025), who proved the converse implication under moment assumptions."]]></description>
<dc:subject>to_read mixing re:codename:catherine_wheel in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a231e8b97c1d/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:mixing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:codename:catherine_wheel"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2512.24999">
    <title>[2512.24999] Basic Inequalities for First-Order Optimization with Applications to Statistical Risk Analysis</title>
    <dc:date>2026-01-07T15:13:39+00:00</dc:date>
    <link>https://arxiv.org/abs/2512.24999</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We introduce \textit{basic inequalities} for first-order iterative optimization algorithms, forming a simple and versatile framework that connects implicit and explicit regularization. While related inequalities appear in the literature, we isolate and highlight a specific form and develop it as a well-rounded tool for statistical analysis. Let f denote the objective function to be optimized. Given a first-order iterative algorithm initialized at θ0 with current iterate θT, the basic inequality upper bounds f(θT)−f(z) for any reference point z in terms of the accumulated step sizes and the distances between θ0, θT, and z. The bound translates the number of iterations into an effective regularization coefficient in the loss function. We demonstrate this framework through analyses of training dynamics and prediction risk bounds. In addition to revisiting and refining known results on gradient descent, we provide new results for mirror descent with Bregman divergence projection, for generalized linear models trained by gradient descent and exponentiated gradient descent, and for randomized predictors. We illustrate and supplement these theoretical findings with experiments on generalized linear models."]]></description>
<dc:subject>to:NB to_read optimization statistics re:HEAS tibshirani.ryan telgarsky.matus via:mraginsky</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:00ba72cd5eb8/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:optimization"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:HEAS"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:tibshirani.ryan"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:telgarsky.matus"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://link.springer.com/article/10.1007/s10955-025-03555-1">
    <title>Measure-Theoretic Time-Delay Embedding | Journal of Statistical Physics</title>
    <dc:date>2025-12-26T14:25:12+00:00</dc:date>
    <link>https://link.springer.com/article/10.1007/s10955-025-03555-1</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The celebrated Takens’ embedding theorem provides a theoretical foundation for reconstructing the full state of a dynamical system from partial observations. However, the classical theorem assumes that the underlying system is deterministic and that observations are noise-free, limiting its applicability in real-world scenarios. Motivated by these limitations, we formulate a measure-theoretic generalization that adopts an Eulerian description of the dynamics and recasts the embedding as a pushforward map between spaces of probability measures. Our mathematical results leverage recent advances in optimal transport. Building on the proposed measure-theoretic time-delay embedding theory, we develop a computational procedure that aims to reconstruct the full state of a dynamical system from time-lagged partial observations, engineered with robustness to handle sparse and noisy data. We evaluate our measure-based approach across several numerical examples, ranging from the classic Lorenz-63 system to real-world applications such as NOAA sea surface temperature reconstruction and ERA5 wind field reconstruction."]]></description>
<dc:subject>to:NB to_read state-space_reconstruction dynamical_systems stochastic_processes re:codename:catherine_wheel</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ddcab3885772/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:state-space_reconstruction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:dynamical_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:stochastic_processes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:codename:catherine_wheel"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.pnas.org/doi/10.1073/pnas.2411731121">
    <title>A local–global principle for nonequilibrium steady states | PNAS</title>
    <dc:date>2025-12-18T02:50:13+00:00</dc:date>
    <link>https://www.pnas.org/doi/10.1073/pnas.2411731121</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The global steady state of a system in thermal equilibrium exponentially favors configurations with lesser energy. This principle is a powerful explanation of self-organization because energy is a local property of configurations. For nonequilibrium systems, there is no such property for which an analogous principle holds, hence no common explanation of the diverse forms of self-organization they exhibit. However, a flurry of recent empirical results has shown that a local property of configurations called “rattling” predicts the steady states of some nonequilibrium systems, leading to claims of a far-reaching principle of nonequilibrium self-organization. But for which nonequilibrium systems is rattling accurate, and why? We develop a theory of rattling in terms of Markov processes that gives simple and precise answers to these key questions. Our results show that rattling predicts a broader class of nonequilibrium steady states than has been claimed and for different reasons than have been suggested. Its predictions hold to an extent determined by the relative variance of, and correlation between, the local and global “parts” of a steady state. We show how these quantities characterize the local-global relationships of various random walks on random graphs, spin-glass dynamics, and models of animal collective behavior. Surprisingly, we find that the core idea of rattling is so general as to apply to equilibrium and nonequilibrium systems alike."]]></description>
<dc:subject>to:NB non-equilibrium self-organization statistical_mechanics stochastic_processes markov_models randall.dana to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:1730b51f604e/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:non-equilibrium"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:self-organization"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistical_mechanics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:stochastic_processes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:markov_models"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:randall.dana"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2512.00175">
    <title>[2512.00175] Comparing Two Proxy Methods for Causal Identification</title>
    <dc:date>2025-12-07T15:10:34+00:00</dc:date>
    <link>https://arxiv.org/abs/2512.00175</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Identifying causal effects in the presence of unmeasured variables is a fundamental challenge in causal inference, for which proxy variable methods have emerged as a powerful solution. We contrast two major approaches in this framework: (1) bridge equation methods, which leverage solutions to integral equations to recover causal targets, and (2) array decomposition methods, which recover latent factors composing counterfactual quantities by exploiting unique determination of eigenspaces. We compare the model restrictions underlying these two approaches and provide insight into implications of the underlying assumptions, clarifying the scope of applicability for each method."]]></description>
<dc:subject>to:NB causal_inference statistics ogburn.elizabeth to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:17f84a7565d6/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:causal_inference"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:ogburn.elizabeth"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2512.02208">
    <title>[2512.02208] Projective limits of probabilistic symmetries and their applications to random graph limits</title>
    <dc:date>2025-12-06T14:27:58+00:00</dc:date>
    <link>https://arxiv.org/abs/2512.02208</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We couple projective limits of probability measures to direct limits of their symmetry groups. We show that the direct limit group is the group of symmetries of the projective limit probability measure. If projective systems of probability measures represent point processes in increasingly larger finite regions of the same infinite space, then we show that under some additional niceness and consistency assumptions, an extension of the direct limit group is the symmetry group of the projective limit point process in the whole infinite space. The application of these results to random graph limits provides ``shortest paths'' to graphons and graphexes as it recovers these random graph limits as trivial corollaries. Another application example encompasses a broad class of limits of random graphs with bounded average degrees. This class includes a representative collection of paradigmatic random graph models that have attracted significant research attention in diverse areas of science. Our approach thus provides a general unified framework to study limits of very different types of random graphs."]]></description>
<dc:subject>to:NB graph_theory graph_limits point_processes symmetry to_read krioukov.dmitri</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:6f35f1d08be2/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:graph_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:graph_limits"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:point_processes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:symmetry"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:krioukov.dmitri"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2512.02193">
    <title>[2512.02193] From monoliths to modules: Decomposing transducers for efficient world modelling</title>
    <dc:date>2025-12-06T14:26:05+00:00</dc:date>
    <link>https://arxiv.org/abs/2512.02193</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["World models have been recently proposed as sandbox environments in which AI agents can be trained and evaluated before deployment. Although realistic world models often have high computational demands, efficient modelling is usually possible by exploiting the fact that real-world scenarios tend to involve subcomponents that interact in a modular manner. In this paper, we explore this idea by developing a framework for decomposing complex world models represented by transducers, a class of models generalising POMDPs. Whereas the composition of transducers is well understood, our results clarify how to invert this process, deriving sub-transducers operating on distinct input-output subspaces, enabling parallelizable and interpretable alternatives to monolithic world modelling that can support distributed inference. Overall, these results lay a groundwork for bridging the structural transparency demanded by AI safety and the computational efficiency required for real-world inference."

