<?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://aclanthology.org/2020.cl-2.7/"/>
	<rdf:li rdf:resource="https://dl.acm.org/doi/full/10.1145/3722548"/>
	<rdf:li rdf:resource="https://openreview.net/forum?id=SBE2q9qwZj"/>
	<rdf:li rdf:resource="https://aeon.co/essays/the-sovereign-individual-and-the-paradox-of-the-digital-age"/>
	<rdf:li rdf:resource="https://www.annualreviews.org/content/journals/10.1146/annurev-polisci-040723-013245"/>
	<rdf:li rdf:resource="https://www.propublica.org/article/inside-ai-tool-doge-veterans-affairs-contracts-sahil-lavingia"/>
	<rdf:li rdf:resource="https://dl.acm.org/doi/10.1145/321033.321035"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/1503.02531"/>
	<rdf:li rdf:resource="https://dl.acm.org/doi/10.1145/1150402.1150464"/>
	<rdf:li rdf:resource="https://www.theverge.com/2024/12/5/24313222/chatgpt-pardon-biden-bush-esquire"/>
	<rdf:li rdf:resource="https://www.nytimes.com/interactive/2024/07/18/technology/spain-domestic-violence-viogen-algorithm.html?smid=nytcore-ios-share&amp;referringSource=articleShare&amp;sgrp=c-cb"/>
	<rdf:li rdf:resource="https://nowpublishers.com/article/Details/MAL-106"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2310.00865"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/1712.03586#"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2301.11562"/>
	<rdf:li rdf:resource="https://knightcolumbia.org/content/understanding-social-media-recommendation-algorithms"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2305.17224"/>
	<rdf:li rdf:resource="https://github.com/igorbrigadir/awesome-twitter-algo"/>
	<rdf:li rdf:resource="https://onlinelibrary.wiley.com/doi/full/10.1111/papa.12233"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2301.07015"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2205.13026"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2206.13102"/>
	<rdf:li rdf:resource="https://www.wired.com/story/welfare-state-algorithms/"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2206.02667"/>
	<rdf:li rdf:resource="https://www.amacad.org/publication/moral-economy-high-tech-modernism"/>
	<rdf:li rdf:resource="https://www.jstor.org/stable/27765899#metadata_info_tab_contents"/>
	<rdf:li rdf:resource="https://www.aiweirdness.com/galactica/"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2201.07203"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2208.08489"/>
	<rdf:li rdf:resource="https://www.foreignaffairs.com/world/spirals-delusion-artificial-intelligence-decision-making"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2207.08815"/>
	<rdf:li rdf:resource="https://www.nytimes.com/2022/06/25/technology/china-surveillance-police.html"/>
	<rdf:li rdf:resource="https://jmlr.org/papers/v23/21-1427.html"/>
	<rdf:li rdf:resource="https://jmlr.org/papers/v23/20-874.html"/>
	<rdf:li rdf:resource="https://www.commerce.senate.gov/services/files/62102355-DC26-4909-BF90-8FB068145F18"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/1906.04493"/>
	<rdf:li rdf:resource="https://logicmag.io/play/my-stepdad's-huge-data-set/"/>
	<rdf:li rdf:resource="https://www.nytimes.com/2022/06/25/technology/china-surveillance-police.html?action=click&amp;module=Well&amp;pgtype=Homepage&amp;section=Business"/>
	<rdf:li rdf:resource="https://www.aiweirdness.com/interview-with-a-squirrel/"/>
	<rdf:li rdf:resource="https://www.annualreviews.org/doi/abs/10.1146/annurev-criminol-030920-112506"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2205.03009"/>
	<rdf:li rdf:resource="https://www.nature.com/articles/s41467-019-10933-3/"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2203.06498"/>
	<rdf:li rdf:resource="https://www.jmlr.org/papers/v23/20-874.html"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2201.10408"/>
	<rdf:li rdf:resource="https://www.sciencedirect.com/science/article/pii/S016792361400061X?casa_token=8mKjKuwIF58AAAAA:DjRmaJuDoBZjjHb2kA3iEoCckvybsakE7Ww6qdBRABxULXlOuE8FIvmSbgMgYO0ZwLasjyow"/>
	<rdf:li rdf:resource="https://archive.ics.uci.edu/ml/datasets/Bank+Marketing"/>
