<?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://thesphinxblog.com/2026/08/11/take-a-chance-on-me/"/>
	<rdf:li rdf:resource="https://www.science.org/doi/10.1126/science.aef8874"/>
	<rdf:li rdf:resource="https://en.wikipedia.org/wiki/Sofic_group"/>
	<rdf:li rdf:resource="https://dl.acm.org/doi/10.1145/359131.359136"/>
	<rdf:li rdf:resource="https://sociologicalscience.com/articles-v13-21-528/"/>
	<rdf:li rdf:resource="https://sociologicalscience.com/articles-v13-24-614/"/>
	<rdf:li rdf:resource="https://sociologicalscience.com/articles-v13-32-825/"/>
	<rdf:li rdf:resource="https://sociologicalscience.com/articles-v13-33-864/"/>
	<rdf:li rdf:resource="https://www.nature.com/articles/s41586-026-10953-2"/>
	<rdf:li rdf:resource="https://www.science.org/doi/full/10.1126/sciadv.adz6502"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2607.17397"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2604.16653"/>
	<rdf:li rdf:resource="https://www.aeaweb.org/articles?id=10.1257%2Faer.20231468"/>
	<rdf:li rdf:resource="https://www.tandfonline.com/doi/full/10.1080/01916599.2026.2697365"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2607.06145"/>
	<rdf:li rdf:resource="https://www.annualreviews.org/content/journals/10.1146/annurev-genom-091416-035340"/>
	<rdf:li rdf:resource="https://journals.aps.org/prx/abstract/10.1103/tmkr-9kl2"/>
	<rdf:li rdf:resource="https://hdsr.mitpress.mit.edu/pub/op7v0ofq/release/1?readingCollection=78528461"/>
	<rdf:li rdf:resource="https://www.quantamagazine.org/is-ai-reasoning-right-for-the-wrong-reasons-20260731/"/>
	<rdf:li rdf:resource="https://en.wikipedia.org/wiki/Congressional_Airport"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2605.09294"/>
	<rdf:li rdf:resource="https://sociologicalscience.com/articles-v13-31-802/"/>
	<rdf:li rdf:resource="https://link.springer.com/article/10.1007/s13171-026-00452-x"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2608.01326"/>
	<rdf:li rdf:resource="https://www.cambridge.org/core/journals/behavioral-and-brain-sciences/article/subjective-selection-superattractors-and-the-origins-of-the-cultural-manifold/2D71AA44CFDB34F171A3D2E2F4410173"/>
	<rdf:li rdf:resource="https://iai.tv/articles/puzzles-reveal-the-limits-of-ai-auid-3638"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2411.10939"/>
	<rdf:li rdf:resource="https://www.nature.com/articles/s41593-025-02031-z"/>
	<rdf:li rdf:resource="https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013130"/>
	<rdf:li rdf:resource="https://www.jstor.org/content/oa_book_monograph/j.ctt1djmg7w"/>
	<rdf:li rdf:resource="https://users.monash.edu.au/~nwormald/papers/de.pdf"/>
	<rdf:li rdf:resource="https://doi.org/10.1214/aoap/1177004612"/>
	<rdf:li rdf:resource="https://doi.org/10.1017/psa.2026.10267"/>
	<rdf:li rdf:resource="https://doi.org/10.1017/psa.2026.10268"/>
	<rdf:li rdf:resource="https://doi.org/10.1017/psa.2026.10266"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2607.14421"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2306.06529"/>
	<rdf:li rdf:resource="https://www.molbiolcell.org/doi/10.1091/mbc.E19-02-0107"/>
	<rdf:li rdf:resource="https://www.santafe.edu/news-center/news/in-memoriam-peter-schuster"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2607.20068"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2607.20751"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2607.21916"/>
	<rdf:li rdf:resource="https://www.annualreviews.org/content/journals/10.1146/annurev-physchem-082423-010652"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2607.16035"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2605.26711"/>
	<rdf:li rdf:resource="https://doi.org/10.1145/3759429.3762631"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2602.04770"/>
	<rdf:li rdf:resource="https://inpreparation.substack.com/p/opinion-i-was-not-allowed-to-type"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2607.01521"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2603.12277"/>
	<rdf:li rdf:resource="https://role-confusion.github.io/"/>
	<rdf:li rdf:resource="https://philpapers.org/archive/POIWDH.pdf"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2606.22748"/>
	<rdf:li rdf:resource="https://link.springer.com/article/10.1007/s10955-025-03550-6"/>
	<rdf:li rdf:resource="https://link.springer.com/article/10.1007/s10955-026-03647-6"/>
	<rdf:li rdf:resource="https://www.cambridge.org/core/journals/philosophy-of-science/article/case-for-time-in-causal-dags/FB8A5FA21249300B5250358B21A3E26D?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles"/>
	<rdf:li rdf:resource="https://www.cambridge.org/core/journals/philosophy-of-science/article/noncausal-explanations-of-social-and-biological-networks/59E80990AA755C7CBBD6E543832681C4?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles"/>
	<rdf:li rdf:resource="https://www.annualreviews.org/content/journals/10.1146/annurev-polisci-032624-025000"/>
	<rdf:li rdf:resource="https://direct.mit.edu/books/monograph/6064/Wired-for-WordsThe-Neural-Architecture-of-Language"/>
	<rdf:li rdf:resource="https://www.youtube.com/watch?v=fO9iRDPXvT4"/>
	<rdf:li rdf:resource="https://nathan.rs/posts/gzip-lm/"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2606.13280"/>
	<rdf:li rdf:resource="https://cacm.acm.org/opinion/artificial-intelligence-for-software-engineering-from-probable-to-provable/"/>
	<rdf:li rdf:resource="https://www.trainjazz.com/"/>
	<rdf:li rdf:resource="https://www.reuters.com/commentary/breakingviews/physical-shocks-are-shrinking-power-money-2026-06-04/"/>
	<rdf:li rdf:resource="https://aclanthology.org/2020.cl-2.7/"/>
	<rdf:li rdf:resource="https://cooking.nytimes.com/recipes/780676171-cucumber-and-onion-salad"/>
	<rdf:li rdf:resource="https://link.springer.com/book/10.1007/978-3-031-97239-3"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2601.05444"/>
	<rdf:li rdf:resource="https://www.science.org/doi/10.1126/sciadv.aao3580"/>
      </rdf:Seq>
    </items>
  </channel><item rdf:about="https://thesphinxblog.com/2026/08/11/take-a-chance-on-me/">
    <title>Take A Chance On Me | Sphinx</title>
    <dc:date>2026-08-13T16:21:26+00:00</dc:date>
    <link>https://thesphinxblog.com/2026/08/11/take-a-chance-on-me/</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>scholarship tychetext have_read via:?</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:0b9023bbe51c/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:scholarship"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:tychetext"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:?"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.science.org/doi/10.1126/science.aef8874">
    <title>Recovering signatures of archaic hominin introgression using ancestral recombination graphs | Science</title>
    <dc:date>2026-08-11T15:32:23+00:00</dc:date>
    <link>https://www.science.org/doi/10.1126/science.aef8874</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Admixture between modern humans and extinct hominins has shaped the genomes of present-day individuals, but reconstructing this history has been constrained by the scarcity of archaic samples and unadmixed outgroup populations. We introduce TRACE, a reference- and outgroup-free approach that uses features of ancestral recombination graphs to identify archaic ancestry. Simulations show TRACE has high precision and low false discovery rates. Applied to 1000 Genomes, TRACE recovers known Neanderthal and Denisovan introgression and uncovers ghost admixture from uncharacterized hominins in both Africans and non-Africans. Ghost ancestry persists in Neanderthal and Denisovan ancestry deserts, challenging their interpretation as Homo sapiens–specific regions. In Oceanians, TRACE finds deep lineages are enriched in Denisovan compared to Neanderthal regions, supporting super-archaic introgression. TRACE enables mapping archaic introgression without archaic genomes."

--- Maybe I'm being excessively skeptical, but if your new method proposes _two_ new lineages which are previously totally unknown and we have no evidence for, you really need super convincing proof in the reliability of the method, no?]]></description>
<dc:subject>to:NB historical_genetics human_evolution</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:9e7b6d281105/</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:historical_genetics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:human_evolution"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://en.wikipedia.org/wiki/Sofic_group">
    <title>Sofic group - Wikipedia</title>
    <dc:date>2026-08-11T15:28:58+00:00</dc:date>
    <link>https://en.wikipedia.org/wiki/Sofic_group</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- I'll be honest, I find the definition here incomprehensible, and I used to understand sofic shifts pretty well.]]></description>
<dc:subject>algebra</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:c644022254be/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algebra"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://dl.acm.org/doi/10.1145/359131.359136">
    <title>Algorithm = logic + control | Communications of the ACM</title>
    <dc:date>2026-08-11T15:06:12+00:00</dc:date>
    <link>https://dl.acm.org/doi/10.1145/359131.359136</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The notion that computation = controlled deduction was first proposed by Pay Hayes [19] and more recently by Bibel [2] and Vaughn-Pratt [31]. A similar thesis that database systems should be regarded as consisting of a relational component, which defines the logic of the data, and a control component, which stores and retrieves it, has been successfully argued by Codd [10]. Hewitt's argument [20] for the programming language PLANNER, though generally regarded as an argument against logic, can also be regarded as an argument for the thesis that algorithms be regarded as consisting of both logic and control components. In this paper we shall explore some of the useful consequences of that thesis."]]></description>
<dc:subject>in_NB computation algorithms re:thought_experiments via:? have_read logic</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:292ce720202f/</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:computation"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithms"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:thought_experiments"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:?"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:logic"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://sociologicalscience.com/articles-v13-21-528/">
    <title>How a Seemingly Innocuous and Intuitive Methodological Choice Confused a Generation of Research on Policy Responsiveness</title>
    <dc:date>2026-08-11T13:46:47+00:00</dc:date>
    <link>https://sociologicalscience.com/articles-v13-21-528/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The finding that government policy is, “virtually unrelated to the desires of the low- and middle-income citizens” (Gilens 2005:789), is one of the most influential social science results of the last two decades. This article offers a new perspective on this finding. I show that the seemingly innocuous decision to restrict analyses to data where different income groups’ policy support differs (i.e., a preference gap exists) introduced Simpson’s paradox, leading to misleading conclusions about whose preferences policy reflects. The same concerns apply to analyses of responsiveness to men and women and to partisan groups. I also present evidence that other common approaches for evaluating policy responsiveness can produce equally misleading conclusions. These findings suggest a need to reconsider conventional wisdom about political influence. The conclusion offers methodological recommendations and discusses implications related to understanding social and economic inequality and support for populist candidates."

--- Should really write up that module about inequality in political participation & responsiveness.