--- There is a legend that the old Soviet mathematical literature contained a long series of papers which began "The following problem arises in the construction of long-range radar.  Let $B$ be a Banach space...", after which radar was never mentioned again.  From a quick scan, the baffle-gab about "alignment" serves a similar function here.  The math seems interesting, and perhaps even useful for statistics / "data-driven modeling"; someday we'll look back and ignore the cultic trappings.
--- More seriously, when reading this with pencil-and-paper, I should refresh my memory for how Krohn-Rhodes decomposition works (something I knew c. 2000 but have not used for a quarter century).]]></description>
<dc:subject>to:NB to_read transducers theoretical_computer_science automata_theory prediction</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:c73c05eb44db/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:transducers"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:theoretical_computer_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:automata_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:prediction"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2212.01987">
    <title>[2212.01987] Fractal dimensions for Iterated Graph Systems</title>
    <dc:date>2025-12-03T20:39:06+00:00</dc:date>
    <link>https://arxiv.org/abs/2212.01987</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Building upon [1], this study aims to introduce fractal geometry into graph theory, and to establish a potential theoretical foundation for complex networks. Specifically, we employ the method of substitution to create and explore fractal-like graphs, termed deterministic or random iterated graph systems. While the concept of substitution is commonplace in fractal geometry and dynamical systems, its analysis in the context of graph theory remains a nascent field.
"By delving into the properties of these systems, including diameter and distal, we derive two primary outcomes. Firstly, within the deterministic iterated graph systems, we establish that the Minkowski dimension and Hausdorff dimension align analytically through explicit formulae. Secondly, in the case of random iterated graph systems, we demonstrate that almost every graph limit exhibits identical Minkowski and Hausdorff dimensions numerically by their Lyapunov exponents.
"The exploration of iterated graph systems holds the potential to unveil novel directions. These findings not only, mathematically, contribute to our understanding of the interplay between fractals and graphs, but also, physically, suggest promising avenues for applications for complex networks."]]></description>
<dc:subject>to:NB fractals networks graph_theory re:fractal_network_asymptotics to_read scooped? via:vaguery</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:0d244921ef96/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:fractals"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:graph_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:fractal_network_asymptotics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:scooped?"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:vaguery"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2510.15511">
    <title>[2510.15511] Language Models are Injective and Hence Invertible</title>
    <dc:date>2025-11-06T15:56:39+00:00</dc:date>
    <link>https://arxiv.org/abs/2510.15511</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Transformer components such as non-linear activations and normalization are inherently non-injective, suggesting that different inputs could map to the same output and prevent exact recovery of the input from a model's representations. In this paper, we challenge this view. First, we prove mathematically that transformer language models mapping discrete input sequences to their corresponding sequence of continuous representations are injective and therefore lossless, a property established at initialization and preserved during training. Second, we confirm this result empirically through billions of collision tests on six state-of-the-art language models, and observe no collisions. Third, we operationalize injectivity: we introduce SipIt, the first algorithm that provably and efficiently reconstructs the exact input text from hidden activations, establishing linear-time guarantees and demonstrating exact invertibility in practice. Overall, our work establishes injectivity as a fundamental and exploitable property of language models, with direct implications for transparency, interpretability, and safe deployment."]]></description>
<dc:subject>to_read large_language_models_(so_called) via:mraginsky in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e4c52e75e008/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://osf.io/preprints/psyarxiv/ep3ub_v3">
    <title>OSF | Network structure explains intellectual discourse across human history</title>
    <dc:date>2025-10-28T17:50:12+00:00</dc:date>
    <link>https://osf.io/preprints/psyarxiv/ep3ub_v3</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The production of knowledge is a collective endeavor. Scientific discovery, for instance, reflects not only the insights of individual scientists but also interactions among scientists. This is often taken to reflect social influences on collective epistemic vitality, the capacity to generate and synthesize new ideas. However, in science, it is difficult to separate intellectual influence from access to material resources, such as equipment and grant funding, since networks of collaboration influence the circulation of both intellectual and non-intellectual resources. Here, as a strict test of how social structure shapes intellectual discourse, we use the three-thousand-year history of a human debate in communities that relied on intellectual argumentation (rather than, say, empirical experiments). Building on the work of historians and sociologists, we digitized and quantified the time-evolving network structure of interaction among intellectuals (N = 3187), broadly construed, from religious debate in ancient India (c. 800 BCE) to 20th century debates about the logical foundations of mathematics in Europe and North America. We find that the production or preservation of knowledge by a community is explained by its network structure but not with overall levels of antagonism, suggesting that how communities are organized matters more for intellectual progress than how contentious they are. Extending tools from collective intelligence to intellectual history, we call for an integration of the science of science, the philosophy of science, and the history of ideas to forge a comprehensive understanding of the social dynamics of knowledge."