	<rdf:li rdf:resource="https://www.nytimes.com/2022/02/19/technology/qanon-messages-authors.html"/>
	<rdf:li rdf:resource="https://hci.stanford.edu/publications/2019/streetlevelalgorithms/streetlevelalgorithms-chi2019.pdf"/>
	<rdf:li rdf:resource="https://press.princeton.edu/books/paperback/9780691177311/experiments-of-the-mind"/>
	<rdf:li rdf:resource="https://www.pnas.org/content/118/22/e2018340118"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2104.04103"/>
	<rdf:li rdf:resource="https://www.quantamagazine.org/computer-scientists-discover-limits-of-major-research-algorithm-20210817/"/>
	<rdf:li rdf:resource="https://www.washingtonpost.com/technology/interactive/2021/how-facebook-algorithm-works?no_nav=true"/>
	<rdf:li rdf:resource="https://www.pnas.org/content/118/32/e2101967118.short"/>
	<rdf:li rdf:resource="https://www.annualreviews.org/doi/abs/10.1146/annurev-soc-090820-020800"/>
	<rdf:li rdf:resource="https://www.technologyreview.com/2021/07/07/1027916/we-tested-ai-interview-tools/?_hsmi=140522126&amp;_hsenc=p2ANqtz-8vjTOrf7wPGyBHzGvNM6HAsRT9_ivd6OAndeGlgd7q_DeKQzL8wzhIWUxeUBctrU37fL6-4CUvXLmfyBLOvUsDCnCCddi0lNRwTMkkbRP71Z9mu90"/>
	<rdf:li rdf:resource="https://nymag.com/intelligencer/2020/09/inside-palantir-technologies-peter-thiel-alex-karp.html"/>
	<rdf:li rdf:resource="https://www.nature.com/articles/s41386-021-01020-7"/>
	<rdf:li rdf:resource="https://bigthink.com/laurie-vazquez/are-you-a-geek-or-a-nerd"/>
	<rdf:li rdf:resource="https://doxa.substack.com/p/phrenology-insurance-claims-and-digital?token=eyJ1c2VyX2lkIjozMTk2MjUwOSwicG9zdF9pZCI6MzcxODg4MDYsIl8iOiI5RlR3eiIsImlhdCI6MTYyMjc3NzYxOSwiZXhwIjoxNjIyNzgxMjE5LCJpc3MiOiJwdWItMjM5NjUzIiwic3ViIjoicG9zdC1yZWFjdGlvbiJ9.r6AUl5BSCgCU4aMl6zL1Pt8xcC3cnNU5E7J6LDGlQbs"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2105.09065"/>
	<rdf:li rdf:resource="https://doi.org/10.1111/rssb.12425"/>
	<rdf:li rdf:resource="https://www.nber.org/papers/w28811"/>
	<rdf:li rdf:resource="https://www.sup.org/books/title/?id=32597"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2105.04648"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2105.04134"/>
	<rdf:li rdf:resource="http://simondedeo.com/?p=705"/>
	<rdf:li rdf:resource="https://www.ucpress.edu/book/9780520382046/netflix-recommends"/>
	<rdf:li rdf:resource="https://arxiv.org/abs/2003.00381"/>
      </rdf:Seq>
    </items>
  </channel><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://dl.acm.org/doi/full/10.1145/3722548">
    <title>Concerning the Responsible Use of AI in the U.S. Criminal Justice System | Communications of the ACM</title>
    <dc:date>2026-06-04T15:53:42+00:00</dc:date>
    <link>https://dl.acm.org/doi/full/10.1145/3722548</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>to:NB kith_and_kin algorithmic_fairness moore.cris 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:a1ce478cb8c7/</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:kith_and_kin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:moore.cris"/>
	<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://openreview.net/forum?id=SBE2q9qwZj">
    <title>Fast Computation of Leave-One-Out Cross-Validation for $k$-NN Regression | OpenReview</title>
    <dc:date>2025-09-02T19:30:09+00:00</dc:date>
    <link>https://openreview.net/forum?id=SBE2q9qwZj</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We describe a fast computation method for leave-one-out cross-validation (LOOCV) for 
$k$-nearest neighbours ($k$-NN) regression. We show that, under a tie-breaking condition for nearest neighbours, the LOOCV estimate of the mean square error for $k$
-NN regression is identical to the mean square error of $(k+1)$-NN regression evaluated on the training data, multiplied by the scaling factor $(k+1)^2/𝑘^2$. Therefore, to compute the LOOCV score, one only needs to fit $(k+1)$-NN regression only once, and does not need to repeat training-validation of $k$-NN regression for the number of training data. Numerical experiments confirm the validity of the fast computation method."