--- ETA after reading: the Simpson's Paradox stuff is convincing, but the mind-blowing thing is the correlation in policy preferences between 10th and 90th percentiles is over 0.9!  As he says, this is just _asking_ for trouble with collinearity.
One could do a nonparametric regression here, with policy support at 10th and 90th percentiles as the two regressors, and because it'd only be 2D, the curse of dimensionality wouldn't really apply.  One would not need to explicitly encode a dummy for whether a policy was more supported by the affluent than the poor or not.  But with that much correlation between the two regressors, one would only be estimating the regression function in a narrow strip of the unit square (basically).
--- Might be worth making into an assignment for logistic regression in a regression class, or ADA? (Hence last tags.)]]></description>
<dc:subject>to:NB inequality political_economy to_teach:statistics_of_inequality_and_discrimination have_read social_science_methodology to_teach:linear_models to_teach:undergrad-ADA</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:392c3f36ff71/</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:inequality"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:political_economy"/>
	<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:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_science_methodology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:linear_models"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:undergrad-ADA"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://sociologicalscience.com/articles-v13-24-614/">
    <title>Declining Inequality and Persistent Inequality Structures</title>
    <dc:date>2026-08-11T13:45:45+00:00</dc:date>
    <link>https://sociologicalscience.com/articles-v13-24-614/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Prior research finds that rising labor market inequality in the United States was abetted by structural changes in the economy: a consolidation of occupation and organizational bases of advantage; rising within-job inequality; and declining pay and employment in middle-earning jobs. In this article, we revisit these structural changes by asking whether they have been reversed as labor market inequality fell over the last decade. Drawing on restricted-use microdata from the Occupational Employment and Wages Statistics, we find that declining inequality is due to declining inequality in occupation premiums. There has been only a small reversal of consolidation and no decrease in inequality within jobs. Low-wage jobs gained on shrinking middle-earning occupations, further eroding union, manufacturing, and public sector wage premiums. These findings demonstrate a novel configuration of labor market inequality, in which pay rose in low-wage jobs, but underlying inequality structures in the economy persisted."]]></description>
<dc:subject>to:NB to_teach:statistics_of_inequality_and_discrimination class_struggles_in_america inequality</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:271312f4fc4b/</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_teach:statistics_of_inequality_and_discrimination"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:class_struggles_in_america"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:inequality"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://sociologicalscience.com/articles-v13-32-825/">
    <title>A Roadmap for Inequality Research: Transparency, Intersectionality, and Multiple Measures of Race</title>
    <dc:date>2026-08-11T13:44:34+00:00</dc:date>
    <link>https://sociologicalscience.com/articles-v13-32-825/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Most quantitative studies of U.S. inequality rely on single measures of race and do not transparently describe them. However, inconsistencies between measures can yield conclusions that differ both substantively and statistically. We ask: when faced with multiple ways to categorize respondents, how should researchers choose? We conduct intersectional analyses of five inequality outcomes, using the 1979 National Longitudinal Survey of Youth, which offers several measures of self-identification and external classification. Strikingly, we find the survey’s screener race variable, ubiquitous in prior research, is never empirically preferred based on model fit across outcomes spanning the labor market (wages, salary, and unemployment), health (depression), and education (school discipline). Instead, the top-performing measure varies by gender, outcome, and fit statistic. The range of potential researcher decisions and the absence of a clear gold-standard highlights the need for greater transparency and more thoughtful decision-making when researchers operationalize race—whether racial categorization is central to the analysis or included primarily as a control variable. To that end, we offer a roadmap of key considerations inequality researchers can consult when designing their approach."

--- I'm not sure (without reading it) that "which measured feature gives the best prediction?" is really the right way to address "which is the better measurement of the underlying variable?"  But the basic point about caution and clarity being warranted when getting discordant results from different measures of (purportedly) the same thing is sound.]]></description>
<dc:subject>to:NB social_measurement race to_teach:statistics_of_inequality_and_discrimination</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:acbb92569074/</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:social_measurement"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:race"/>
	<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://sociologicalscience.com/articles-v13-33-864/">
    <title>Information Diets Are More Diverse in Attention Than in Engagement</title>
    <dc:date>2026-08-11T13:40:28+00:00</dc:date>
    <link>https://sociologicalscience.com/articles-v13-33-864/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["What political content do we pay attention to online? Diverse political information is essential for democratic competence, yet online media raises concerns about fragmented information diets. Research on selective exposure highlights how social media can foster ideological echo chambers, while other studies emphasize incidental exposure to diverse viewpoints. A critical measurement challenge is that public platform traces often observe engagement (e.g., likes or shares) more readily than lower-visibility forms of attention—i.e., what users notice or choose to read without necessarily interacting publicly. In this study, we address this gap with a social media clone platform that separately records attention and engagement. We ran the study in both the United Kingdom (once) and the United States (thrice). Across both contexts, we found that the ideology-engagement association is significantly larger than the ideology-attention association. This underscores the importance of measuring attention, rather than solely engagement, to accurately assess the diversity of online information diets."]]></description>
<dc:subject>to:NB social_media polarization preference_falsification</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:5b541a1f8727/</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:social_media"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:polarization"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:preference_falsification"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.nature.com/articles/s41586-026-10953-2">
    <title>Operational Tropical Cyclone Forecasting with AI | Nature</title>
    <dc:date>2026-08-10T18:40:33+00:00</dc:date>
    <link>https://www.nature.com/articles/s41586-026-10953-2</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Tropical cyclones are among the most dangerous and costly weather phenomena, yet forecasting them remains a profound scientific challenge. Here, we introduce WeatherNext Cyclones (WN-C), an AI operational weather model producing state-of-the-art ensemble forecasts for track, intensity, and size of tropical cyclones worldwide. Trained on a combination of global analysis data1 and a global database of historical tropical cyclones2,3, WN-C generates large ensembles of possible global weather and cyclone scenarios extending 15 days into the future. Evaluated on tropical cyclones from 2023–2025, the track, intensity and wind radii predictions from WN-C offer an average of a day or more of lead time advantage over leading operational models, an improvement in accuracy comparable to the progress seen over the last decade of operational development. We achieved these results using inputs orders of magnitude coarser than regional models, suggesting that high resolution is not a strict prerequisite for state-of-the-art intensity forecasting and that this coarser atmospheric data contains more intensity signal than previously recognised. Including predictions from WN-C in a weighted-average consensus ensemble substantially improves its skill. The scalability of WN-C enables up to 1,000-member ensembles which better capture rare events over conventional 50-member ensembles. By providing state-of-the-art operational ensemble guidance to human forecasters, this work represents a step-change towards more reliable and timely forecasts and warnings that can help protect lives and mitigate the devastating impacts of tropical cyclones."]]></description>
<dc:subject>to:NB prediction neural_networks meteorology</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:5512310d2847/</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:neural_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:meteorology"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.science.org/doi/full/10.1126/sciadv.adz6502">
    <title>Untrustworthy sources on Facebook and Instagram in 2020: Concentrated exposure but no attitudinal effects | Science Advances</title>
    <dc:date>2026-08-10T18:39:37+00:00</dc:date>
    <link>https://www.science.org/doi/full/10.1126/sciadv.adz6502</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Despite concern about exposure to content from untrustworthy sources on social media, little is known about the frequency or effects of exposure to their content. We examine 2020 data from all active US adults on Facebook and Instagram to measure exposure to content from Pages, groups, and web domains on Facebook and public accounts on Instagram that repeatedly publish misinformation. We find that average users saw relatively little content in their feeds from these untrustworthy sources; exposure was highly concentrated. A multimonth field experiment during the 2020 election among consenting users reduced feed-based exposure to content from untrustworthy sources by approximately 70% on both platforms but had no measurable effects on numerous preregistered outcomes, even among participants with high pretreatment exposure. Our results demonstrate that a feasible platform intervention can successfully reduce exposure to content from untrustworthy sources but suggest that these changes are unlikely to have immediate effects on attitudes and beliefs."]]></description>
<dc:subject>to:NB social_media epidemiology_of_representations</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:559f8f2ed2e7/</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:social_media"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:epidemiology_of_representations"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2607.17397">
    <title>[2607.17397] The unintended consequences of large language models as a labor-augmenting technology in science</title>
    <dc:date>2026-08-10T18:20:38+00:00</dc:date>
    <link>https://arxiv.org/abs/2607.17397</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["As a labor-augmenting technology, large language models (LLMs) have the potential to accelerate scientific activity across the research pipeline. But even if LLMs perform on par with human experts at selected tasks, their use will bring unintended consequences as they alter the balance of frictions and inducements that steer the allocation of research effort across projects. Here we develop a simple mathematical model to illustrate. In fields where LLMs are useful primarily as tools for discovering promising projects, researchers will become more selective about what they publish; where they facilitate the process of publishing existing data, researchers will become less selective. By allowing scientists to work more quickly, LLMs raise the opportunity cost of researcher time, creating incentives to refine papers less thoroughly before moving on. Enticing as it is to imagine that, by saving us time on mundane tasks, LLMs will provide us with more time to think deeply and develop projects completely, our results temper such hopes."]]></description>
<dc:subject>to:NB large_language_models_(so_called) science_as_a_social_process bergstrom.carl</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:2283b6baa54f/</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:science_as_a_social_process"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bergstrom.carl"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2604.16653">
    <title>[2604.16653] Continuous transformations of probability measures and their transport representations</title>
    <dc:date>2026-08-10T18:20:09+00:00</dc:date>
    <link>https://arxiv.org/abs/2604.16653</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Given a function F transforming a probability measure μ into another one F(μ), we study the existence and regularity of a transport representation of it. That is, we ask whether we can represent the image F(μ) of the input probability measure μ as the push-forward of μ by a map f(⋅,μ) which may depend on μ; and furthermore, how regular f can be chosen depending on F. Even if F is continuous and a transport representative exists, it cannot necessarily be chosen in a continuous way; however, if F is Lipschitz continuous with respect to the Wasserstein distance, then f can be chosen continuous. We provide several examples to illustrate the sharpness of our assumptions. This question is motivated by approximation results for transformations of probability distributions with transformers."]]></description>
<dc:subject>to:NB probability via:mraginsky</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:3dd3c141c6d9/</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:probability"/>
	<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%2Faer.20231468">
    <title>When Product Markets Become Collective Traps: The Case of Social Media - American Economic Association</title>
    <dc:date>2026-08-10T18:18:39+00:00</dc:date>
    <link>https://www.aeaweb.org/articles?id=10.1257%2Faer.20231468</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Individuals might experience negative utility from not consuming a popular product. With such externalities to nonusers, standard consumer surplus measures, which take aggregate consumption as given, fail to appropriately capture consumer welfare. We propose an approach to account for these externalities and apply it to estimate consumer welfare from two social media platforms: TikTok and Instagram. Incentivized experiments with college students indicate positive welfare based on the standard measure but negative welfare when accounting for these nonuser externalities. Our findings highlight the existence of product market traps, where active users of a platform prefer it not to exist."]]></description>
<dc:subject>to:NB collective_action social_media re:actually-dr-internet-is-the-name-of-the-monsters-creator</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:b79be7e6abaa/</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:collective_action"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_media"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:actually-dr-internet-is-the-name-of-the-monsters-creator"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.tandfonline.com/doi/full/10.1080/01916599.2026.2697365">
    <title>The bee and the architect: scientists against scientism, a Transnational European Movement (1976–1978): History of European Ideas: Vol 0, No 0 - Get Access</title>
    <dc:date>2026-08-10T18:17:34+00:00</dc:date>
    <link>https://www.tandfonline.com/doi/full/10.1080/01916599.2026.2697365</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The Radical Science Movement emerged in the 1970s to challenge established scientific practices and the social consequences of technological modernity. This article investigates prominent radical scientists in Italy, Britain, West Germany, and France, such as Marcello Cini, Hilary and Steven Rose, the Max Planck researchers in Starnberg, and Jean-Marc Lévy-Leblond. Inspired by Marx, Kuhn, Marcuse and Heisenberg, radical scientists challenged the notion of scientific neutrality. They denounced the distortion of science by capitalism and advocated for a new science aligned with social goals. Their primary target was social democrats and communists who embraced productivism and mastery over nature and believed that growth of scientific power would lead to socialism. They criticised both the USA and the USSR while expressing admiration for China and Vietnam. Radical scientists revealed that, far from being a liberal profession, scientific work was as alienating and oppressive as any job under capitalism. However, their analysis had many weaknesses, such as being vague in depicting an alternative science, embracing pseudoscience, and exoticising distant countries. These intellectual contributions emerged at a turning point in Western culture, when the science-based optimism about the future and progress began to give way to scepticism about rationality and fear about technological risks."]]></description>
<dc:subject>to:NB history_of_science wisdom_of_the_east</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:14d0a20291aa/</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:history_of_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:wisdom_of_the_east"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2607.06145">
    <title>[2607.06145] Prompting Complexity: Shortest Prompts for Texts and Behaviors in LLMs</title>
    <dc:date>2026-08-10T18:16:28+00:00</dc:date>
    <link>https://arxiv.org/abs/2607.06145</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["In this paper, we define the quantity of prompting complexity: for a fixed instruction-tuned language model, what is the shortest plausible prompt that makes deterministic decoding produce a target text? It is an LM-relative analogue of resource-bounded Kolmogorov complexity: the prompt is a program, the model interface is the interpreter, and information omitted from the prompt is supplied by the model's weights, training distribution, tokenizer, template, and decoding rule. Unlike classical Kolmogorov complexity, this measure is intentionally non-universal. In the finite-context setting it is computable by enumeration, but there is no model-independent invariance theorem; the same text may be cheap for one model and inaccessible or expensive for another. To keep the search space aligned with prompt engineering, we restrict programs to plausible human-readable texts rather than arbitrary token strings. We extend the exact definition to soft prompting complexity for approximate outputs, yielding a lossy notion of model-relative text compression and a formal target for prompt optimization. We also define prompting distance by comparing shortest generating prompts, and behavioral prompting complexity for reaching any output satisfying a specification. Based on these formulations, we define a research agenda for empirically studying which texts and behaviors are accessible from short plausible prompts under a fixed LM interface."