--- These are ambitious conclusions to draw from four cases.]]></description>
<dc:subject>to_read color_me_skeptical social_networks social_life_of_the_mind</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:581c5eeb9956/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:color_me_skeptical"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_life_of_the_mind"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2510.16250">
    <title>[2510.16250] One-Bit Quantization for Random Features Models</title>
    <dc:date>2025-10-25T20:06:30+00:00</dc:date>
    <link>https://arxiv.org/abs/2510.16250</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Recent advances in neural networks have led to significant computational and memory demands, spurring interest in one-bit weight compression to enable efficient inference on resource-constrained devices. However, the theoretical underpinnings of such compression remain poorly understood. We address this gap by analyzing one-bit quantization in the Random Features model, a simplified framework that corresponds to neural networks with random representations. We prove that, asymptotically, quantizing weights of all layers except the last incurs no loss in generalization error, compared to the full precision random features model. Our findings offer theoretical insights into neural network compression. We also demonstrate empirically that one-bit quantization leads to significant inference speed ups for the Random Features models even on a laptop GPU, confirming the practical benefits of our work. Additionally, we provide an asymptotically precise characterization of the generalization error for Random Features with an arbitrary number of layers. To the best of our knowledge, our analysis yields more general results than all previous works in the related literature."]]></description>
<dc:subject>to_read random_features neural_networks in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:8c8e1f29f66c/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:random_features"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neural_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://sociologicalscience.com/articles-v12-28-685/">
    <title>Complex Contagion in Social Networks: Causal Evidence from a Country-Scale Field Experiment</title>
    <dc:date>2025-10-25T20:05:30+00:00</dc:date>
    <link>https://sociologicalscience.com/articles-v12-28-685/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Complex contagion rests on the idea that individuals are more likely to adopt a behavior if they experience social reinforcement from multiple sources. We develop a test for complex contagion, conceptualized as social reinforcement, and then use it to examine whether empirical data from a country-scale randomized controlled viral marketing field experiment show evidence of complex contagion. The experiment uses a peer encouragement design in which individuals were randomly exposed to either one or two friends who were encouraged to share a coupon for a mobile data product. Using three different analytical methods to address the empirical challenges of causal identification, we provide strong support for complex contagion: the contagion process cannot be understood as independent cascades but rather as a process in which signals from multiple sources amplify each other through synergistic interdependence. We also find social network embeddedness is an important structural moderator that shapes the effectiveness of social reinforcement."]]></description>
<dc:subject>to:NB contagion diffusion_of_innovations social_networks re:homophily_and_confounding lazer.david to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:69d0da3f7294/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:contagion"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:diffusion_of_innovations"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:homophily_and_confounding"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:lazer.david"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://onlinelibrary.wiley.com/doi/full/10.1111/jtsa.70025?campaign=wolearlyview">
    <title>Density‐Valued ARMA Models by Spline Mixtures - Matsuda - Journal of Time Series Analysis - Wiley Online Library</title>
    <dc:date>2025-10-25T20:02:50+00:00</dc:date>
    <link>https://onlinelibrary.wiley.com/doi/full/10.1111/jtsa.70025?campaign=wolearlyview</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["This paper proposes a novel framework for modeling time series of probability density functions by extending autoregressive moving average (ARMA) models to density-valued data. The method is based on a transformation approach, wherein each density function on a compact domain $[0,1]^d$ is approximated by a B-spline mixture representation. Through generalized logit and softmax mappings, the space of density functions is transformed into an unconstrained Euclidean space, enabling the application of classical time series techniques. We define ARMA-type dynamics in the transformed space. Estimation is carried out via least squares for density-valued AR models and Whittle likelihood for ARMA models, with asymptotic normality derived under the joint divergence of the time horizon and basis dimension. The proposed methodology is applied to spatiotemporal human population data in Tokyo, where meaningful temporal structures in the distributional dynamics are successfully captured."]]></description>
<dc:subject>to:NB splines time_series nonparametrics to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:2c29c195dd77/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:splines"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:time_series"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:nonparametrics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2403.02579">
    <title>[2403.02579] Geometric Dynamics of Signal Propagation Predict Trainability of Transformers</title>
    <dc:date>2025-10-22T19:00:48+00:00</dc:date>
    <link>https://arxiv.org/abs/2403.02579</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We investigate forward signal propagation and gradient back propagation in deep, randomly initialized transformers, yielding simple necessary and sufficient conditions on initialization hyperparameters that ensure trainability of deep transformers. Our approach treats the evolution of the representations of n tokens as they propagate through the transformer layers in terms of a discrete time dynamical system of n interacting particles. We derive simple update equations for the evolving geometry of this particle system, starting from a permutation symmetric simplex. Our update equations show that without MLP layers, this system will collapse to a line, consistent with prior work on rank collapse in transformers. However, unlike prior work, our evolution equations can quantitatively track particle geometry in the additional presence of nonlinear MLP layers, and it reveals an order-chaos phase transition as a function of initialization hyperparameters, like the strength of attentional and MLP residual connections and weight variances. In the ordered phase the particles are attractive and collapse to a line, while in the chaotic phase the particles are repulsive and converge to a regular n-simplex. We analytically derive two Lyapunov exponents: an angle exponent that governs departures from the edge of chaos in this particle system, and a gradient exponent that governs the rate of exponential growth or decay of backpropagated gradients. We show through experiments that, remarkably, the final test loss at the end of training is well predicted just by these two exponents at the beginning of training, and that the simultaneous vanishing of these two exponents yields a simple necessary and sufficient condition to achieve minimal test loss."]]></description>
<dc:subject>to_read interacting_particle_systems large_language_models_(so_called) re:large_language_models_in_statistical_perspective in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:79276f50856c/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:interacting_particle_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:large_language_models_in_statistical_perspective"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://proceedings.mlr.press/v195/bosch23a.html">
    <title>Precise Asymptotic Analysis of Deep Random Feature Models</title>
    <dc:date>2025-09-27T10:58:52+00:00</dc:date>
    <link>https://proceedings.mlr.press/v195/bosch23a.html</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We provide exact asymptotic expressions for the performance of regression by an L-layer deep random feature (RF) model, where the input is mapped through multiple random embedding and non-linear activation functions. For this purpose, we establish two key steps: First, we prove a novel universality result for RF models and deterministic data, by which we demonstrate that a deep random feature model is equivalent to a deep linear Gaussian model that matches it in the first and second moments, at each layer. Second, we make use of the convex Gaussian Min-Max theorem multiple times to obtain the exact behavior of deep RF models. We further characterize the variation of the eigendistribution in different layers of the equivalent Gaussian model, demonstrating that depth has a tangible effect on model performance despite the fact that only the last layer of the model is being trained."]]></description>
<dc:subject>to:NB to_read random_features via:mraginsky</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:5b927f9752dd/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:random_features"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.aeaweb.org/articles?id=10.1257/aer.20230458">
    <title>A Stepping Stone Approach to Norm Transitions - American Economic Association</title>
    <dc:date>2025-09-22T17:14:04+00:00</dc:date>
    <link>https://www.aeaweb.org/articles?id=10.1257/aer.20230458</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We propose a model to study when an intermediate action can serve as a stepping stone that enables the elimination of a harmful norm. While the intermediate action may facilitate the first "step," it may also become a new norm. We derive intuitive conditions for stepping stones, which depend on the relative size of social penalties and intrinsic utility benefits. We propose an econometric approach to testing whether an intermediate action is a stepping stone, and apply it to original data on female genital cutting in Somalia. The analysis shows that the intermediate action may become the new norm."]]></description>
<dc:subject>to:NB to_read young.h_peyton institutions cultural_evolution re:do-institutions-evolve evolutionary_game_theory</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:97fbe7c5d5da/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:young.h_peyton"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:institutions"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:cultural_evolution"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:do-institutions-evolve"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:evolutionary_game_theory"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2509.04664">
    <title>[2509.04664] Why Language Models Hallucinate</title>
    <dc:date>2025-09-22T16:18:57+00:00</dc:date>
    <link>https://arxiv.org/abs/2509.04664</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Like students facing hard exam questions, large language models sometimes guess when uncertain, producing plausible yet incorrect statements instead of admitting uncertainty. Such "hallucinations" persist even in state-of-the-art systems and undermine trust. We argue that language models hallucinate because the training and evaluation procedures reward guessing over acknowledging uncertainty, and we analyze the statistical causes of hallucinations in the modern training pipeline. Hallucinations need not be mysterious -- they originate simply as errors in binary classification. If incorrect statements cannot be distinguished from facts, then hallucinations in pretrained language models will arise through natural statistical pressures. We then argue that hallucinations persist due to the way most evaluations are graded -- language models are optimized to be good test-takers, and guessing when uncertain improves test performance. This "epidemic" of penalizing uncertain responses can only be addressed through a socio-technical mitigation: modifying the scoring of existing benchmarks that are misaligned but dominate leaderboards, rather than introducing additional hallucination evaluations. This change may steer the field toward more trustworthy AI systems."]]></description>
<dc:subject>to:NB to_read via:everyone to_teach:statistics_and_generative_ai large_language_models_(so_called)</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e279ffa5ce6d/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:everyone"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:statistics_and_generative_ai"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://link.springer.com/article/10.1007/BF02333197">
    <title>On some alleged philosophical implications of mathematical logic (Benes, 1953) | Philosophical Studies</title>
    <dc:date>2025-09-05T16:19:56+00:00</dc:date>
    <link>https://link.springer.com/article/10.1007/BF02333197</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>to:NB to_read mathematics logic philosophy via:mraginsky</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:647cf3750676/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:mathematics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:logic"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:philosophy"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://academic.oup.com/jeea/article-abstract/22/5/2261/7582277?redirectedFrom=PDF&amp;login=false">
    <title>Social Conflict and the Evolution of Unequal Conventions | Journal of the European Economic Association | Oxford Academic</title>
    <dc:date>2025-09-05T16:13:38+00:00</dc:date>
    <link>https://academic.oup.com/jeea/article-abstract/22/5/2261/7582277?redirectedFrom=PDF&amp;login=false</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We propose a theory of social norms (or conventions) that implement substantial levels of inequality between men and women, ethnic groups, and classes and that persist over long periods of time despite being inefficient and not supported by formal institutions. Consistent with historical cases, we extend the standard asymmetric stochastic evolutionary game model to allow subpopulation sizes to differ and idiosyncratic rejection of a status quo convention to be intentional to some degree (rather than purely random as in the standard evolutionary models). In this setting, if idiosyncratic play is sufficiently intentional and the subordinate class is sufficiently large relative to the elite, then risk-dominated conventions that are both more unequal and inefficient relative to alternative conventions will be stochastically stable and may persist for long periods. We show that the same is true in a general bipartite network of the population if most of the subordinate groups interactions are local, while the elite is more “cosmopolitan”. We apply the model to the evolution of wage conventions on the bipartite network of workers and employers, and find that an unequal monopsonistic wage convention is robust to the idiosyncratic play of workers that otherwise might displace it."

--- Need to see how/if this differs from earlier Naidu&Bowles papers...]]></description>
<dc:subject>to:NB to_read cultural_evolution institutions inequality bowles.samuel naidu.suresh to_teach:statistics_of_inequality_and_discrimination via:henry_farrell</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:de97228cfc12/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:cultural_evolution"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:institutions"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:inequality"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bowles.samuel"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:naidu.suresh"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:statistics_of_inequality_and_discrimination"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:henry_farrell"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2502.05148">
    <title>[2502.05148] An Annotated Reading of 'The Singer of Tales' in the LLM Era</title>
    <dc:date>2025-09-05T16:07:32+00:00</dc:date>
    <link>https://arxiv.org/abs/2502.05148</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The Parry-Lord oral-formulaic theory was a breakthrough in understanding how oral narrative poetry is learned, composed, and transmitted by illiterate bards. In this paper, we provide an annotated reading of the mechanism underlying this theory from the lens of large language models (LLMs) and generative artificial intelligence (AI). We point out the the similarities and differences between oral composition and LLM generation, and comment on the implications to society and AI policy."]]></description>
<dc:subject>oral_epics re:gopnikism large_language_models_(so_called) via:henry_farrell to_read in_NB re:the_singer_of_tales_and_the_house_of_intellect</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:2c883327eae9/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:oral_epics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:gopnikism"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:henry_farrell"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:the_singer_of_tales_and_the_house_of_intellect"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5409144">
    <title>Computational Hermeneutics: Evaluating Generative AI as a Cultural Technology by Cody Kommers, Ruth Ahnert, Maria Antoniak, Emmanouil Benetos, Steve Benford, Mercedes Bunz, Baptiste Caramiaux, Shauna Concannon, Martin Disley, James Dobson, Yali Du, Edgar </title>
    <dc:date>2025-09-05T15:05:09+00:00</dc:date>
    <link>https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5409144</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Generative AI (GenAI) systems are increasingly recognized as cultural technologies, yet current evaluation frameworks often treat culture as a variable to be measured rather than fundamental to the system’s operation. Drawing on hermeneutic theory from the humanities, we argue that GenAI systems function as “context machines” that must inherently address three interpretive challenges: situatedness (meaning only emerges in context), plurality (multiple valid interpretations coexist), and ambiguity (interpretations naturally conflict). We present computational hermeneutics as an emerging framework offering an interpretive account of what GenAI systems do, and how they might do it better. We offer three principles for hermeneutic evaluation—that benchmarks should be iterative, not one-off; include people, not just machines; and measure cultural context, not just model output. This perspective offers a nascent paradigm for designing and evaluating contemporary AI systems: shifting from standardized questions about accuracy to contextual ones about meaning."]]></description>
<dc:subject>to:NB to_read large_language_models_(so_called) humanities interpretation re:gopnikism</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a9eb550e3ac3/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:humanities"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:interpretation"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:gopnikism"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://openreview.net/forum?id=SBE2q9qwZj">
    <title>Fast Computation of Leave-One-Out Cross-Validation for $k$-NN Regression | OpenReview</title>
    <dc:date>2025-09-02T19:30:09+00:00</dc:date>
    <link>https://openreview.net/forum?id=SBE2q9qwZj</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We describe a fast computation method for leave-one-out cross-validation (LOOCV) for 
$k$-nearest neighbours ($k$-NN) regression. We show that, under a tie-breaking condition for nearest neighbours, the LOOCV estimate of the mean square error for $k$
-NN regression is identical to the mean square error of $(k+1)$-NN regression evaluated on the training data, multiplied by the scaling factor $(k+1)^2/𝑘^2$. Therefore, to compute the LOOCV score, one only needs to fit $(k+1)$-NN regression only once, and does not need to repeat training-validation of $k$-NN regression for the number of training data. Numerical experiments confirm the validity of the fast computation method."