!!!]]></description>
<dc:subject>to:NB to_read nearest_neighbors to_teach:data-mining to_teach:undergrad-ADA via:?</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a7ddd207a4f6/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to:NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:nearest_neighbors"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:undergrad-ADA"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:?"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://aeon.co/essays/the-sovereign-individual-and-the-paradox-of-the-digital-age">
    <title>The sovereign individual and the paradox of the digital age | Aeon Essays</title>
    <dc:date>2025-08-22T13:06:47+00:00</dc:date>
    <link>https://aeon.co/essays/the-sovereign-individual-and-the-paradox-of-the-digital-age</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>have_read healy.kieran networked_life to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:4fd386526e12/</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:healy.kieran"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networked_life"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.annualreviews.org/content/journals/10.1146/annurev-polisci-040723-013245">
    <title>AI as Governance | Annual Reviews</title>
    <dc:date>2025-06-20T19:38:37+00:00</dc:date>
    <link>https://www.annualreviews.org/content/journals/10.1146/annurev-polisci-040723-013245</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>in_NB algorithmic_fairness artificial_intelligence re:shoggothim farrell.henry kith_and_kin have_read to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:c53762fb1437/</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:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:artificial_intelligence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:shoggothim"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:farrell.henry"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kith_and_kin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.propublica.org/article/inside-ai-tool-doge-veterans-affairs-contracts-sahil-lavingia">
    <title>Inside the AI Tool Used by DOGE to Review Veterans Affairs Contracts — ProPublica</title>
    <dc:date>2025-06-15T15:50:36+00:00</dc:date>
    <link>https://www.propublica.org/article/inside-ai-tool-doge-veterans-affairs-contracts-sahil-lavingia</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>utter_stupidity us_politics large_language_models_(so_called) programming classifiers to_teach:data-mining have_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:8dfe8878ce5a/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:utter_stupidity"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:us_politics"/>
	<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:classifiers"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://dl.acm.org/doi/10.1145/321033.321035">
    <title>On Relevance, Probabilistic Indexing and Information Retrieval | Journal of the ACM [1960]</title>
    <dc:date>2025-05-17T11:53:39+00:00</dc:date>
    <link>https://dl.acm.org/doi/10.1145/321033.321035</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["This paper reports on a novel technique for literature indexing and searching in a mechanized library system. The notion of relevance is taken as the key concept in the theory of information retrieval and a comparative concept of relevance is explicated in terms of the theory of probability. The resulting technique called “Probabilistic Indexing,” allows a computing machine, given a request for information, to make a statistical inference and derive a number (called the “relevance number”) for each document, which is a measure of the probability that the document will satisfy the given request. The result of a search is an ordered list of those documents which satisfy the request ranked according to their probable relevance.
"The paper goes on to show that whereas in a conventional library system the cross-referencing (“see” and “see also”) is based solely on the “semantical closeness” between index terms, statistical measures of closeness between index terms can be defined and computed. Thus, given an arbitrary request consisting of one (or many) index term(s), a machine can elaborate on it to increase the probability of selecting relevant documents that would not otherwise have been selected.
"Finally, the paper suggests an interpretation of the whole library problem as one where the request is considered as a clue on the basis of which the library system makes a concatenated statistical inference in order to provide as an output an ordered list of those documents which most probably satisfy the information needs of the user."