--- Cf. [https://pinboard.in/u:cshalizi/b:0b7cc7c67740]]]></description>
<dc:subject>to:NB large_language_models_(so_called) no_relation</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:3c3d7acc876c/</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:no_relation"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.annualreviews.org/content/journals/10.1146/annurev-genom-091416-035340">
    <title>On the Evolution of Lactase Persistence in Humans | Annual Reviews</title>
    <dc:date>2026-08-10T18:12:47+00:00</dc:date>
    <link>https://www.annualreviews.org/content/journals/10.1146/annurev-genom-091416-035340</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Lactase persistence—the ability of adults to digest the lactose in milk—varies widely in frequency across human populations. This trait represents an adaptation to the domestication of dairying animals and the subsequent consumption of their milk. Five variants are currently known to underlie this phenotype, which is monogenic in Eurasia but mostly polygenic in Africa. Despite being a textbook example of regulatory convergent evolution and gene-culture coevolution, the story of lactase persistence is far from clear: Why are lactase persistence frequencies low in Central Asian herders but high in some African hunter-gatherers? Why was lactase persistence strongly selected for even though milk processing can reduce the amount of lactose? Are there other factors, outside of an advantage of caloric intake, that contributed to the selective pressure for lactase persistence? It is time to revisit what we know and still do not know about lactase persistence in humans."]]></description>
<dc:subject>to:NB historical_genetics</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:6675c66a76b0/</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:historical_genetics"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://journals.aps.org/prx/abstract/10.1103/tmkr-9kl2">
    <title>Limits of Inference in Complex Systems: When Stochastic Models Become Indistinguishable | Phys. Rev. X</title>
    <dc:date>2026-08-10T18:05:50+00:00</dc:date>
    <link>https://journals.aps.org/prx/abstract/10.1103/tmkr-9kl2</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Robust inference for stochastic dynamical systems is often hampered by sparse sampling and the absence of closed-form likelihoods. We introduce a Monte Carlo path-inference framework that leverages full-path statistics and bridge processes to deliver reliable parameter estimation and model selection from coarsely sampled time series, without requiring analytical solutions. Crucially, we couple mechanistic stochastic models with their inference procedures to quantify how experimental design—specifically, sampling frequency and dataset size—governs estimator precision and model distinguishability. This analysis reveals optimal sampling regimes and sharp, resolution-dependent limits beyond which competing models become empirically indistinguishable. We validate the approach across four disparate systems—trajectories of optically trapped particles, human microbiome dynamics, social-media topic mentions, and forest population time series—recovering parameters and identifying when inference is fundamentally constrained by measurement resolution, thereby clarifying ongoing debates about dominant noise sources in these systems. Together, these results establish path-based Monte Carlo as a practical, general tool for inference and model discrimination in complex systems and provide principled guidelines for designing measurements that maximize information under real-world constraints."

--- Curious to see what this amounts to in my terms.]]></description>
<dc:subject>to:NB statistical_inference_for_stochastic_processes statistics</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:c7fa781757f0/</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:statistical_inference_for_stochastic_processes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistics"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://hdsr.mitpress.mit.edu/pub/op7v0ofq/release/1?readingCollection=78528461">
    <title>Shaping Data Science: Statistics and Geometry · Issue 8.3, Summer 2026</title>
    <dc:date>2026-08-10T18:03:51+00:00</dc:date>
    <link>https://hdsr.mitpress.mit.edu/pub/op7v0ofq/release/1?readingCollection=78528461</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>geometry statistics track_down_references</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:f98cde1f69c3/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:geometry"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:track_down_references"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.quantamagazine.org/is-ai-reasoning-right-for-the-wrong-reasons-20260731/">
    <title>Is AI Reasoning Right for the Wrong Reasons? | Quanta Magazine</title>
    <dc:date>2026-08-10T18:03:23+00:00</dc:date>
    <link>https://www.quantamagazine.org/is-ai-reasoning-right-for-the-wrong-reasons-20260731/</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>have_read large_language_models_(so_called)</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ca8402958002/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_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://en.wikipedia.org/wiki/Congressional_Airport">
    <title>Congressional Airport - Wikipedia</title>
    <dc:date>2026-08-10T16:47:55+00:00</dc:date>
    <link>https://en.wikipedia.org/wiki/Congressional_Airport</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- I'd always wondered why that was "Congressional Plaza".  The idea that it could've been the site of DCA, with tons of airplanes approaching over Rockville Pike, is... curious.]]></description>
<dc:subject>have_read maryland via:aeo</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:d3504d0a0947/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:maryland"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:aeo"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2605.09294">
    <title>[2605.09294] Towards Effective Theory of LLMs: A Representation Learning Approach</title>
    <dc:date>2026-08-10T16:43:11+00:00</dc:date>
    <link>https://arxiv.org/abs/2605.09294</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We propose Representational Effective Theory (RET), a framework for describing large language model computation in terms of learned macrostates rather than microscopic details. RET learns these macrostates from hidden-state trajectories using a BYOL/JEPA-style self-supervised objective, coarse-graining activations into macrovariables that preserve higher-level structure relevant for prediction and interpretation. We evaluate whether these macrovariables are practically relevant for interpretability: RET yields temporally consistent states that reveal "mental-state" trajectories of reasoning, capture high-level semantic structure, support early prediction of behavioral outcomes such as sycophancy, and provide causal handles for steering generations toward interpretable computational phases. Together, these results suggest that LLM computation admits useful effective descriptions via RET: high-level, dynamically meaningful variables that support interpretation, prediction, and intervention."

--- Guess I should walk over to Gates and talk to them.]]></description>
<dc:subject>to:NB large_language_models_(so_called) macro_from_micro</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ab773838ea20/</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:macro_from_micro"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://sociologicalscience.com/articles-v13-31-802/">
    <title>Leveraging Genomic Data to Document Within-Race Attractiveness Penalties Among Black Americans</title>
    <dc:date>2026-08-10T15:08:02+00:00</dc:date>
    <link>https://sociologicalscience.com/articles-v13-31-802/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["In recent years, scholars of racial inequality have increasingly sought to move beyond simply quantifying discrete racial disparities and instead measure social stratification as a function of continuous racialized characteristics that vary both within and between racial groups. In this article, we draw on a sample of genotyped respondents from the Add Health study and construct genetic similarity proportions, individual-level measures that correlate with racialized physical features that vary across the expansive family tree of humanity (skin tone, facial structure, hair texture, etc.). We then investigate the relationship between these proportions and interviewer-rated physical attractiveness among self-identified Black Americans (N=2,087). Our findings highlight the existence of substantial attractiveness penalties related to having higher levels of Sub-Saharan African (as opposed to European) genetic similarity."

--- Last tag is really "to mention in the footnotes".
--- Notes after reading: I do not like using linear regression to predict a 5-level Likert scale.  Results for a logistic regression of predicting "Very Attractive" vs. not were similar, but they really should have done an ordinal regression.  Also, they neglect a simple and compelling explanation for why there might be small group differences in rated attractiveness (a finding they confirm) and yet large group disparities in outcomes like dating, viz., competition.  (Well, they _are_ sociologists.)]]></description>
<dc:subject>to:NB racism historical_genetics to_teach:statistics_of_inequality_and_discrimination</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:d069d50e4b75/</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:racism"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:historical_genetics"/>
	<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://link.springer.com/article/10.1007/s13171-026-00452-x">
    <title>Bootstrapping Generalization Error Bounds for Time Series | Sankhya A | Springer Nature Link</title>
    <dc:date>2026-08-10T14:49:53+00:00</dc:date>
    <link>https://link.springer.com/article/10.1007/s13171-026-00452-x</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We consider the problem of constructing confidence intervals for the risk of forecasting the future of a stationary, ergodic stochastic process, using a model estimated from the past of the process. We show that a bootstrap procedure provides valid confidence intervals for the risk, when the data source is sufficiently mixing, and the loss function and the estimator are suitably smooth. Autoregressive (AR(d)) models estimated by least squares obey the necessary regularity conditions, even when mis-specified, and simulations show that the finite-sample coverage of our bounds quickly converges to the theoretical, asymptotic level. As an intermediate step, we derive sufficient conditions for asymptotic independence between empirical distribution functions formed by splitting a realization of a stochastic process, of independent interest."