!!!]]></description>
<dc:subject>to:NB to_read nearest_neighbors to_teach:data-mining to_teach:undergrad-ADA via:?</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a7ddd207a4f6/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:nearest_neighbors"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:undergrad-ADA"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:?"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2508.21055">
    <title>[2508.21055] Modern aspects of Markov chains: entropy, curvature and the cutoff phenomenon</title>
    <dc:date>2025-09-02T02:49:02+00:00</dc:date>
    <link>https://arxiv.org/abs/2508.21055</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The cutoff phenomenon is an abrupt transition from out of equilibrium to equilibrium undergone by certain Markov processes in the limit where the size of the state space tends to infinity: instead of decaying gradually over time, their distance to equilibrium remains close to its maximal value for a while and suddenly drops to zero as the time parameter reaches a critical threshold. Discovered four decades ago in the context of card shuffling, this surprising phenomenon has since then been observed in a variety of models, from random walks on groups or complex networks to interacting particle systems. It is now believed to be universal among fast-mixing high-dimensional processes. Yet, current proofs are heavily model-dependent, and identifying the general conditions that trigger a cutoff remains one of the biggest challenges in the quantitative analysis of finite Markov chains. The purpose of these lecture notes is to provide a self-contained introduction to this fascinating question, and to describe its recently-uncovered relations with entropy, curvature and concentration."]]></description>
<dc:subject>to:NB to_read markov_models stochastic_processes</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:4b52c6b91f38/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:markov_models"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:stochastic_processes"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.turing.ac.uk/news/publications/doing-ai-differently">
    <title>Doing AI differently | The Alan Turing Institute</title>
    <dc:date>2025-08-25T13:52:18+00:00</dc:date>
    <link>https://www.turing.ac.uk/news/publications/doing-ai-differently</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Artificial Intelligence is rapidly becoming global infrastructure – shaping decisions in healthcare, education, industry and everyday life. Yet current AI systems face a fundamental limitation: they are shaped by narrow operational metrics that fail to reflect the diversity, ambiguity and richness of human experience.
"This white paper presents a research vision that positions interpretive depth as essential to building AI systems capable of engaging meaningfully with cultural complexity – while recognising that no technical solution alone can resolve the challenges these systems face in diverse human contexts.
"Accompanying the white paper is a policy note and a methodology report – links to all publications can be found below.
"Doing AI Differently calls for a fundamental shift in AI development – one that positions the humanities, arts and qualitative social sciences as integral, rather than supplemental, to technical innovation.
"Three critical challenges
"The qualitative turn: AI is no longer limited to structured prediction or optimisation – it now operates in tasks that require contextual judgement, cultural nuance, and interpretive reasoning.
"The homogenisation problem: The dominance of a few AI architectures propagates design limitations across countless applications and can entrench social inequalities by reinforcing narrow models of reasoning and representation.
"The transformation of human cognition: As we engage with complex, interconnected systems of artificial and human agents, AI is reshaping human thinking and work in ways that risk diminishing rather than enhancing human agency and capabilities.
"The core innovations we envision
"Interpretive technologies: AI systems that represent multiple valid perspectives rather than producing monolithic outputs, enabling more nuanced, culturally sensitive reasoning across diverse contexts.
"Alternative architectures for AI: Expanding the AI design space beyond current homogeneous approaches through diverse reasoning paradigms grounded in heterogeneous cognitive, cultural and planetary processes.
"Human-AI ensembles: Developing frameworks for sophisticated, collaborative human-AI systems that strengthen our collective intelligence and enhance rather than replace human capabilities in complex decision-making.
"This is more than a report – it is a call to action and a plan for change. We invite researchers, institutions and funders to join us in this crucial endeavour to unite the humanities, data science and engineering in shaping the future of AI."

The Doing AI Differently Initiative is led by the Alan Turing Institute, University of Edinburgh and the UK’s Arts & Humanities Research Council (AHRC-UKRI), in collaboration with international partners.

Prepared as part of the Sustainability Mission at the Alan Turing Institute.