--- 1960!]]></description>
<dc:subject>to_read information_retrieval relevance to_teach:data-mining via:mraginsky in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:0515f0733905/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:information_retrieval"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:relevance"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:mraginsky"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/1503.02531">
    <title>[1503.02531] Distilling the Knowledge in a Neural Network</title>
    <dc:date>2025-03-04T18:58:35+00:00</dc:date>
    <link>https://arxiv.org/abs/1503.02531</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["A very simple way to improve the performance of almost any machine learning algorithm is to train many different models on the same data and then to average their predictions. Unfortunately, making predictions using a whole ensemble of models is cumbersome and may be too computationally expensive to allow deployment to a large number of users, especially if the individual models are large neural nets. Caruana and his collaborators have shown that it is possible to compress the knowledge in an ensemble into a single model which is much easier to deploy and we develop this approach further using a different compression technique. We achieve some surprising results on MNIST and we show that we can significantly improve the acoustic model of a heavily used commercial system by distilling the knowledge in an ensemble of models into a single model. We also introduce a new type of ensemble composed of one or more full models and many specialist models which learn to distinguish fine-grained classes that the full models confuse. Unlike a mixture of experts, these specialist models can be trained rapidly and in parallel."]]></description>
<dc:subject>to:NB ensemble_methods to_read to_teach:data-mining via:rvenkat</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a950b4f51e90/</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:ensemble_methods"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:rvenkat"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://dl.acm.org/doi/10.1145/1150402.1150464">
    <title>Model compression | Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining</title>
    <dc:date>2025-03-04T18:57:34+00:00</dc:date>
    <link>https://dl.acm.org/doi/10.1145/1150402.1150464</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Often the best performing supervised learning models are ensembles of hundreds or thousands of base-level classifiers. Unfortunately, the space required to store this many classifiers, and the time required to execute them at run-time, prohibits their use in applications where test sets are large (e.g. Google), where storage space is at a premium (e.g. PDAs), and where computational power is limited (e.g. hea-ring aids). We present a method for "compressing" large, complex ensembles into smaller, faster models, usually without significant loss in performance."

--- KDD '06!  WTH didn't I know about this?]]></description>
<dc:subject>to:NB ensemble_methods to_read to_teach:data-mining via:rvenkat</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:5a3d0fd453f6/</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:ensemble_methods"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:rvenkat"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.theverge.com/2024/12/5/24313222/chatgpt-pardon-biden-bush-esquire">
    <title>Stop using generative AI as a search engine - The Verge</title>
    <dc:date>2024-12-06T13:56:06+00:00</dc:date>
    <link>https://www.theverge.com/2024/12/5/24313222/chatgpt-pardon-biden-bush-esquire</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- The "to_teach" tags are just the courses I'm gearing up for in the next few semesters; I suspect I will be giving _this_ lesson (and having it ignored) for many years to come.]]></description>
<dc:subject>large_language_models_(so_called) to_teach to_teach:data-mining to_teach:undergrad-ADA to_teach:statistics_of_inequality_and_discrimination information_retrieval have_read via:?</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:4dc951b569f7/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:undergrad-ADA"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:statistics_of_inequality_and_discrimination"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:information_retrieval"/>
	<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.nytimes.com/interactive/2024/07/18/technology/spain-domestic-violence-viogen-algorithm.html?smid=nytcore-ios-share&amp;referringSource=articleShare&amp;sgrp=c-cb">
    <title>An Algorithm Told Police She Was Safe. Then Her Husband Killed Her. - The New York Times</title>
    <dc:date>2024-07-19T04:32:41+00:00</dc:date>
    <link>https://www.nytimes.com/interactive/2024/07/18/technology/spain-domestic-violence-viogen-algorithm.html?smid=nytcore-ios-share&amp;referringSource=articleShare&amp;sgrp=c-cb</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- I'll certainly teach this, but it's worth emphasizing something that this only hints at in its quotes.  Suppose this scoring system is (1) generally very reliable but not perfect, while (2) police protection works when it's provided.  It then follows that a very large share of the women who do get killed will be ones who were scored at low risk, precisely because all the high-risk cases are being effectually protected!]]></description>
<dc:subject>to:NB to_teach:data-mining risk_assessment violence ethical_and_political_issues_in_data_mining decision-making have_read via:aeo</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:6ffe6f5184b4/</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:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:risk_assessment"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:violence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:ethical_and_political_issues_in_data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:decision-making"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:aeo"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://nowpublishers.com/article/Details/MAL-106">
    <title>now publishers - Causal Fairness Analysis: A Causal Toolkit for Fair Machine Learning</title>
    <dc:date>2024-03-02T19:20:27+00:00</dc:date>
    <link>https://nowpublishers.com/article/Details/MAL-106</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Decision-making systems based on AI and machine learning have been used throughout a wide range of real-world scenarios, including healthcare, law enforcement, education, and finance. It is no longer far-fetched to envision a future where autonomous systems will drive entire business decisions and, more broadly, support large-scale decision-making infrastructure to solve society’s most challenging problems. Issues of unfairness and discrimination are pervasive when decisions are being made by humans, and remain (or are potentially amplified) when decisions are made using machines with little transparency, accountability, and fairness. In this monograph, we introduce a framework for causal fairness analysis with the intent of filling in this gap, i.e., understanding, modeling, and possibly solving issues of fairness in decision-making settings.