--- Ungated: [http://arxiv.org/abs/1711.02834]

--- (This isn't my longest gap between a preprint and a journal version, but it's up there.)]]></description>
<dc:subject>in_NB time_series bootstrap prediction data_splitting self-promotion kith_and_kin lunde.robert</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:62d6e1157d0a/</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:time_series"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bootstrap"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:prediction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_splitting"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:self-promotion"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kith_and_kin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:lunde.robert"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2608.01326">
    <title>[2608.01326] Context Compaction Theory</title>
    <dc:date>2026-08-10T14:42:28+00:00</dc:date>
    <link>https://arxiv.org/abs/2608.01326</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Large Language Models (LLMs) have a bounded context window. The context window is the maximum input size an LLM can consume for a single inference. AI agents rely on a process called context compaction to fit their state within the context window when calling an LLM. Despite its ubiquity, context compaction has received essentially no formal analysis. In this paper, we initiate a formal study of context compaction. We first introduce a framework consisting of two games that capture the two algorithmic strategies for context compaction used by contemporary AI agents in practice. The Context Selection Game models context compaction algorithms that select a subset of an agent's accumulated state to retain. The Context Generation Game models context compaction algorithms that summarize an agent's state by an arbitrary message of bounded length. We then prove an equivalence between the Context Generation Game and one-way communication complexity. The minimum context compaction budget for answering a set of queries within a target error is equal to the one-way communication complexity of the induced communication problem at the same error. Known bounds from communication complexity therefore transfer directly to context compaction. We also show that the Context Selection Game corresponds to a restricted class of one-way communication protocols. Any gap between selection and generation is therefore a gap between two classes of communication protocols. We prove that there exists a set of queries for which generation needs strictly less budget than selection. The equivalence between the Context Generation Game and one-way communication also lets us measure how well a deployed context compaction algorithm performs on a query relative to the optimal strategy. As an example, we present a case study that evaluates Anthropic's context compaction endpoint on set membership queries."]]></description>
<dc:subject>to:NB large_language_models_(so_called) information_theory mitzenmacher.michael to_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:199b02f76217/</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:information_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:mitzenmacher.michael"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.cambridge.org/core/journals/behavioral-and-brain-sciences/article/subjective-selection-superattractors-and-the-origins-of-the-cultural-manifold/2D71AA44CFDB34F171A3D2E2F4410173">
    <title>Subjective selection, super-attractors, and the origins of the cultural manifold | Behavioral and Brain Sciences | Cambridge Core</title>
    <dc:date>2026-08-10T14:41:47+00:00</dc:date>
    <link>https://www.cambridge.org/core/journals/behavioral-and-brain-sciences/article/subjective-selection-superattractors-and-the-origins-of-the-cultural-manifold/2D71AA44CFDB34F171A3D2E2F4410173</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Human societies consistently develop complex, near-universal traditions that exhibit striking similarities. These “super-attractors” span magic and religion (e.g., shamanism, supernatural punishment beliefs), esthetics (e.g., heroic tales, dance songs), and social institutions (e.g., justice, corporate groups), and collectively constitute the “cultural manifold.” Here, I argue that the components of the cultural manifold develop primarily through “subjective selection,” or the production and selective retention of variants evaluated as useful for satisfying goals. Whereas previous explanations emphasize objective individual or group-level benefits, I highlight how subjective selection drives complex cultural convergences worldwide, attesting to the importance of subjective selection in shaping human culture."

--- BBS target article, with plenty of commentaries.]]></description>
<dc:subject>to:NB to_read cultural_evolution re:do-institutions-evolve</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:58e7633a7563/</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:re:do-institutions-evolve"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://iai.tv/articles/puzzles-reveal-the-limits-of-ai-auid-3638">
    <title>Puzzles reveal the limits of AI | Cristopher Moore » IAI TV</title>
    <dc:date>2026-08-10T14:40:56+00:00</dc:date>
    <link>https://iai.tv/articles/puzzles-reveal-the-limits-of-ai-auid-3638</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>have_read to:NB artificial_intelligence problem_solving kith_and_kin moore.cristopher</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:0125d9affcb1/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:artificial_intelligence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:problem_solving"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kith_and_kin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:moore.cristopher"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2411.10939">
    <title>[2411.10939] Evaluating Generative AI Systems is a Social Science Measurement Challenge</title>
    <dc:date>2026-08-10T14:40:03+00:00</dc:date>
    <link>https://arxiv.org/abs/2411.10939</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Across academia, industry, and government, there is an increasing awareness that the measurement tasks involved in evaluating generative AI (GenAI) systems are especially difficult. We argue that these measurement tasks are highly reminiscent of measurement tasks found throughout the social sciences. With this in mind, we present a framework, grounded in measurement theory from the social sciences, for measuring concepts related to the capabilities, impacts, opportunities, and risks of GenAI systems. The framework distinguishes between four levels: the background concept, the systematized concept, the measurement instrument(s), and the instance-level measurements themselves. This four-level approach differs from the way measurement is typically done in ML, where researchers and practitioners appear to jump straight from background concepts to measurement instruments, with little to no explicit systematization in between. As well as surfacing assumptions, thereby making it easier to understand exactly what the resulting measurements do and do not mean, this framework has two important implications for evaluating evaluations: First, it can enable stakeholders from different worlds to participate in conceptual debates, broadening the expertise involved in evaluating GenAI systems. Second, it brings rigor to operational debates by offering a set of lenses for interrogating the validity of measurement instruments and their resulting measurements."]]></description>
<dc:subject>to:NB measurement social_measurement large_language_models_(so_called) kith_and_kin jacobs.abigail_z.</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:c39bcf5cb31e/</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:social_measurement"/>
	<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:jacobs.abigail_z."/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.nature.com/articles/s41593-025-02031-z">
    <title>A neural manifold view of the brain | Nature Neuroscience</title>
    <dc:date>2026-08-05T04:06:04+00:00</dc:date>
    <link>https://www.nature.com/articles/s41593-025-02031-z</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Animal behavior arises from the coordinated activity of neural populations that span the entire brain. The activity of large neural populations from an increasing number of brain regions, behaviors and species shows low-dimensional structure. We posit that this structure arises as a result of neural manifolds. Neural manifolds are mathematical descriptions of a meaningful biological entity: the possible collective states of a population of neurons given the constraints, both intrinsic (for example, connectivity) and extrinsic (for example, behavior), to the neural circuit. Here, we explore the link between neural manifolds and behavior, and discuss the insights that the neural manifold framework can provide into brain function. To conclude, we explore existing conceptual gaps in this framework and discuss their implications when building an integrative view of brain function. We thus position neural manifolds as a crucial framework with which to describe how the brain generates behavior."

--- The last tag is madly ambitious.]]></description>
<dc:subject>to:NB manifold_learning neuroscience neural_coding_and_decoding re:codename:catherine_wheel</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:03bb6504d5c8/</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:manifold_learning"/>
	<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:re:codename:catherine_wheel"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013130">
    <title>Information theoretic measures of neural and behavioural coupling predict representational drift | PLOS Computational Biology</title>
    <dc:date>2026-08-05T04:04:19+00:00</dc:date>
    <link>https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013130</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["In many parts of the brain, population tuning to stimuli and behaviour gradually changes over the course of days to weeks in a phenomenon known as representational drift. The tuning stability of individual cells varies over the population, and it remains unclear what drives this heterogeneity. We investigate how a neuron’s tuning stability relates to its shared variability with other neurons in the population using two published datasets from posterior parietal cortex and visual cortex. We quantified the contribution of pairwise interactions to behaviour or stimulus encoding by partial information decomposition, which breaks down the mutual information between the pairwise neural activity and the external variable into components uniquely provided by each neuron and by their interactions. Information shared by the two neurons is termed ‘redundant’, and information requiring knowledge of the state of both neurons is termed ‘synergistic’. We found that a neuron’s tuning stability is positively correlated with the strength of its average pairwise redundancy with the population. We hypothesize that subpopulations of neurons show greater stability because they are tuned to salient features common across multiple tasks. Regardless of the mechanistic implications of our work, the stability–redundancy relationship may support improved longitudinal neural decoding in technology that has to track population dynamics over time, such as brain–machine interfaces."]]></description>
<dc:subject>to:NB neural_coding_and_decoding information_theory</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:be3a18a79790/</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:neural_coding_and_decoding"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:information_theory"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.jstor.org/content/oa_book_monograph/j.ctt1djmg7w">
    <title>The 1870 Ghost Dance | JSTOR</title>
    <dc:date>2026-07-30T17:58:16+00:00</dc:date>
    <link>https://www.jstor.org/content/oa_book_monograph/j.ctt1djmg7w</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>in_NB books:noted downloaded american_history 19th_century_history millenarianism native_american_history</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:65b895f52866/</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: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:millenarianism"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:native_american_history"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://users.monash.edu.au/~nwormald/papers/de.pdf">
    <title>The differential equation method for random graph processes and greedy algorithms (Wormald, 1997)</title>
    <dc:date>2026-07-30T15:24:22+00:00</dc:date>
    <link>https://users.monash.edu.au/~nwormald/papers/de.pdf</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>to_read re:do-institutions-evolve convergence_of_stochastic_processes stochastic_processes graph_theory wormald.nicholas via:cris_moore in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:b340150fbbd9/</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:re:do-institutions-evolve"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:convergence_of_stochastic_processes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:stochastic_processes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:graph_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:wormald.nicholas"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:cris_moore"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://doi.org/10.1214/aoap/1177004612">
    <title>Differential Equations for Random Processes and Random Graphs</title>
    <dc:date>2026-07-30T15:23:14+00:00</dc:date>
    <link>https://doi.org/10.1214/aoap/1177004612</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["General criteria are given to ensure that in a family of discrete random processes, given parameters exhibit convergence to the solution of a system of differential equations. As one application we consider random graph processes in which the maximum degree is bounded and show that the numbers of vertices of given degree exhibit this convergence as the total number of vertices tends to infinity. Two other applications are to random processes which generate independent sets of vertices in random $r$-regular graphs. In these cases, we deduce almost sure lower bounds on the size of independent sets of vertices in random $r$-regular graphs."]]></description>
<dc:subject>to_read convergence_of_stochastic_processes stochastic_processes via:cris_moore re:do-institutions-evolve wormald.nicholas in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e9d6878a8322/</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:convergence_of_stochastic_processes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:stochastic_processes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:cris_moore"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:do-institutions-evolve"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:wormald.nicholas"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://doi.org/10.1017/psa.2026.10267">
    <title>Navigating Epistemic Monocultures in AI-Driven Science: A Simulation Study | Philosophy of Science | Cambridge Core</title>
    <dc:date>2026-07-30T15:18:40+00:00</dc:date>
    <link>https://doi.org/10.1017/psa.2026.10267</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["AI integration into scientific communities promises accelerated discovery but raises concerns about detrimental homogenization. We develop an NK landscape model to explore these promises and risks.We find that non-personalized AI systems that offer uniform guidance yield benefits only under a narrow conjunction of problem structure, practices, and baseline research capabilities, becoming harmful otherwise. We implement two proposed mitigations: randomization and personalization. While randomization’s utility remains restricted to decomposable problems, personalization can enhance diversity, enabling benefits across a broader range of conditions. Crucially, these benefits are not automatic: they depend on effective institutional adaptation, requiring new standards, protocols, and practices."]]></description>
<dc:subject>to:NB sociology_of_science diversity nk_model</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:bdcd0b37799f/</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:sociology_of_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:diversity"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:nk_model"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://doi.org/10.1017/psa.2026.10268">
    <title>What use the Realist for a Theory of Truth? | Philosophy of Science | Cambridge Core</title>
    <dc:date>2026-07-30T15:16:56+00:00</dc:date>
    <link>https://doi.org/10.1017/psa.2026.10268</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Recently, Hasok Chang has argued that we should adopt a pragmatic account of truth on which truth is constituted by practical success. Central to his argument is the claim that truth in this sense can be pursued and that other senses of truth can’t be. In this paper, I present a dilemma for Chang’s view. Either practical success constitutes truth in a way that implies infallibility—a conclusion that I’ll argue is untenable—or the argument from pursuit fails. I end by suggesting that scientific realism alone commits us to very little with respect to truth."