Funded by Arts & Humanities Research Council and Lloyd’s Register Foundation.]]></description>
<dc:subject>to:NB to_read artificial_intelligence large_language_models_(so_called) humanities</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ade8f35f82e9/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:artificial_intelligence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:humanities"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/1807.03819">
    <title>[1807.03819] Universal Transformers</title>
    <dc:date>2025-08-20T03:25:31+00:00</dc:date>
    <link>https://arxiv.org/abs/1807.03819</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Recurrent neural networks (RNNs) sequentially process data by updating their state with each new data point, and have long been the de facto choice for sequence modeling tasks. However, their inherently sequential computation makes them slow to train. Feed-forward and convolutional architectures have recently been shown to achieve superior results on some sequence modeling tasks such as machine translation, with the added advantage that they concurrently process all inputs in the sequence, leading to easy parallelization and faster training times. Despite these successes, however, popular feed-forward sequence models like the Transformer fail to generalize in many simple tasks that recurrent models handle with ease, e.g. copying strings or even simple logical inference when the string or formula lengths exceed those observed at training time. We propose the Universal Transformer (UT), a parallel-in-time self-attentive recurrent sequence model which can be cast as a generalization of the Transformer model and which addresses these issues. UTs combine the parallelizability and global receptive field of feed-forward sequence models like the Transformer with the recurrent inductive bias of RNNs. We also add a dynamic per-position halting mechanism and find that it improves accuracy on several tasks. In contrast to the standard Transformer, under certain assumptions, UTs can be shown to be Turing-complete. Our experiments show that UTs outperform standard Transformers on a wide range of algorithmic and language understanding tasks, including the challenging LAMBADA language modeling task where UTs achieve a new state of the art, and machine translation where UTs achieve a 0.9 BLEU improvement over Transformers on the WMT14 En-De dataset."]]></description>
<dc:subject>to_read large_language_models_(so_called) in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:2ebbe7c393ac/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2508.05776">
    <title>[2508.05776] Whither symbols in the era of advanced neural networks?</title>
    <dc:date>2025-08-16T13:20:46+00:00</dc:date>
    <link>https://arxiv.org/abs/2508.05776</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Some of the strongest evidence that human minds should be thought about in terms of symbolic systems has been the way they combine ideas, produce novelty, and learn quickly. We argue that modern neural networks -- and the artificial intelligence systems built upon them -- exhibit similar abilities. This undermines the argument that the cognitive processes and representations used by human minds are symbolic, although the fact that these neural networks are typically trained on data generated by symbolic systems illustrates that such systems play an important role in characterizing the abstract problems that human minds have to solve. This argument leads us to offer a new agenda for research on the symbolic basis of human thought."]]></description>
<dc:subject>to:NB cognitive_science artificial_intelligence large_language_models_(so_called) neural_networks via:melanie_mitchell to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:35961d1fced6/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:cognitive_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:artificial_intelligence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neural_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:melanie_mitchell"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.nature.com/articles/s41586-025-09199-1">
    <title>The dynamics and geometry of choice in the premotor cortex | Nature</title>
    <dc:date>2025-08-15T22:37:22+00:00</dc:date>
    <link>https://www.nature.com/articles/s41586-025-09199-1</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The brain represents sensory variables in the coordinated activity of neural populations, in which tuning curves of single neurons define the geometry of the population code1,2. Whether the same coding principle holds for dynamic cognitive variables remains unknown because internal cognitive processes unfold with a unique time course on single trials observed only in the irregular spiking of heterogeneous neural populations3,4,5,6,7,8. Here we show the existence of such a population code for the dynamics of choice formation in the primate premotor cortex. We developed an approach to simultaneously infer population dynamics and tuning functions of single neurons to the population state. Applied to spike data recorded during decision-making, our model revealed that populations of neurons encoded the same dynamic variable predicting choices, and heterogeneous firing rates resulted from the diverse tuning of single neurons to this decision variable. The inferred dynamics indicated an attractor mechanism for decision computation. Our results reveal a unifying geometric principle for neural encoding of sensory and dynamic cognitive variables."]]></description>
<dc:subject>to:NB to_read neuroscience neural_coding_and_decoding neural_control_of_action decision-making</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:51b6338e7543/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neuroscience"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neural_coding_and_decoding"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neural_control_of_action"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:decision-making"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://martinuzzifrancesco.github.io/posts/a-brief-introduction-to-reservoir-computing/">
    <title>A brief introduction to Reservoir Computing | Francesco Martinuzzi</title>
    <dc:date>2025-08-07T15:03:08+00:00</dc:date>
    <link>https://martinuzzifrancesco.github.io/posts/a-brief-introduction-to-reservoir-computing/</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>to:NB to_read reservoir_computing dynamical_systems prediction re:codename:catherine_wheel</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:b949b5cf35ce/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:reservoir_computing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:dynamical_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:prediction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:codename:catherine_wheel"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2412.13212">
    <title>[2412.13212] An introduction to reservoir computing</title>
    <dc:date>2025-08-07T14:51:00+00:00</dc:date>
    <link>https://arxiv.org/abs/2412.13212</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["There is a growing interest in the development of artificial neural networks that are implemented in a physical system. A major challenge in this context is that these networks are difficult to train since training here would require a change of physical parameters rather than simply of coefficients in a computer program. For this reason, reservoir computing, where one employs high-dimensional recurrent networks and trains only the final layer, is widely used in this context. In this chapter, I introduce the basic concepts of reservoir computing. Moreover, I present some important physical implementations coming from electronics, photonics, spintronics, mechanics, and biology. Finally, I provide a brief discussion of quantum reservoir computing."]]></description>
<dc:subject>to:NB reservoir_computing to_read re:codename:catherine_wheel</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:f06002878a5c/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:reservoir_computing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:codename:catherine_wheel"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/1808.04962">
    <title>[1808.04962] Recent Advances in Physical Reservoir Computing: A Review</title>
    <dc:date>2025-08-07T14:49:55+00:00</dc:date>
    <link>https://arxiv.org/abs/1808.04962</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Reservoir computing is a computational framework suited for temporal/sequential data processing. It is derived from several recurrent neural network models, including echo state networks and liquid state machines. A reservoir computing system consists of a reservoir for mapping inputs into a high-dimensional space and a readout for pattern analysis from the high-dimensional states in the reservoir. The reservoir is fixed and only the readout is trained with a simple method such as linear regression and classification. Thus, the major advantage of reservoir computing compared to other recurrent neural networks is fast learning, resulting in low training cost. Another advantage is that the reservoir without adaptive updating is amenable to hardware implementation using a variety of physical systems, substrates, and devices. In fact, such physical reservoir computing has attracted increasing attention in diverse fields of research. The purpose of this review is to provide an overview of recent advances in physical reservoir computing by classifying them according to the type of the reservoir. We discuss the current issues and perspectives related to physical reservoir computing, in order to further expand its practical applications and develop next-generation machine learning systems."]]></description>
<dc:subject>to:NB reservoir_computing to_read re:codename:catherine_wheel</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:89a2d8414d87/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:reservoir_computing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:codename:catherine_wheel"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.nature.com/articles/s41467-024-45187-1">
    <title>Emerging opportunities and challenges for the future of reservoir computing | Nature Communications</title>
    <dc:date>2025-08-07T14:44:41+00:00</dc:date>
    <link>https://www.nature.com/articles/s41467-024-45187-1</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Reservoir computing originates in the early 2000s, the core idea being to utilize dynamical systems as reservoirs (nonlinear generalizations of standard bases) to adaptively learn spatiotemporal features and hidden patterns in complex time series. Shown to have the potential of achieving higher-precision prediction in chaotic systems, those pioneering works led to a great amount of interest and follow-ups in the community of nonlinear dynamics and complex systems. To unlock the full capabilities of reservoir computing towards a fast, lightweight, and significantly more interpretable learning framework for temporal dynamical systems, substantially more research is needed. This Perspective intends to elucidate the parallel progress of mathematical theory, algorithm design and experimental realizations of reservoir computing, and identify emerging opportunities as well as existing challenges for large-scale industrial adoption of reservoir computing, together with a few ideas and viewpoints on how some of those challenges might be resolved with joint efforts by academic and industrial researchers across multiple disciplines."]]></description>
<dc:subject>to:NB reservoir_computing re:codename:catherine_wheel to_read prediction dynamical_systems</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:951abc62d069/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:reservoir_computing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:codename:catherine_wheel"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:prediction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:dynamical_systems"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://academic.oup.com/sf/advance-article/doi/10.1093/sf/soaf079/8171706?login=false">
    <title>Competing social influence in contested diffusion: contention and the spread of the early reformation1 | Social Forces | Oxford Academic</title>
    <dc:date>2025-08-06T18:02:45+00:00</dc:date>
    <link>https://academic.oup.com/sf/advance-article/doi/10.1093/sf/soaf079/8171706?login=false</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The spread of radical institutional change does not often result from one-sided pro-innovation influence; countervailing influence networks in support of the status quo can suppress adoption. We develop a model of multiplex and competing network diffusion to describe how competing actors compete through multiple types of networks. Specifically, we hypothesize three types of contested diffusion: market competition, inoculation, and firefighting. To apply the contested-diffusion model to real data, we look at the contest between Martin Luther and Desiderius Erasmus, the two most influential intellectuals of early 16th-century Europe. In the early phase of the Reformation, these two figures utilized influence networks, affecting which cities in the Holy Roman Empire adopted reform. Using newly digitalized data on both leaders’ correspondence networks, their travels, the dispersion of their followers, and parallel processes of exchange among places through trade routes, we employ empirical tests of our theoretical model. We find that although Luther’s network is strongly associated with the spread of the Reformation, Erasmus’s network is associated with the stifling of the Reformation. This is consistent with a “firefighting” mechanism of contested diffusion, whereby the countervailing force suppresses innovations only after they have begun to spread."]]></description>
<dc:subject>to:NB early_modern_european_history social_networks diffusion_of_innovations sociology luther.martin erasmus.desiderius to_read re:do-institutions-evolve</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:fa4788783ab3/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:early_modern_european_history"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:diffusion_of_innovations"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:sociology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:luther.martin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:erasmus.desiderius"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:do-institutions-evolve"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2406.01382">
    <title>[2406.01382] Do Large Language Models Perform the Way People Expect? Measuring the Human Generalization Function</title>
    <dc:date>2025-08-05T15:15:26+00:00</dc:date>
    <link>https://arxiv.org/abs/2406.01382</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["What makes large language models (LLMs) impressive is also what makes them hard to evaluate: their diversity of uses. To evaluate these models, we must understand the purposes they will be used for. We consider a setting where these deployment decisions are made by people, and in particular, people's beliefs about where an LLM will perform well. We model such beliefs as the consequence of a human generalization function: having seen what an LLM gets right or wrong, people generalize to where else it might succeed. We collect a dataset of 19K examples of how humans make generalizations across 79 tasks from the MMLU and BIG-Bench benchmarks. We show that the human generalization function can be predicted using NLP methods: people have consistent structured ways to generalize. We then evaluate LLM alignment with the human generalization function. Our results show that -- especially for cases where the cost of mistakes is high -- more capable models (e.g. GPT-4) can do worse on the instances people choose to use them for, exactly because they are not aligned with the human generalization function."]]></description>
<dc:subject>to_read large_language_models_(so_called) human-computer_interaction vafa.keyon mullainathan.sendhil in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:7c8ec70f8c6b/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:human-computer_interaction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:vafa.keyon"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:mullainathan.sendhil"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://philsci-archive.pitt.edu/26062/">
    <title>Fairness and Signaling in Bargaining Games - PhilSci-Archive</title>
    <dc:date>2025-08-05T13:26:58+00:00</dc:date>
    <link>https://philsci-archive.pitt.edu/26062/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Cultural evolutionary models of bargaining can elucidate issues related to fairness and justice, and especially how fair and unfair conventions and norms might arise in human societies. One line of this research shows how the presence of social categories in such models creates inequitable equilibria that are not possible in models without social categories. This is taken to help explain why in human groups with social categories, inequity is the rule rather than the exception. But in previous models, it is typically assumed that these categories are rigid---in the sense that they cannot be altered, and easily observable---in the sense that all agents can identify each others' category membership. In reality, social categories are not always so tidy. We introduce evolutionary models where the tags connected with social categories can be flexible, variable, or difficult to observe, i.e., where these tags can carry different amounts of information about group membership. We show how alterations to these tags can undermine the stability of unfair conventions. We argue that these results can inform projects intended to ameliorate inequity, especially projects that seek to alter the properties of tags by promoting experimentation, imitation, and play with identity markers."]]></description>
<dc:subject>to:NB to_read evolutionary_game_theory inequality to_teach:statistics_of_inequality_and_discrimination re:do-institutions-evolve o'connor.cailin</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e203827a0dd2/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:evolutionary_game_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:inequality"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:statistics_of_inequality_and_discrimination"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:do-institutions-evolve"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:o'connor.cailin"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.jair.org/index.php/jair/article/view/10640">
    <title>From Frequency to Meaning: Vector Space Models of Semantics | Journal of Artificial Intelligence Research (2010)</title>
    <dc:date>2025-08-05T13:21:08+00:00</dc:date>
    <link>https://www.jair.org/index.php/jair/article/view/10640</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Computers understand very little of the meaning of human language. This profoundly limits our ability to give instructions to computers, the ability of computers to explain their actions to us, and the ability of computers to analyse and process text. Vector space models (VSMs) of semantics are beginning to address these limits. This paper surveys the use of VSMs for semantic processing of text. We organize the literature on VSMs according to the structure of the matrix in a VSM. There are currently three broad classes of VSMs, based on term-document, word-context, and pair-pattern matrices, yielding three classes of applications. We survey a broad range of applications in these three categories and we take a detailed look at a specific open source project in each category. Our goal in this survey is to show the breadth of applications of VSMs for semantics, to provide a new perspective on VSMs for those who are already familiar with the area, and to provide pointers into the literature for those who are less familiar with the field."]]></description>
<dc:subject>to:NB natural_language_processing to_read turney.peter_d.</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a4a6080e4e01/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:natural_language_processing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:turney.peter_d."/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2507.06268">
    <title>[2507.06268] A Collectivist, Economic Perspective on AI</title>
    <dc:date>2025-08-05T13:18:46+00:00</dc:date>
    <link>https://arxiv.org/abs/2507.06268</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Information technology is in the midst of a revolution in which omnipresent data collection and machine learning are impacting the human world as never before. The word "intelligence" is being used as a North Star for the development of this technology, with human cognition viewed as a baseline. This view neglects the fact that humans are social animals, and that much of our intelligence is social and cultural in origin. A related issue is that the current view treats the social consequences of technology as an afterthought. The path forward is not merely more data and compute, and not merely more attention paid to cognitive or symbolic representations, but a thorough blending of economic and social concepts with computational and inferential concepts, in the service of system-level designs in which social welfare is a first-class citizen, and with the aspiration that a new human-centric engineering field will emerge."