"The main insight of our approach will be to link the quantification of the disparities present in the observed data with the underlying, often unobserved, collection of causal mechanisms that generate the disparity in the first place, a challenge we call the Fundamental Problem of Causal Fairness Analysis (FPCFA). In order to solve the FPCFA, we study the problem of decomposing variations and empirical measures of fairness that attribute such variations to structural mechanisms and different units of the population. Our effort culminates in the Fairness Map, the first systematic attempt to organize and explain the relationship between various criteria found in the literature. Finally, we study which causal assumptions are minimally needed for performing causal fairness analysis and propose the Fairness Cookbook, which allows one to assess the existence of disparate impact and disparate treatment."

--- Ungated: [https://causalai.net/r90.pdf]]]></description>
<dc:subject>to:NB algorithmic_fairness causal_inference to_read to_teach:statistics_of_inequality_and_discrimination to_teach:data-mining bareinboim.elias</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:58b48843b8d2/</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:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:causal_inference"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<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:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bareinboim.elias"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2310.00865">
    <title>[2310.00865] Data Science at the Singularity</title>
    <dc:date>2023-11-16T16:52:03+00:00</dc:date>
    <link>https://arxiv.org/abs/2310.00865</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["A purported `AI Singularity' has been in the public eye recently. Mass media and US national political attention focused on `AI Doom' narratives hawked by social media influencers. The European Commission is announcing initiatives to forestall `AI Extinction'. In my opinion, `AI Singularity' is the wrong narrative for what's happening now; recent happenings signal something else entirely. Something fundamental to computation-based research really changed in the last ten years. In certain fields, progress is dramatically more rapid than previously, as the fields undergo a transition to frictionless reproducibility (FR). This transition markedly changes the rate of spread of ideas and practices, affects mindsets, and erases memories of much that came before.
"The emergence of frictionless reproducibility follows from the maturation of 3 data science principles in the last decade. Those principles involve data sharing, code sharing, and competitive challenges, however implemented in the particularly strong form of frictionless open services. Empirical Machine Learning (EML) is todays leading adherent field, and its consequent rapid changes are responsible for the AI progress we see. Still, other fields can and do benefit when they adhere to the same principles.
"Many rapid changes from this maturation are misidentified. The advent of FR in EML generates a steady flow of innovations; this flow stimulates outsider intuitions that there's an emergent superpower somewhere in AI. This opens the way for PR to push worrying narratives: not only `AI Extinction', but also the supposed monopoly of big tech on AI research. The helpful narrative observes that the superpower of EML is adherence to frictionless reproducibility practices; these practices are responsible for the striking progress in AI that we see everywhere."]]></description>
<dc:subject>to:NB donoho.david reproducibility computational_statistics to_teach:statcomp to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ff1a86af9c29/</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:donoho.david"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:reproducibility"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:computational_statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:statcomp"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/1712.03586#">
    <title>[1712.03586] Fairness in Machine Learning: Lessons from Political Philosophy</title>
    <dc:date>2023-10-24T13:46:25+00:00</dc:date>
    <link>https://arxiv.org/abs/1712.03586#</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["What does it mean for a machine learning model to be `fair', in terms which can be operationalised? Should fairness consist of ensuring everyone has an equal probability of obtaining some benefit, or should we aim instead to minimise the harms to the least advantaged? Can the relevant ideal be determined by reference to some alternative state of affairs in which a particular social pattern of discrimination does not exist? Various definitions proposed in recent literature make different assumptions about what terms like discrimination and fairness mean and how they can be defined in mathematical terms. Questions of discrimination, egalitarianism and justice are of significant interest to moral and political philosophers, who have expended significant efforts in formalising and defending these central concepts. It is therefore unsurprising that attempts to formalise `fairness' in machine learning contain echoes of these old philosophical debates. This paper draws on existing work in moral and political philosophy in order to elucidate emerging debates about fair machine learning."]]></description>
<dc:subject>in_NB political_philosophy algorithmic_fairness to_read via:wiggins 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:52018eac4bb1/</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:political_philosophy"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:wiggins"/>
	<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://arxiv.org/abs/2301.11562">
    <title>[2301.11562] Is My Prediction Arbitrary? The Confounding Effects of Variance in Fair Classification Benchmarks</title>
    <dc:date>2023-09-15T19:39:01+00:00</dc:date>
    <link>https://arxiv.org/abs/2301.11562</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Variance in predictions across different trained models is a significant, under-explored source of error in fair classification. In practice, the variance on some data examples is so large that decisions can be effectively arbitrary. To investigate this problem, we take an experimental approach and make four overarching contributions: We 1) Define a metric called self-consistency, derived from variance, which we use as a proxy for measuring and reducing arbitrariness; 2) Develop an ensembling algorithm that abstains from classification when a prediction would be arbitrary; 3) Conduct the largest to-date empirical study of the role of variance (vis-a-vis self-consistency and arbitrariness) in fair classification; and, 4) Release a toolkit that makes the US Home Mortgage Disclosure Act (HMDA) datasets easily usable for future research. Altogether, our experiments reveal shocking insights about the reliability of conclusions on benchmark datasets. Most fairness classification benchmarks are close-to-fair when taking into account the amount of arbitrariness present in predictions -- before we even try to apply common fairness interventions. This finding calls into question the practical utility of common algorithmic fairness methods, and in turn suggests that we should fundamentally reconsider how we choose to measure fairness in machine learning."