--- After skimming, the more obvious objection to this sort of pragmatism would seem to be circularity: accepting "theory A is more practically successful than theory B" would need to be _more practically successful_ than "theory B is more practically successful than theory A", etc., ad inifinitum.  OR: judgments of practical success / relative practical success need to have a more ordinary sort of truth or falsity, but why those and not lower-level theories?  (IIRC, this objection to defining "truth" as "what works" originates with Russell vs. William James, but I'm not sure of that.)  Anyway, this is going of a skim of this paper's account of Chang, who I should just read in my infinite free time.

--- ETA: I should have read more carefully, this all comes up in sec. 5 of the paper.]]></description>
<dc:subject>to:NB pragmatism philosophy_of_science track_down_references have_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:6a5d203f62e9/</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:pragmatism"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:philosophy_of_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:track_down_references"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://doi.org/10.1017/psa.2026.10266">
    <title>Exploring the Maximally Sensitive Priors | Philosophy of Science | Cambridge Core</title>
    <dc:date>2026-07-30T14:24:44+00:00</dc:date>
    <link>https://doi.org/10.1017/psa.2026.10266</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We explore the Maximum Sensitivity method (MaxSen) of selecting priors for Bayesian inference in the face of severe lack of evidence (Konek, 2013). MaxSen minimizes the need to rely on the epistemic luck of true parameter values falling closely to their prior estimate.We explore the impact of the choice of the scoring method on the recommendations made by MaxSen. It turns out that if we use Kullback-Leibler Divergence (KLD) as a distance measure, the recommendations are the same as that of MaxEnt. We estimate how quickly the recommended prior distribution concentration increases with the planned sample size."]]></description>
<dc:subject>to:NB maximum_entropy bayesianism</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:3ee5065c5069/</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:maximum_entropy"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bayesianism"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2607.14421">
    <title>[2607.14421] Temporal Memory in Repeating Fast Radio Bursts: Epsilon-Machine Reconstruction of Causal Structure in Burst Timing</title>
    <dc:date>2026-07-29T03:29:44+00:00</dc:date>
    <link>https://arxiv.org/abs/2607.14421</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The emission mechanism of fast radio bursts (FRBs) remains unknown. Whether the bursts from a repeating FRB arrive at random or in a structured sequence is a key constraint on that mechanism. We apply ε-machine reconstruction, a tool from computational mechanics that infers the minimal model capturing all predictive information in a stochastic process. Applied to the waiting-time sequences of three repeating FRBs (FRB~20121102A and FRB~20201124A from FAST; FRB~20220912A from CHIME), the method yields the statistical complexity Cμ, the minimum number of bits required for optimal prediction. Both FAST sources carry roughly one bit of temporal memory (significant against permutation surrogates, p≤0.01; per-source false-discovery-rate-adjusted p≤0.028), while FRB~20220912A is consistent with memoryless emission. FRB~20201124A's memory spans hours-to-days across four sessions, FRB~20121102A's spans hours-to-weeks across thirty-nine, and neither source shows defensible within-session predictive memory. For FRB~20121102A the ordering of those sessions is itself predictive (session-shuffle p=0.02), whereas FRB~20201124A's signal reflects the contrast between heterogeneous sessions rather than their order. A simulated windowing test shows that CHIME's short transit observations would suppress comparable structure in the FAST data, leaving FRB~20220912A's null result ambiguous. This first application of ε-machine reconstruction to astrophysical transients yields a model-independent constraint: the bursting of at least two of these repeaters is not memoryless, but is governed by a hidden state that occupies distinct activity-rate regimes varying across observing sessions, behaviour that any viable physical model must reproduce."]]></description>
<dc:subject>to:NB prediction_processes to_read via:? astrophysics</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:2f7ac8c21cba/</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_processes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:?"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:astrophysics"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2306.06529">
    <title>[2306.06529] Neural Injective Functions for Multisets, Measures and Graphs via a Finite Witness Theorem</title>
    <dc:date>2026-07-29T03:26:49+00:00</dc:date>
    <link>https://arxiv.org/abs/2306.06529</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Injective multiset functions have a key role in the theoretical study of machine learning on multisets and graphs. Yet, there remains a gap between the provably injective multiset functions considered in theory, which typically rely on polynomial moments, and the multiset functions used in practice, which rely on neural moments — whose injectivity on multisets has not been studied to date.
"In this paper, we bridge this gap by showing that moments of neural networks do define injective multiset functions, provided that an analytic non-polynomial activation is used. The number of moments required by our theory is optimal essentially up to a multiplicative factor of two. To prove this result, we state and prove a finite witness theorem, which is of independent interest.
"As a corollary to our main theorem, we derive new approximation results for functions on multisets and measures, and new separation results for graph neural networks. We also provide two negative results: (1) moments of piecewise-linear neural networks cannot be injective multiset functions; and (2) even when moment-based multiset functions are injective, they can never be bi-Lipschitz."