--- Without reading beyond the abstract (yet), I will just repeat that if we'd settled on "complex information processing" instead of "artificial intelligence", we'd be much less confused.]]></description>
<dc:subject>to:NB to_read jordan.michael_i. artificial_intelligence machine_learning mechanism_design collaborative_filtering institutions collective_cognition</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:9ecd3e67d562/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:jordan.michael_i."/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:artificial_intelligence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:machine_learning"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:mechanism_design"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:collaborative_filtering"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:institutions"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:collective_cognition"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.sciencedirect.com/science/article/abs/pii/S0065245808604180">
    <title>Speculations Concerning the First Ultraintelligent Machine - ScienceDirect</title>
    <dc:date>2025-08-05T13:07:54+00:00</dc:date>
    <link>https://www.sciencedirect.com/science/article/abs/pii/S0065245808604180</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Publisher Summary: An ultra-intelligent machine is a machine that can far surpass all the intellectual activities of any man however clever. The design of machines is one of these intellectual activities; therefore, an ultra-intelligent machine could design even better machines. To design an ultra-intelligent machine one needs to understand more about the human brain or human thought or both. The physical representation of both meaning and recall, in the human brain, can be to some extent understood in terms of a subassembly theory, this being a modification of Hebb's cell assembly theory. The subassembly theory sheds light on the physical embodiment of memory and meaning, and there can be little doubt that both needs embodiment in an ultra-intelligent machine. The subassembly theory leads to reasonable and interesting explanations of a variety of psychological effects."]]></description>
<dc:subject>to:NB to_read artificial_intelligence good.i.j. via:mraginsky</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:7b869735df81/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:artificial_intelligence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:good.i.j."/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.cambridge.org/core/journals/american-political-science-review/article/social-media-social-control-and-the-politics-of-public-shaming/2BC3349DF48F25D83ADD3271FF2FCEB6">
    <title>Social Media, Social Control, and the Politics of Public Shaming | American Political Science Review | Cambridge Core</title>
    <dc:date>2025-08-05T13:01:35+00:00</dc:date>
    <link>https://www.cambridge.org/core/journals/american-political-science-review/article/social-media-social-control-and-the-politics-of-public-shaming/2BC3349DF48F25D83ADD3271FF2FCEB6</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["While there is disagreement over the value of public shaming, scholars largely agree that social media introduce pathologies. But while scholars rightly identify the effects of online public shaming (OPS), they misidentify the cause. Rather than solely a problem of scale, OPS’s effects are also shaped by the network structure within which they take place. In this article, I argue that the social conditions necessary for productive public shaming are more likely to obtain in a closed social network structure. Using the cases of Twitter, Wikipedia, and Reddit, I show how the design of social media platforms facilitates different network structures among users, with differing results for OPS. In evaluating OPS by way of network structure, I argue, we can not only better understand why OPS works productively in some cases and not in others, but also derive lessons for how to deploy, discuss, and respond to it more effectively."

]]></description>
<dc:subject>to:NB to_read networked_life social_media shame sociology institutions</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e46bacd32165/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networked_life"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_media"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:shame"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:sociology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:institutions"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.stat.berkeley.edu/~brill/Papers/jtsa2012.pdf">
    <title>The Nicholson blowfly experiments: Some history and EDA</title>
    <dc:date>2025-07-29T14:02:16+00:00</dc:date>
    <link>https://www.stat.berkeley.edu/~brill/Papers/jtsa2012.pdf</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The renowned Australian entomologist Alexander J. Nicholson carried out a series of experiments in the 1950s with
the intent of learning more about a sheep pest, the blowfly. The results presented here are driven by analyses of
the data that Nicholson collected. The situation is of special interest because it involves a system that is nonlinear,
has time lags and might be described as non-stationary. There are other complicating aspects including that: the
data are aggregate referring to a sum of interacting cohorts, age effects exist, the data are measured at discrete
times yet the phenomenon exists in continuous time and a structural change may be taking place. In the work, the
spectrogram and complex demodulation prove to be useful tools since the phenomenon is varying, depending on
both time and period (or frequency). These tools have in common the notion of an evolutionary spectrum. The
goals are to explore some of Nicholson’s data and to illustrate how the tools of complex demodulation and the
spectrogram and subject matter can elicit information from time-series data."]]></description>
<dc:subject>to:NB to_read time_series to_teach:data_over_space_and_time re:codename:catherine_wheel brillinger.david_r.</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e5a367e433cf/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:time_series"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data_over_space_and_time"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:codename:catherine_wheel"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:brillinger.david_r."/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://jid.journals.yorku.ca/index.php/jid/article/view/40591">
    <title>Axioms and Intuitions about Societal Inequality : What does the Gini Coefficient Measure? | Journal of Income Distribution®</title>
    <dc:date>2025-07-29T13:49:21+00:00</dc:date>
    <link>https://jid.journals.yorku.ca/index.php/jid/article/view/40591</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We show that Corrado Gini’s “concentration ratio” is an informative measure of experienced inequality that (as he pointed out)  varies from one (his “maximum concentration”) to zero (“minimum concentration”), a feature that does not hold (except in infinite populations) for the measure advocated in the contribution to this symposium by our colleague, Debraj Ray.  Through a social network representation of inequality and  a series of examples, we clarify the differing intuitions about the nature of inequality that alternative measures of inequality capture."]]></description>
<dc:subject>to:NB to_read inequality social_measurement economics bowles.samuel carlin.wendy to_teach:statistics_of_inequality_and_discrimination</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:7563ea1351cb/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:inequality"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_measurement"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:economics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bowles.samuel"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:carlin.wendy"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:statistics_of_inequality_and_discrimination"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.jstor.org/stable/4123261">
    <title>The Market for Quacks on JSTOR</title>
    <dc:date>2025-07-28T14:15:34+00:00</dc:date>
    <link>https://www.jstor.org/stable/4123261</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["A group of n "quacks" plays a price-competition game, facing a continuum of "patients" who recover with probability a, whether they acquire a quack's "treatment". If patients chose rationally, the market would be inactive. I assume, however, that patients choose according to a boundedly rational procedure, which reflects "anecdotal" reasoning. This element of bounded rationality has significant implications. The market for quacks is active, and patients suffer a welfare loss which behaves non-monotonically w.r.t. n and a. In an extended model that endogenizes the quacks' choice of "treatments", the quacks minimize the force of price competition by offering maximally differentiated treatments. The patients' welfare loss is robust to market interventions, which would crowd out low-quality firms in standard models. Thus, as long as the patients' quality of reasoning is not lifted above the anecdotal level, ordinary competition policies may be ineffective."