--- "Variance" here is defined slightly non-standardly, as E[loss(Y_1, Y_2)] where Y_1 and Y_2 are (distinct) draws from the distribution.  (If loss is squared error, this comes out to twice the usual definition of variance.)  "Self-consistency" is just the probability that two models, bootstrapped from the same data set, give the same classification for a given individual.]]></description>
<dc:subject>algorithmic_fairness via:rvenkat classifiers have_read ensemble_methods uncertainty_for_neural_networks in_NB to_teach:data-mining</dc:subject>
<dc:identifier>https://pinboard.in/u:cshalizi/b:bec8057ed430/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:rvenkat"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:classifiers"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:ensemble_methods"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:uncertainty_for_neural_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://knightcolumbia.org/content/understanding-social-media-recommendation-algorithms">
    <title>Understanding Social Media Recommendation Algorithms | Knight First Amendment Institute</title>
    <dc:date>2023-06-06T01:49:35+00:00</dc:date>
    <link>https://knightcolumbia.org/content/understanding-social-media-recommendation-algorithms</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>recommender_systems in_NB to_teach:data-mining naranyan.arvind</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:b4f86a8f7ba0/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:naranyan.arvind"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2305.17224">
    <title>[2305.17224] Fast and Minimax Optimal Estimation of Low-Rank Matrices via Non-Convex Gradient Descent</title>
    <dc:date>2023-06-05T03:07:07+00:00</dc:date>
    <link>https://arxiv.org/abs/2305.17224</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We study the problem of estimating a low-rank matrix from noisy measurements, with the specific goal of achieving minimax optimal error. In practice, the problem is commonly solved using non-convex gradient descent, due to its ability to scale to large-scale real-world datasets. In theory, non-convex gradient descent is capable of achieving minimax error. But in practice, it often converges extremely slowly, such that it cannot even deliver estimations of modest accuracy within reasonable time. On the other hand, methods that improve the convergence of non-convex gradient descent, through rescaling or preconditioning, also greatly amplify the measurement noise, resulting in estimations that are orders of magnitude less accurate than what is theoretically achievable with minimax optimal error. In this paper, we propose a slight modification to the usual non-convex gradient descent method that remedies the issue of slow convergence, while provably preserving its minimax optimality. Our proposed algorithm has essentially the same per-iteration cost as non-convex gradient descent, but is guaranteed to converge to minimax error at a linear rate that is immune to ill-conditioning. Using our proposed algorithm, we reconstruct a 60 megapixel dataset for a medical imaging application, and observe significantly decreased reconstruction error compared to previous approaches."]]></description>
<dc:subject>to:NB low-rank_approximation computational_statistics optimization to_teach:data-mining minimax</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:143024a0fac1/</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:low-rank_approximation"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:computational_statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:optimization"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:minimax"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://github.com/igorbrigadir/awesome-twitter-algo">
    <title>GitHub - igorbrigadir/awesome-twitter-algo: The release of the Twitter algorithm, annotated for recsys</title>
    <dc:date>2023-05-02T20:39:10+00:00</dc:date>
    <link>https://github.com/igorbrigadir/awesome-twitter-algo</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>twitter recommender_systems to_teach:data-mining have_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:661b081ee535/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:twitter"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://onlinelibrary.wiley.com/doi/full/10.1111/papa.12233">
    <title>Reconciling Algorithmic Fairness Criteria - Beigang - 2023 - Philosophy &amp; Public Affairs - Wiley Online Library</title>
    <dc:date>2023-05-02T19:24:32+00:00</dc:date>
    <link>https://onlinelibrary.wiley.com/doi/full/10.1111/papa.12233</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- I should re-read this carefully, but on first examination this is decidedly underwhelming.  If you read Chouldechova's paper [http://arxiv.org/abs/1610.07524] (or my class notes: [https://www.stat.cmu.edu/~cshalizi/dm/22/lectures/24/lecture-24.pdf], p. 13), she shows a three-way relationship between error-rate disparities, predictive-value disparities, and the ratio of "success" probabilities in the groups.  When, but only when, the "positive" outcome is equally prevalent in the two groups, then parity of error rates <=> parity of predictive values.  By matching on outcomes (!), Beigang has constructed distributed where the prevalences must be equal, so Chouldechova implies parity of error rates and predictive values.  But matching on outcomes is absurd.  I'd even say that matching on propensity scores would be absurd here.  Now since serious people are pushing this paper, it's quite likely that I'm missing something here, so I'll re-read, but at the very least this seems like bad exposition.]]></description>