--- The most relevant stuff for me is in the appendix.]]></description>
<dc:subject>to_read mathematics neural_networks re:codename:catherine_wheel via:mw-s</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:56c4e8142d0d/</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:mathematics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neural_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:codename:catherine_wheel"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mw-s"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.molbiolcell.org/doi/10.1091/mbc.E19-02-0107">
    <title>Biophysics at the coffee shop: lessons learned working with George Oster | Molecular Biology of the Cell</title>
    <dc:date>2026-07-28T03:14:00+00:00</dc:date>
    <link>https://www.molbiolcell.org/doi/10.1091/mbc.E19-02-0107</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Over the past 50 years, the use of mathematical models, derived from physical reasoning, to describe molecular and cellular systems has evolved from an art of the few to a cornerstone of biological inquiry. George Oster stood out as a pioneer of this paradigm shift from descriptive to quantitative biology not only through his numerous research accomplishments, but also through the many students and postdocs he mentored over his long career. Those of us fortunate enough to have worked with George agree that his sharp intellect, physical intuition, and passion for scientific inquiry not only inspired us as scientists but also greatly influenced the way we conduct research. We would like to share a few important lessons we learned from George in honor of his memory and with the hope that they may inspire future generations of scientists."]]></description>
<dc:subject>to:NB to_read physics biology molecular_biology oster.george</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:8fbfa21e1e5f/</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:physics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:biology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:molecular_biology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:oster.george"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.santafe.edu/news-center/news/in-memoriam-peter-schuster">
    <title>In Memoriam: Peter Schuster | Santa Fe Institute</title>
    <dc:date>2026-07-27T19:16:06+00:00</dc:date>
    <link>https://www.santafe.edu/news-center/news/in-memoriam-peter-schuster</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- I never knew him that well, but well enough to be saddened by this news.]]></description>
<dc:subject>schuster.peter in_memoriam have_read self-organization molecular_biology</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e34116f65149/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:schuster.peter"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_memoriam"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:self-organization"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:molecular_biology"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2607.20068">
    <title>[2607.20068] Catastrophic disruption cascades driven by the nonlinearity of systemic risk</title>
    <dc:date>2026-07-27T19:10:29+00:00</dc:date>
    <link>https://arxiv.org/abs/2607.20068</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Whether the COVID-19 pandemic or the Iran war, recent events have highlighted the systemic fragility of supply chains. Due to highly specific and mutual buyer-supplier dependencies, even the failure of a single firm can cause system-wide economic disruptions in the form of cascading failures up and down the supply chain network. Only recently has it become possible to quantify the systemic impact of the failure of individual firms on the total supply chain. Here, we demonstrate that the systemic risk contributions of combinations of firm failures can be drastically larger than the sum of the damage caused by the firms individually. Using a unique data set that allows us to reconstruct the national supply chain network of Ecuador at the firm-level, we find that combined failures can produce systemic risk amplifications of up to a factor of 257. However, only a tiny fraction of 0.14\% of pairs exhibit a more than 4-fold amplification of systemic risk. We develop a simple method to identify firm combinations that lead to large systemic risk amplifications. The origin of these amplifications is a breakdown of the substitutability of defaulted suppliers. We discuss the implications of the existence of rare but strong systemic risk amplification for situations that simultaneously affect multiple firms, such as natural disasters and wars."]]></description>
<dc:subject>economic_networks networks havlin.shlomo thurner.stefan ecuador in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:9b16299cf281/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:economic_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:havlin.shlomo"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:thurner.stefan"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:ecuador"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2607.20751">
    <title>[2607.20751] Local permutation tests for conditional independence: an adaptive binning perspective</title>
    <dc:date>2026-07-27T19:07:05+00:00</dc:date>
    <link>https://arxiv.org/abs/2607.20751</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["In this work, we study the problem of testing conditional independence between random variables X and Y given a confounder Z. The local permutation test (LPT) offers a principled approach to this problem by partitioning the Z-space into pre-specified bins, and permuting the X and Y data within each bin, to assess the significance of an observed test statistic. However, when the partitions are pre-fixed, the resulting partition can be poorly balanced, as some bins may contain most of the samples while others contain only a few. This motivates the use of data-adaptive binning strategies, such as equisized bins with a fixed (typically small) number of points. We study this natural and practically important extension of LPT, providing finite-sample bounds on the Type I error for an arbitrary test statistic, providing stronger validity results than previously known. We also show that LPT attains power comparable to the oracle likelihood ratio tests derived from the Neyman-Pearson lemma. Within a linear confounder model class, we further analyze the effect of bin size and demonstrate that constant bin sizes can match the performance of partitions with growing bin-size. These results, further supported by extensive numerical simulations, position the proposed data-adaptive strategy as both practically implementable and statistically efficient."]]></description>
<dc:subject>conditional_independence_testing statistics in_NB barber.rina_foygel</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:3c6ad09cc8c5/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:conditional_independence_testing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:barber.rina_foygel"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2607.21916">
    <title>[2607.21916] Estimating dynamic models by matching random features</title>
    <dc:date>2026-07-27T03:17:45+00:00</dc:date>
    <link>https://arxiv.org/abs/2607.21916</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Scientists increasingly express their ideas as dynamic models of complex processes. It is often much easier to simulate these models than to calculate the probability of their generating a particular outcome, making likelihood-based estimation infeasible. Existing likelihood-free approaches rely either on manually chosen summary statistics or on representations learned by neural networks. The former is error-prone and laborious, while the latter is computationally intensive, leaving many scientists in a difficult position. We show that, for a large class of dynamic models, parameters can be estimated by matching a small number of random features of the observed and simulated data. Specifically, we adapt results from nonlinear dynamics to show that models with a p-dimensional parameter can generically be identified from just 2p+1 random features. We introduce two estimators for stationary and nonstationary processes, respectively, and we establish their consistency under mild regularity conditions. More broadly, our results serve as the foundation for a new class of random feature methods for simulation-based estimation and inference."]]></description>
<dc:subject>self-promotion random_features simulation-based_inference re:codename:catherine_wheel</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a69db1549c31/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:self-promotion"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:random_features"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:simulation-based_inference"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:codename:catherine_wheel"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.annualreviews.org/content/journals/10.1146/annurev-physchem-082423-010652">
    <title>Reaction Coordinates Are Optimal Channels of Energy Flow | Annual Reviews</title>
    <dc:date>2026-07-24T20:11:01+00:00</dc:date>
    <link>https://www.annualreviews.org/content/journals/10.1146/annurev-physchem-082423-010652</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Reaction coordinates (RCs) are the few essential coordinates of a protein that control its functional processes, such as allostery, enzymatic reaction, and conformational change. They are critical for understanding protein function and provide optimal enhanced sampling of protein conformational changes and states. Since the pioneering work in the late 1990s, identifying the correct and objectively provable RCs has been a central topic in molecular biophysics and chemical physics. This review summarizes the major advances in identifying RCs over the past 25 years, focusing on methods aimed at finding RCs that meet the rigorous committor criterion, widely accepted as the true RCs. Notably, the newly developed physics-based energy flow theory and generalized work functional method provide a general and rigorous approach for identifying true RCs, revealing their physical nature as the optimal channels of energy flow in biomolecules."]]></description>
<dc:subject>molecular_biology thermodynamics statistical_mechanics dimension_reduction in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:8551067d5b3a/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:molecular_biology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:thermodynamics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistical_mechanics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:dimension_reduction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2607.16035">
    <title>[2607.16035] Dynamic models with $p$ parameters are identified by $2p+1$ random features</title>
    <dc:date>2026-07-20T14:34:32+00:00</dc:date>
    <link>https://arxiv.org/abs/2607.16035</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["A foundational principle in nonlinear dynamics is that the structure of a dynamical system can be recovered from a small number of generic measurements or coordinates. We develop an analogous principle for the identification of dynamic models for time series {\em with noise}, which builds on previous identification results for noiseless dynamical systems. The noise is allowed to be non-iid, non-Gaussian, and dependent on the state. Our results cover noisily observed differential equations and discrete-time dynamical systems, as well as stochastic models with process noise. We illustrate the utility of this identification principle using a Lorenz-63 model and a Hénon map model, both with observational noise."]]></description>
<dc:subject>in_NB self-promotion random_features simulation-based_inference re:codename:catherine_wheel</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:86f11637ce33/</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:self-promotion"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:random_features"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:simulation-based_inference"/>
	<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/2605.26711">
    <title>[2605.26711] The Need for an External Observer Formalizing the Sufficiency Gap: A Mathematical Extension of Mixture Identifiability and Contextual Grounding in Sequence Models</title>
    <dc:date>2026-07-17T13:57:10+00:00</dc:date>
    <link>https://arxiv.org/abs/2605.26711</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We construct a binary mixed-regime process with one deterministic textual regime and one random regime governed by an unobserved latent state. Even an ideal infinite-capacity sequence predictor that exactly recovers the text-only marginal law can become overconfident when the observed prefix is compatible with the wrong latent regime. The resulting entropy difference is not an ordinary optimization error; it is a sufficiency gap caused by marginalization over an unobserved state. We then formalize retrieval, tool use, and external grounding through an auxiliary binary signal with fidelity γ∈[1/2,1]. The resulting Bayesian update yields a contextual dominance threshold: a corrective signal reverses the posterior odds induced by the textual history exactly when its fidelity exceeds the text-only posterior weight assigned to the misleading regime. This threshold reduces, but does not generally eliminate, the sufficiency gap; complete closure requires perfect revelation of the relevant latent state or an equivalent verification mechanism. The analysis clarifies why temperature scaling cannot restore missing context, why grounding mechanisms must be both informative and learnably usable by the model, and why autonomous sequence models require structurally decoupled observers or verifiers in high-stakes domains."]]></description>
<dc:subject>via:rvenkat prediction inference_to_latent_objects large_language_models_(so_called)</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:39c519a8fb94/</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:prediction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:inference_to_latent_objects"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://doi.org/10.1145/3759429.3762631">
    <title>Gauguin, Descartes, Bayes: A Diurnal Golem’s Brain | Proceedings of the 2025 ACM SIGPLAN International Symposium on New Ideas, New Paradigms, and Reflections on Programming and Software</title>
    <dc:date>2026-07-17T01:09:33+00:00</dc:date>
    <link>https://doi.org/10.1145/3759429.3762631</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["A "quine" is a deterministic program that prints itself. In this essay, I will show you a "gauguine": a probabilistic program that infers itself. A gauguine is repeatedly asked to guess its own source code. Initially, its chances of guessing correctly are of course minuscule. But as the gauguine observes more and more of its own previous guesses, it detects patterns of behavior and gains information about its inner workings. This information allows it to bootstrap self-knowledge, and ultimately discover its own source code. We will discuss how—and why—we might write a gauguine, and what we stand to learn by constructing one."
]]></description>
<dc:subject>to:NB to_read theoretical_computer_science learning_theory tenebaum.joshua_b. at_a_loss_for_words</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:17a2f07043d0/</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:theoretical_computer_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:learning_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:tenebaum.joshua_b."/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:at_a_loss_for_words"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2602.04770">
    <title>[2602.04770] Generative Modeling via Drifting</title>
    <dc:date>2026-07-15T01:56:10+00:00</dc:date>
    <link>https://arxiv.org/abs/2602.04770</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Generative modeling can be formulated as learning a mapping f such that its pushforward distribution matches the data distribution. The pushforward behavior can be carried out iteratively at inference time, for example in diffusion and flow-based models. In this paper, we propose a new paradigm called Drifting Models, which evolve the pushforward distribution during training and naturally admit one-step inference. We introduce a drifting field that governs the sample movement and achieves equilibrium when the distributions match. This leads to a training objective that allows the neural network optimizer to evolve the distribution. In experiments, our one-step generator achieves state-of-the-art results on ImageNet at 256 x 256 resolution, with an FID of 1.54 in latent space and 1.61 in pixel space. We hope that our work opens up new opportunities for high-quality one-step generation."

--- Trying to read this, and I am going to need to do it with pencil and paper, not because it's so dense but because it is so badly written.  Not sure it'll be worth the effort.]]></description>
<dc:subject>to:NB probability neural_networks via:mraginsky density_estimation</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:984baf98df6f/</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:probability"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neural_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:density_estimation"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://inpreparation.substack.com/p/opinion-i-was-not-allowed-to-type">
    <title>Opinion: I Was Not Allowed To Type Prompts Into ChatGPT During My Chalk Talk And This Is Discrimination</title>
    <dc:date>2026-07-15T01:48:37+00:00</dc:date>
    <link>https://inpreparation.substack.com/p/opinion-i-was-not-allowed-to-type</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- I have yet to hear this from job candidates; undergrads are another matter.
--- To be clear, this is a satirical publication. ]]></description>
<dc:subject>have_read funny:academic funny:laughing_instead_of_screaming large_language_models_(so_called)</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:f3ca464b7d3b/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:funny:academic"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:funny:laughing_instead_of_screaming"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2607.01521">
    <title>[2607.01521] Selling the Stock, Not the Cream: The Soviet Émigré Career Premium of the 1990s</title>
    <dc:date>2026-07-15T01:44:26+00:00</dc:date>
    <link>https://arxiv.org/abs/2607.01521</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["In the early-mid 1990s, scientists emigrating from the former Soviet Union to the United States -- especially physicists, engineers, chemists, and biologists -- frequently secured prestigious and visible positions, including professorships, named chairs, and laboratory leadership; comparable scientists arriving after about 2000 built more modest, less visible, and often non-academic careers. Against the common view that this reflects the people -- the elite having left first -- this article sets aside the thin apex of Nobel- and Fields-level émigrés and examines the larger cohort of capable but non-stellar scientists, showing that similar scientists fared differently by year of arrival. The explanation therefore lies in the structure of the receiving market, not primarily in individual ability. Reading premium appointments backward from later Nobel-level recognition risks survivorship bias: celebrated successes obscure the broader demand for Soviet scientific capital. I weigh four conditions that favoured the 1990s cohort and had largely closed by the mid-2000s: technology transfer and the export of a finite, distinctive stock of Soviet expertise that commanded a career premium; the favourable immigration regime created by the Soviet Scientists Immigration Act of 1992; the surge of U.S.-trained Chinese and Indian competitors; and the securitizing aftermath of 11~September 2001. All four mattered, but technology transfer and knowledge export were primary: their premium opened the window, and their depletion -- as exported knowledge was published and absorbed into global science -- removed the demand on which the other factors depended. A further cross-cutting mechanism, the cultural ``ghettoization'' of émigrés into co-national laboratory enclaves, capped their visibility and independent advancement. The imbalance between émigré generations was structural, not personal."]]></description>
<dc:subject>to:NB sociology_of_science post-soviet_life via:rvenkat</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:3dcebbaab06c/</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:sociology_of_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:post-soviet_life"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:rvenkat"/>
</rdf:Bag></taxo:topics>
</item>
<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://role-confusion.github.io/">
    <title>Prompt Injection as Role Confusion</title>
    <dc:date>2026-07-15T01:42:04+00:00</dc:date>
    <link>https://role-confusion.github.io/</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>have_read large_language_models_(so_called) via:kjhealy tracked_down_references</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:54fdeebe063c/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_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:kjhealy"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:tracked_down_references"/>
</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://arxiv.org/abs/2606.22748">
    <title>[2606.22748] AI Fiction in the Wild</title>
    <dc:date>2026-06-26T02:37:28+00:00</dc:date>
    <link>https://arxiv.org/abs/2606.22748</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Some professional authors are beginning to use AI tools to help produce their fiction writing. Are readers using AI to generate fiction, too? Drawing on over 500,000 anonymized, English-language ChatGPT-user conversations (arXiv:2405.01470), we find that more than one third of the conversations involve some form of fiction generation -- including original stories, roleplay, fanfiction, and erotica. This AI-generated fiction is notably dominated by power users. We identify common fiction generation patterns and profiles among these users, including what we call "infinite story demanders," who repeatedly request and revise variations of the same or similar narratives over extended periods of time. We show that users especially gravitate toward fanfiction and erotica, and that they are broadly drawn to generic forms, repetition, immediacy, and niche combinations of story elements. Our findings motivate two theoretical provocations. First, we argue that AI technologies may lead to a shift in the conventional relationship between the author and reader, potentially producing what we call a "solipsistic reader-writer," who both generates and consumes fiction within a closed conversational loop, interacting with a machine rather than a human other. Second, we note that LLMs enable interactivity, play, and permutation in ways that are seemingly pleasurable for users, raising questions about where AI will fit into contemporary storytelling and entertainment ecosystems. We situate these developments within broader transformations in literature and media, including self-publishing, fanfiction, and pornography, and suggest that AI-generated fiction shares structural affinities with on-demand, personalized, and repetitive cultural forms."