--- One presumes this applies to educational policy, management consulting, etc., etc.]]></description>
<dc:subject>to:NB to_read psychoceramics market_failures_in_everything economics via:?</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ef22bfe65165/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:psychoceramics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:market_failures_in_everything"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:economics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:?"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="http://computationalculture.net/situating-bayesian-knowledge/">
    <title>Situating Bayesian Knowledge: A Case Study of Modelling Pollutant Transfers from Land to Water – Computational Culture</title>
    <dc:date>2025-07-17T14:36:14+00:00</dc:date>
    <link>http://computationalculture.net/situating-bayesian-knowledge/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Bayesian statistics is an alternative to classical, frequentist statistics. Some have argued that the Bayesian framework embodies a necessarily ‘subjective’ perspective which accounts for the context dependence and relativity of knowledge and contrasts it with a frequentist approach. We argue that such epistemological discussions can actually obscure all the ways in which Bayesian knowledge is partial and, in this way, similar to frequentist knowledge. In this contribution, we explore the contingency and performativity of knowing in Bayesian ways by revisiting an application of Bayesian modelling in a case of pollutant transfers from land to water. We query this material from an STS perspective, thinking through a concrete Bayesian modelling process, the various choices made and their alternatives. We ponder how specific practices that play out in Bayesian modelling–model building, data preparation, setting the prior, defining the likelihood function, sampling from the posterior, and checking the model–produce knowledges that can be situated within and produce, for example, partial perspectives on the issue in question, on knowledge and the good, and within social and material contexts. We touch upon discussions on mathematical affordances of Bayesian modelling such as the lack of a built-in mechanism for updating the space of models. Ultimately, we discuss how Bayesian modelling practices enact aspects of the world, including ‘natural’, ‘social’, ‘political’ and ‘ethical’ ‘objects’, and can (re)configure (social) relations. We demonstrate the value of collaborating with actors that can unsettle the Bayesian workflow to iteratively preserve onto-epistemic openings."

--- I may need AEO to help me understand this...]]></description>
<dc:subject>to:NB to_read to_read_maybe the_french_disease bayesianism hydrology spatio-temporal_statistics re:phil-of-bayes_paper</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e7e6be4024ca/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read_maybe"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:the_french_disease"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bayesianism"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:hydrology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:spatio-temporal_statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:phil-of-bayes_paper"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://aigi.ox.ac.uk/publications/chain-of-thought-is-not-explainability/">
    <title>Chain-of-Thought Is Not Explainability - Oxford Martin AIGI</title>
    <dc:date>2025-07-17T11:57:13+00:00</dc:date>
    <link>https://aigi.ox.ac.uk/publications/chain-of-thought-is-not-explainability/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Chains-of-thought (CoT) allow language models to verbalise multi-step rationales before producing their final answer. While this technique often boosts task performance and offers an impression of transparency into the model’s reasoning, we argue that rationales generated by current CoT techniques can be misleading and are neither necessary nor sufficient for trustworthy interpretability.
"By analysing faith-fulness in terms of whether CoTs are not only human-interpretable, but also reflect underlying model reasoning in a way that supports responsible use, we synthesise evidence from previous studies. We show that verbalised chains are frequently unfaithful, diverging from the true hidden computations that drive a model’s predictions, and giving an incorrect picture of how models arrive at conclusions. Despite this, CoT is increasingly relied upon in high-stakes domains such as medicine, law, and autonomous systems—our analysis of 1,000 recent CoT-centric papers find that ~25% explicitly treat CoT as an interpretability technique—and among them, papers in high-stakes domains specifically hinge on such interpretability claim heavily.
"Building on prior work in interpretability, we make three proposals:(i) avoid treating CoT as being sufficient for interpretability without additional verification while continuing to use CoT for its communicative benefits (ii) adopt rigorous methods that assess faithfulness for downstream decision-making, and(iii) develop causal validation methods (e.g., activation patching, counterfactual interventions, verified models) to ground explanations in model internals."

--- OOH, there's an sense in which an LLM is simply incapable of hiding what it's doing, because there's no internal workspace, nothing _really_ latent, just the stream of tokens.  So you are in fact seeing every step it goes through as it comes up with its output.  OTOH, from a brief look at the paper, they're talking about phenomena like changing the order of options in a multiple-choice question changing the answers, or parts of the prompt influencing the answer but not being referenced by the chain-of-thought.  So the issue is something like "does the output of the text generator, run in verbose mode, accurately describe what the text generator is doing?" and the answer is "not really".]]></description>
<dc:subject>large_language_models_(so_called) to_read via:absfac chain-of-thought_(so_called) in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:59d6c1ea2c6a/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:absfac"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:chain-of-thought_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.tandfonline.com/doi/full/10.1080/26939169.2025.2497547">
    <title>Full article: Developing Students’ Statistical Expertise Through Writing in the Age of AI</title>
    <dc:date>2025-07-08T15:34:52+00:00</dc:date>
    <link>https://www.tandfonline.com/doi/full/10.1080/26939169.2025.2497547</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["As large language models (LLMs) such as GPT have become more accessible, concerns about their potential effects on students’ learning have grown. In data science education, the specter of students’ turning to LLMs raises multiple issues, as writing is a means not just of conveying information but of developing their statistical reasoning. In our study, we engage with questions surrounding LLMs and their pedagogical impact by: (a) quantitatively and qualitatively describing how select LLMs write report introductions and complete data analysis reports; and (b) comparing patterns in texts authored by LLMs to those authored by students and by published researchers. Our results show distinct differences between machine-generated and human-generated writing, as well as between novice and expert writing. Those differences are evident in how writers manage information, modulate confidence, signal importance, and report statistics. The findings can help inform classroom instruction, whether that instruction is aimed at dissuading the use of LLMs or at guiding their use as a productivity tool. It also has implications for students’ development as statistical thinkers and writers. What happens when they offload the work of data science to a model that doesn’t write quite like a data scientist? Supplementary materials for this article are available online."]]></description>
<dc:subject>to:NB to_read teaching large_language_models_(so_called) kith_and_kin reinhart.alex weinberg.gordon</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a2f8d1b7308c/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:teaching"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kith_and_kin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:reinhart.alex"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:weinberg.gordon"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2412.20292">
    <title>[2412.20292] An analytic theory of creativity in convolutional diffusion models</title>
    <dc:date>2025-07-01T15:05:41+00:00</dc:date>
    <link>https://arxiv.org/abs/2412.20292</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We obtain an analytic, interpretable and predictive theory of creativity in convolutional diffusion models. Indeed, score-matching diffusion models can generate highly original images that lie far from their training data. However, optimal score-matching theory suggests that these models should only be able to produce memorized training examples. To reconcile this theory-experiment gap, we identify two simple inductive biases, locality and equivariance, that: (1) induce a form of combinatorial creativity by preventing optimal score-matching; (2) result in fully analytic, completely mechanistically interpretable, local score (LS) and equivariant local score (ELS) machines that, (3) after calibrating a single time-dependent hyperparameter can quantitatively predict the outputs of trained convolution only diffusion models (like ResNets and UNets) with high accuracy (median r2 of 0.95,0.94,0.94,0.96 for our top model on CIFAR10, FashionMNIST, MNIST, and CelebA). Our model reveals a locally consistent patch mosaic mechanism of creativity, in which diffusion models create exponentially many novel images by mixing and matching different local training set patches at different scales and image locations. Our theory also partially predicts the outputs of pre-trained self-attention enabled UNets (median r2∼0.77 on CIFAR10), revealing an intriguing role for attention in carving out semantic coherence from local patch mosaics."

--- Via multiple referers, including rvenkat, whose comments at [https://pinboard.in/u:rvenkat/b:70e511b7e535] I will cache here for ease of my reference later:

"+ using equivariance and symmetries to generate novel images is not exactly explaining aesthetic concept of creativity
"+ image grammars and dictionary models tend to do the same
"+ uttering words like morphogenesis and turing patterns as a quanta magazine journalist may make your article fancy, but do you have to?"]]></description>
<dc:subject>to:NB to_read generative_diffusion_models neural_networks symmetry to_teach:statistics_and_generative_ai via:multiple</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e2510f893f3f/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:generative_diffusion_models"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neural_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:symmetry"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:statistics_and_generative_ai"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:multiple"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.cambridge.org/core/journals/philosophy-of-science/article/demarcating-defining-and-diagnosing-pseudoscience/78BE2C62ABD782B7572154AAFF9F7D13?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles">
    <title>Demarcating, defining, and diagnosing pseudoscience | Philosophy of Science | Cambridge Core</title>
    <dc:date>2025-06-26T14:39:56+00:00</dc:date>
    <link>https://www.cambridge.org/core/journals/philosophy-of-science/article/demarcating-defining-and-diagnosing-pseudoscience/78BE2C62ABD782B7572154AAFF9F7D13?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Karl Popper introduced a metaphor of demarcation for identification of claims that should not be classified as scientific. This metaphor still dominates the philosophical discussion on pseudoscience. We show that it has hampered the discussion in several ways, most importantly by blocking the insight that determining whether some particular claim is pseudoscientific usually requires specialized scientific expertise. We conclude that it would be better to give up this metaphor and leave room for the two tasks of defining pseudoscience (a task for philosophers) and diagnosing potential cases of pseudoscience (a task for experts in the respective areas of science)."]]></description>
<dc:subject>to:NB to_read philosophy_of_science pseudoscience re:on_cranks</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:2fd81041edaf/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:philosophy_of_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:pseudoscience"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:on_cranks"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2505.17120">
    <title>[2505.17120] Self-Interpretability: LLMs Can Describe Complex Internal Processes that Drive Their Decisions, and Improve with Training</title>
    <dc:date>2025-06-16T22:29:29+00:00</dc:date>
    <link>https://arxiv.org/abs/2505.17120</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We have only limited understanding of how and why large language models (LLMs) respond in the ways that they do. Their neural networks have proven challenging to interpret, and we are only beginning to tease out the function of individual neurons and circuits within them. However, another path to understanding these systems is to investigate and develop their capacity to introspect and explain their own functioning. Here, we show that i) contemporary LLMs are capable of providing accurate, quantitative descriptions of their own internal processes during certain kinds of decision-making, ii) that it is possible to improve these capabilities through training, and iii) that this training generalizes to at least some degree. To do so, we fine-tuned GPT-4o and GPT-4o-mini to make decisions in a wide variety of complex contexts (e.g., choosing between condos, loans, vacations, etc.) according to randomly-generated, quantitative preferences about how to weigh different attributes during decision-making (e.g., the relative importance of natural light versus quiet surroundings for condos). We demonstrate that the LLMs can accurately report these preferences (i.e., the weights that they learned to give to different attributes during decision-making). Next, we demonstrate that these LLMs can be fine-tuned to explain their decision-making even more accurately. Finally, we demonstrate that this training generalizes: It improves the ability of the models to accurately explain what they are doing as they make other complex decisions, not just decisions they have learned to make via fine-tuning. This work is a step towards training LLMs to accurately and broadly report on their own internal processes -- a possibility that would yield substantial benefits for interpretability, control, and safety."

--- So many, many questions begged here...]]></description>
<dc:subject>to_read color_me_skeptical large_language_models_(so_called) decision-making in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:c057bde2bf0e/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:color_me_skeptical"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:decision-making"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.sciencedirect.com/science/article/pii/S1090513812001237">
    <title>Infant and child death in the human environment of evolutionary adaptation - ScienceDirect</title>
    <dc:date>2025-06-16T20:59:25+00:00</dc:date>
    <link>https://www.sciencedirect.com/science/article/pii/S1090513812001237</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The precise quantitative nature of the Environment of Evolutionary Adaptedness (EEA) is difficult to reconstruct. The EEA represents a multitude of different geographic and temporal environments, of which a large number often need to be surveyed in order to draw sound conclusions. We examine a large number of both hunter–gatherer (N = 20) and historical (N = 43) infant and child mortality rates to generate a reliable quantitative estimate of their levels in the EEA. Using data drawn from a wide range of geographic locations, cultures, and times, we estimate that approximately 27% of infants failed to survive their first year of life, while approximately 47.5% of children failed to survive to puberty across in the EEA. These rates represent a serious selective pressure faced by humanity that may be underappreciated by many evolutionary psychologists. Additionally, a cross-species comparison found that human child mortality rates are roughly equivalent to Old World monkeys, higher than orangutan or bonobo rates and potentially higher than those of chimpanzees and gorillas. These findings are briefly discussed in relation to life history theory and evolved adaptations designed to lower high childhood mortality."

--- I don't care about the evol. psych. bit here, but I do want to give The Kids in inequality a bit of a sense of what things used to be like...
--- (The arithmetic of patriarchy: If ~1/2 of births fail to survive to reach puberty, getting 2 kids per woman to puberty means 4 births; in fact you'd need more because what matters is getting _through_ puberty to reproduction.  So that's a minimum of something like 5 births per woman just to maintain a stable population, each of which will consumes several years for pregnancy and breast-feeding; at least 10 years and maybe more like 15.  But this starts tracking at birth, so it doesn't account for pregnancies that don't lead to births.  On average, then, with even a moderate rate of stillbirths, women would have to 15--20 years pregnant or caring for newborns, just to avoid population shrinkage.  That's an average across all women, so if some don't have children, for whatever reason, the rest have to have more to make up for it, adding to the time demand on them.  Reduce that death rate and things look very different.)]]></description>
<dc:subject>to:NB to_read demography anthropology to_teach:statistics_of_inequality_and_discrimination the_nightmare_from_which_we_are_trying_to_awake</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:033c141061f8/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:demography"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:anthropology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:statistics_of_inequality_and_discrimination"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:the_nightmare_from_which_we_are_trying_to_awake"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00415/107615/PAQ-65-Million-Probably-Asked-Questions-and-What">
    <title>PAQ: 65 Million Probably-Asked Questions and What You Can Do With Them | Transactions of the Association for Computational Linguistics | MIT Press</title>
    <dc:date>2025-06-15T16:08:49+00:00</dc:date>
    <link>https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00415/107615/PAQ-65-Million-Probably-Asked-Questions-and-What</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Open-domain Question Answering models that directly leverage question-answer (QA) pairs, such as closed-book QA (CBQA) models and QA-pair retrievers, show promise in terms of speed and memory compared with conventional models which retrieve and read from text corpora. QA-pair retrievers also offer interpretable answers, a high degree of control, and are trivial to update at test time with new knowledge. However, these models fall short of the accuracy of retrieve-and-read systems, as substantially less knowledge is covered by the available QA-pairs relative to text corpora like Wikipedia. To facilitate improved QA-pair models, we introduce Probably Asked Questions (PAQ), a very large resource of 65M automatically generated QA-pairs. We introduce a new QA-pair retriever, RePAQ, to complement PAQ. We find that PAQ preempts and caches test questions, enabling RePAQ to match the accuracy of recent retrieve-and-read models, whilst being significantly faster. Using PAQ, we train CBQA models which outperform comparable baselines by 5%, but trail RePAQ by over 15%, indicating the effectiveness of explicit retrieval. RePAQ can be configured for size (under 500MB) or speed (over 1K questions per second) while retaining high accuracy. Lastly, we demonstrate RePAQ’s strength at selective QA, abstaining from answering when it is likely to be incorrect. This enables RePAQ to “back-off” to a more expensive state-of-the-art model, leading to a combined system which is both more accurate and 2x faster than the state-of-the-art model alone."]]></description>
<dc:subject>to:NB information_retrieval nearest_neighbors to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:aea657f20142/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:information_retrieval"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:nearest_neighbors"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
</rdf:RDF>