--- "generic forms, repetition, immediacy, and niche combinations of story elements": well, yes, of course.  The Internet has created, since at least the 1980s, large populations of readers and writers of written smut which display precisely those characteristics.  This is obvious if you browsed AO3, and equally obvious if you browsed the Alt.Sex.Stories Text Repository --- or, ahem, so I am told.  These are the communities which invented tags!  I am not in the least bit surprised to see that this is also what people ask from the bots.]]></description>
<dc:subject>to:NB via:henry_farrell large_language_models_(so_called) nerdworld pr0n</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:003b6ac2e092/</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:via:henry_farrell"/>
	<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:nerdworld"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:pr0n"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://link.springer.com/article/10.1007/s10955-025-03550-6">
    <title>On Interactions for Large Scale Interacting Systems | Journal of Statistical Physics | Springer Nature Link</title>
    <dc:date>2026-06-25T18:26:51+00:00</dc:date>
    <link>https://link.springer.com/article/10.1007/s10955-025-03550-6</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Statistical mechanics explains the properties of macroscopic phenomena based on the movements of microscopic particles such as atoms and molecules. Movements of microscopic particles can be represented by large-scale interacting systems. In this article, we systematically study combinatorial objects which we call interactions, given as symmetric directed graphs representing the possible transitions of states on adjacent sites of large-scale interacting systems. Such interactions underlie various standard stochastic processes such as the exclusion processes, generalized exclusion processes, multi-species exclusion processes, lattice gas with energy processes, and the multi-lane exclusion processes. We introduce the notion of equivalences of interactions using their space of conserved quantities. This allows for the classification of interactions reflecting the expected macroscopic properties. In particular, we prove that when the set of local states consists of two, three or four elements, then the number of equivalence classes of separable interactions are respectively one, two and five. We also define the wedge sums and box products of interactions, which give systematic methods for constructing new interactions from existing ones. Furthermore, we prove that the irreducibly quantified condition for interactions, which implicitly plays an important role in the theory of hydrodynamic limits, is preserved by wedge sums and box products. Our results provide a systematic method to construct and classify interactions, offering abundant examples suitable for considering hydrodynamic limits."]]></description>
<dc:subject>to:NB combinatorics interacting_particle_systems</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:027686d3c088/</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:combinatorics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:interacting_particle_systems"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://link.springer.com/article/10.1007/s10955-026-03647-6">
    <title>Large Deviations for Subgraphs in Inhomogeneous Random Graphs | Journal of Statistical Physics | Springer Nature Link</title>
    <dc:date>2026-06-25T18:24:18+00:00</dc:date>
    <link>https://link.springer.com/article/10.1007/s10955-026-03647-6</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Inhomogeneous random graphs are fundamental models for real-world networks, where prescribed degrees are imposed as soft constraints. A common assumption in such models is that the degree distribution follows a power-law, capturing the heavy-tailed nature observed in many contexts. While various graph functionals have been studied in this setting, inhomogeneity makes their analysis significantly more challenging. Here, we investigate the large deviations of subgraph counts in inhomogeneous random graphs. Rare events concerning these functionals translate into quantifying the probability that extremely large hubs appear in the graph. This can be achieved by defining a specific optimization problem that captures the most likely way to generate numerous additional subgraphs. When the expected number of subgraphs is sublinear in the graph size, polynomially large deviations are possible, and in this case, we can derive sharp results on clique counts."]]></description>
<dc:subject>to:NB large_deviations graph_theory graph_limits</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:b8c7a77f8a95/</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_deviations"/>
	<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:Bag></taxo:topics>
</item>
<item rdf:about="https://www.cambridge.org/core/journals/philosophy-of-science/article/case-for-time-in-causal-dags/FB8A5FA21249300B5250358B21A3E26D?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles">
    <title>The Case For Time in Causal DAGs | Philosophy of Science | Cambridge Core</title>
    <dc:date>2026-06-24T12:55:55+00:00</dc:date>
    <link>https://www.cambridge.org/core/journals/philosophy-of-science/article/case-for-time-in-causal-dags/FB8A5FA21249300B5250358B21A3E26D?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We make the case for incorporating a notion of time into causal directed acyclic graphs (DAGs). We demonstrate that nontemporal causal DAGs are ambiguous and obstruct justification of the acyclicity assumption. Assuming that causes precede effects, causal relationships are relative to the time order, and causal DAGs require temporal qualification. We propose a formalization via composite causal variables that refer to quantities at one or multiple time points. We emphasize that the acyclicity assumption requires different justifications depending on whether the time order allows cycles. We conclude by discussing implications for the interpretation and applicability of DAGs as causal models."]]></description>
<dc:subject>to:NB graphical_models causality philosophy_of_science</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:90a1d2ee3253/</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:graphical_models"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:causality"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:philosophy_of_science"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.cambridge.org/core/journals/philosophy-of-science/article/noncausal-explanations-of-social-and-biological-networks/59E80990AA755C7CBBD6E543832681C4?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles">
    <title>Non-causal Explanations of Social and Biological Networks | Philosophy of Science | Cambridge Core</title>
    <dc:date>2026-06-24T12:54:36+00:00</dc:date>
    <link>https://www.cambridge.org/core/journals/philosophy-of-science/article/noncausal-explanations-of-social-and-biological-networks/59E80990AA755C7CBBD6E543832681C4?WT.mc_id=New%2520Cambridge%2520Alert%2520-%2520Articles</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["One argument that all explanations are causal explanations is that no extant analysis of non-causal explanations can respect mandatory restrictions on explanation. After articulating two such restrictions, Asymmetry and Directionality, we consider how they work with respect to an explanation from network theory that applies to both social and brain networks. We argue that there are two viable ways to make sense of this explanation. One approach is broadly ontic, while another is pragmatic."]]></description>
<dc:subject>to:NB philosophy_of_science networks explanation</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ab6037e35687/</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:philosophy_of_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:explanation"/>
</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://direct.mit.edu/books/monograph/6064/Wired-for-WordsThe-Neural-Architecture-of-Language">
    <title>Wired for Words: The Neural Architecture of Language | Books Gateway | MIT Press</title>
    <dc:date>2026-06-18T14:10:56+00:00</dc:date>
    <link>https://direct.mit.edu/books/monograph/6064/Wired-for-WordsThe-Neural-Architecture-of-Language</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["A critical synthesis of over 150 years of research on the brain’s networks that enable us to communicate through language.
"The neural architecture of language has been a hotly debated topic in neurology, cognitive neuroscience, linguistics, and philosophy since the early 1800s. Is language separable from intelligence? Is it enabled by dedicated and localizable neural networks? Do we speak and understand with our left hemisphere? How did language emerge? Is language grounded in sensorimotor systems, or is it abstract and amodal? Will we ever have a clear picture of how syntax, the pinnacle of human linguistic prowess, is organized neurologically?
"Wired for Words answers these questions and more. Gregory Hickok tells the stories behind the big ideas, revealing the source of both modern progress and persistent myths. Drawing on decades of research using tools and insights from neurology, functional imaging, neurosurgery, linguistics, psychology, and engineering, Hickok builds a new understanding of the neural architecture—the components and connection patterns—of the brain’s language system from sound to meaning to speech."]]></description>
<dc:subject>in_NB books:noted linguistics neuropsychology to_download</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:295cd5ae0a19/</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:linguistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neuropsychology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_download"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.youtube.com/watch?v=fO9iRDPXvT4">
    <title>The Most Arrogant Science Book Ever Written - YouTube</title>
    <dc:date>2026-06-18T13:57:09+00:00</dc:date>
    <link>https://www.youtube.com/watch?v=fO9iRDPXvT4</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- As the person who alerted me to this put it, the sincerest form of flattery.  (I am credited with authoring the review at the beginning, so I suppose I don't have _too_ much to complain about, and sending a take-down notice would just be churlish.)  But the idea that it's worth someone's while to narrate a book review I wrote in 2002, because it gets hundreds of thousands of views, is very strange to me (to put it mildly).]]></description>
<dc:subject>self-centered not_exactly_self-promotion networked_life</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:8409e874e8d6/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:self-centered"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:not_exactly_self-promotion"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networked_life"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://nathan.rs/posts/gzip-lm/">
    <title>Can gzip be a language model?</title>
    <dc:date>2026-06-17T16:28:26+00:00</dc:date>
    <link>https://nathan.rs/posts/gzip-lm/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- Cf. [https://bactra.org/notebooks/nn-attention-and-transformers.html#gllz], but the beam-search trick is a good one.]]></description>
<dc:subject>to:NB re:gllz via:kjhealy to_teach:statistics_and_generative_ai have_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:da162a9639cd/</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:re:gllz"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:kjhealy"/>
	<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:have_read"/>
</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://cacm.acm.org/opinion/artificial-intelligence-for-software-engineering-from-probable-to-provable/">
    <title>Artificial Intelligence for Software Engineering: From Probable to Provable – Communications of the ACM</title>
    <dc:date>2026-06-17T16:06:58+00:00</dc:date>
    <link>https://cacm.acm.org/opinion/artificial-intelligence-for-software-engineering-from-probable-to-provable/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Here we go again: No programmers will be needed anymore! AI will generate the code! If you have been around for a while, you may feel a sense of déjà vu. That same line advertised COBOL in the 1960s, 4GLs in the 1970s, CASE tools in the 1980s, component-based development in the 1990s, model-driven architecture in the 2000s, and low-code/no-code in the 2010s. Some of these approaches did improve programming, but they did not replace programming, let alone programmers. They simply introduced higher levels of abstraction or new tools, sometimes taking advantage of a restricted application domain. Is it the same this time, or do artificial intelligence (AI) and vibe coding upend the game? More generally, can AI and software engineering enter into a successful marriage?
"Warning and spoiler alert: Even though the following discussion starts out by examining limitations of AI for software construction, do not just expect a critique. Its aim is positive, in support of AI-supported software engineering. Its core thesis (here I am really spilling the beans) is that a successful solution requires combining AI with formal verification. (End of spoiler.)"

--- This makes sense, but I keep getting hung up on why we didn't do all of this with "genetic programming" in the 1990s.  "Prompt engineering is requirements specification", yes, but then you define a fitness function (*) and let the code evolve.  If you have good verification tools, you include that in the fitness function.  Maybe code is _so_ repetitive that training things to semi-memorize large chunks of all the code on the Internet is better than evolving from scratch, but I'd really like to see the cost-benefit on that...

*: Or a vector of fitness functions and drive to the Pareto frontier.  (I think I finally get why Bill T. was so into multi-objective optimization for genetic programming.)]]></description>
<dc:subject>to:NB large_language_models_(so_called) programming via:mraginsky have_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:f0450c258fd1/</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:programming"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.trainjazz.com/">
    <title>Every train, a note.</title>
    <dc:date>2026-06-17T16:01:41+00:00</dc:date>
    <link>https://www.trainjazz.com/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Every dot is a real subway train. Eight hundred of them, give or take, form a small jazz combo (walking bass, piano, sax, vibes, brushes) that has been playing without pause for over a hundred years. On the platforms they are hot, screaming, full of complaint. This is the music inside the noise.
"The harmony moves through a slow chorus. A note is placed precisely where the train happens to be along its route. Rush hour fills the band with held tones; at 3 a.m. the silences grow longer. Whatever is playing now has not played before and will not play again.
"Share your location and the trains nearest you grow louder. The piece rearranges itself around your body. You are listening to a portrait of where you stand, played by the city you are standing in."

--- Actually sounds decent, unlike most such stunts.]]></description>
<dc:subject>music new_york_city have_listened</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:5585d2bf4f76/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:music"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:new_york_city"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_listened"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.reuters.com/commentary/breakingviews/physical-shocks-are-shrinking-power-money-2026-06-04/">
    <title>Physical shocks are shrinking the power of money | Reuters</title>
    <dc:date>2026-06-17T16:00:46+00:00</dc:date>
    <link>https://www.reuters.com/commentary/breakingviews/physical-shocks-are-shrinking-power-money-2026-06-04/</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>book_reviews economics money track_down_references</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:fc12e76a2400/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:book_reviews"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:economics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:money"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:track_down_references"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://aclanthology.org/2020.cl-2.7/">
    <title>Fair Is Better than Sensational: Man Is to Doctor as Woman Is to Doctor - ACL Anthology</title>
    <dc:date>2026-06-17T16:00:06+00:00</dc:date>
    <link>https://aclanthology.org/2020.cl-2.7/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Analogies such as man is to king as woman is to X are often used to illustrate the amazing power of word embeddings. Concurrently, they have also been used to expose how strongly human biases are encoded in vector spaces trained on natural language, with examples like man is to computer programmer as woman is to homemaker. Recent work has shown that analogies are in fact not an accurate diagnostic for bias, but this does not mean that they are not used anymore, or that their legacy is fading. Instead of focusing on the intrinsic problems of the analogy task as a bias detection tool, we discuss a series of issues involving implementation as well as subjective choices that might have yielded a distorted picture of bias in word embeddings. We stand by the truth that human biases are present in word embeddings, and, of course, the need to address them. But analogies are not an accurate tool to do so, and the way they have been most often used has exacerbated some possibly non-existing biases and perhaps hidden others. Because they are still widely popular, and some of them have become classics within and outside the NLP community, we deem it important to provide a series of clarifications that should put well-known, and potentially new analogies, into the right perspective."

--- The most astonishing thing to me here is realizing that in the usual "A is to B as C is to D" protocols, lots of experiments _prohibited_ D from being the same as B, so e.g. in "Man is to doctor as Woman is to ?", the answer _could not_ be "doctor".  (This of course connects to the authors' point that it's often really unclear what an acceptable, un-biased answer might possibly be.)]]></description>
<dc:subject>to:NB have_read analogy algorithmic_fairness to_teach:data-mining to_teach:statistics_of_inequality_and_discrimination</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ff3f7072cb0a/</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:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:analogy"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithmic_fairness"/>
	<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:statistics_of_inequality_and_discrimination"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://cooking.nytimes.com/recipes/780676171-cucumber-and-onion-salad">
    <title>Cucumber and Onion Salad Recipe</title>
    <dc:date>2026-06-17T15:57:16+00:00</dc:date>
    <link>https://cooking.nytimes.com/recipes/780676171-cucumber-and-onion-salad</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[INGREDIENTS
Yield: 6 servings
4 English cucumbers, sliced into 1/8-inch-thick rounds (about 10 cups)
2 tablespoons kosher salt (such as Diamond Crystal)
1.5 cups white vinegar
1/4 cup sugar
2 teaspoons fresh ground black pepper
1 small white onion, thinly sliced (about 1½ cups)

PREPARATION
Step 1
Toss cucumbers and salt together in a large bowl. Transfer cucumbers to a strainer, then place the strainer in the sink. Let cucumbers sit for 30 minutes as they release their water, stirring occasionally to help them drain. 

Step 2
Meanwhile, whisk the vinegar, sugar and pepper in a large bowl until the sugar dissolves. 

Step 3
Add the drained cucumbers and sliced onion to the bowl of vinegar marinade. Use your hands or tongs to toss well—really get in there and make sure the marinade is distributed. Cover and refrigerate for at least 1 hour before serving, stirring once halfway through to ensure everything gets evenly marinated.


--- AEO says it reminds her of her grandparents (in central PA, not the South...)]]></description>
<dc:subject>food recipes have_made</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:59b740127f14/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:food"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recipes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_made"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://link.springer.com/book/10.1007/978-3-031-97239-3">
    <title>Signature Methods in Finance: An Introduction with Computational Applications | Springer Nature Link</title>
    <dc:date>2026-06-13T04:04:14+00:00</dc:date>
    <link>https://link.springer.com/book/10.1007/978-3-031-97239-3</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["This Open Access volume offers an accessible entry point into the fast-growing field of signature methods in finance. It is written for early-career researchers and quantitatively minded practitioners—quant analysts and applied researchers—seeking a clear, practical introduction. It highlights recent developments and includes coding examples to help readers apply signature methods in practice.
"The advantages of modeling financial markets from a path-wise perspective, rather than as a traditional series of returns, are increasingly gaining recognition. Signature methods provide a parsimonious description of paths of stochastic processes and, through the signature kernel, open a rich and compelling framework at the interface between machine learning and mathematical finance."]]></description>
<dc:subject>to:NB books:noted path_signatures time_series stochastic_processes finance kernel_methods</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a5cf1e7772a2/</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:path_signatures"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:time_series"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:stochastic_processes"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:finance"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kernel_methods"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2601.05444">
    <title>[2601.05444] What Functions Does XGBoost Learn?</title>
    <dc:date>2026-06-04T18:10:45+00:00</dc:date>
    <link>https://arxiv.org/abs/2601.05444</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["This paper establishes a rigorous theoretical foundation for the function class implicitly learned by XGBoost, bridging the gap between its empirical success and our theoretical understanding. We introduce an infinite-dimensional function class d,s∞−ST that extends finite ensembles of bounded-depth regression trees, together with a complexity measure Vd,s∞−XGB(⋅) that generalizes the L1 regularization penalty used in XGBoost. We show that every optimizer of the XGBoost objective is also an optimizer of an equivalent penalized regression problem over d,s∞−ST with penalty Vd,s∞−XGB(⋅), providing an interpretation of XGBoost as implicitly targeting a broader function class. We also develop a smoothness-based interpretation of d,s∞−ST and Vd,s∞−XGB(⋅) in terms of Hardy--Krause variation. We prove that the least squares estimator over {f∈d,s∞−ST:Vd,s∞−XGB(f)≤V} achieves a nearly minimax-optimal rate of convergence n−2/3(logn)4(min(s,d)−1)/3, thereby avoiding the curse of dimensionality. Our results provide the first rigorous characterization of the function space underlying XGBoost, clarify its connection to classical notions of variation, and identify an important open problem: whether the XGBoost algorithm itself achieves minimax optimality over this class."]]></description>
<dc:subject>to:NB functional_analysis boosting ensemble_methods decision_trees via:msw</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:85d5009a07b7/</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:functional_analysis"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:boosting"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:ensemble_methods"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:decision_trees"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:msw"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.science.org/doi/10.1126/sciadv.aao3580">
    <title>Trends and fluctuations in the severity of interstate wars | Science Advances</title>
    <dc:date>2026-06-04T15:54:36+00:00</dc:date>
    <link>https://www.science.org/doi/10.1126/sciadv.aao3580</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Since 1945, there have been relatively few large interstate wars, especially compared to the preceding 30 years, which included both World Wars. This pattern, sometimes called the long peace, is highly controversial. Does it represent an enduring trend caused by a genuine change in the underlying conflict-generating processes? Or is it consistent with a highly variable but otherwise stable system of conflict? Using the empirical distributions of interstate war sizes and onset times from 1823 to 2003, we parameterize stationary models of conflict generation that can distinguish trends from statistical fluctuations in the statistics of war. These models indicate that both the long peace and the period of great violence that preceded it are not statistically uncommon patterns in realistic but stationary conflict time series. This fact does not detract from the importance of the long peace or the proposed mechanisms that explain it. However, the models indicate that the postwar pattern of peace would need to endure at least another 100 to 140 years to become a statistically significant trend. This fact places an implicit upper bound on the magnitude of any change in the true likelihood of a large war after the end of the Second World War. The historical patterns of war thus seem to imply that the long peace may be substantially more fragile than proponents believe, despite recent efforts to identify mechanisms that reduce the likelihood of interstate wars."]]></description>
<dc:subject>to:NB war kith_and_kin clauset.aaron social_measurement time_series</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:e174e28da7ca/</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:war"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kith_and_kin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:clauset.aaron"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_measurement"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:time_series"/>
</rdf:Bag></taxo:topics>
</item>
</rdf:RDF>