<dc:subject>have_read to_reread algorithmic_fairness color_me_skeptical via:multiple to_teach:data-mining to_teach:statistics_of_inequality_and_discrimination matching in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:c5132029450e/</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_reread"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:color_me_skeptical"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:multiple"/>
	<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:li rdf:resource="https://pinboard.in/u:cshalizi/t:matching"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2301.07015">
    <title>[2301.07015] Simplistic Collection and Labeling Practices Limit the Utility of Benchmark Datasets for Twitter Bot Detection</title>
    <dc:date>2023-05-01T20:37:24+00:00</dc:date>
    <link>https://arxiv.org/abs/2301.07015</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Accurate bot detection is necessary for the safety and integrity of online platforms. It is also crucial for research on the influence of bots in elections, the spread of misinformation, and financial market manipulation. Platforms deploy infrastructure to flag or remove automated accounts, but their tools and data are not publicly available. Thus, the public must rely on third-party bot detection. These tools employ machine learning and often achieve near perfect performance for classification on existing datasets, suggesting bot detection is accurate, reliable and fit for use in downstream applications. We provide evidence that this is not the case and show that high performance is attributable to limitations in dataset collection and labeling rather than sophistication of the tools. Specifically, we show that simple decision rules -- shallow decision trees trained on a small number of features -- achieve near-state-of-the-art performance on most available datasets and that bot detection datasets, even when combined together, do not generalize well to out-of-sample datasets. Our findings reveal that predictions are highly dependent on each dataset's collection and labeling procedures rather than fundamental differences between bots and humans. These results have important implications for both transparency in sampling and labeling procedures and potential biases in research using existing bot detection tools for pre-processing."]]></description>
<dc:subject>to:NB classifiers networked_life deceiving_us_has_become_an_industrial_process decision_trees to_teach:data-mining philip_k_dick_and_the_fake_humans_rules_everything_around_me</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:26709234aea1/</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:classifiers"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networked_life"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:deceiving_us_has_become_an_industrial_process"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:decision_trees"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:philip_k_dick_and_the_fake_humans_rules_everything_around_me"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2205.13026">
    <title>[2205.13026] Preference Dynamics Under Personalized Recommendations</title>
    <dc:date>2023-03-22T03:02:55+00:00</dc:date>
    <link>https://arxiv.org/abs/2205.13026</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Many projects (both practical and academic) have designed algorithms to match users to content they will enjoy under the assumption that user's preferences and opinions do not change with the content they see. Evidence suggests that individuals' preferences are directly shaped by what content they see -- radicalization, rabbit holes, polarization, and boredom are all example phenomena of preferences affected by content. Polarization in particular can occur even in ecosystems with "mass media," where no personalization takes place, as recently explored in a natural model of preference dynamics by~\citet{hkazla2019geometric} and~\citet{gaitonde2021polarization}. If all users' preferences are drawn towards content they already like, or are repelled from content they already dislike, uniform consumption of media leads to a population of heterogeneous preferences converging towards only two poles.
"In this work, we explore whether some phenomenon akin to polarization occurs when users receive \emph{personalized} content recommendations. We use a similar model of preference dynamics, where an individual's preferences move towards content the consume and enjoy, and away from content they consume and dislike. We show that standard user reward maximization is an almost trivial goal in such an environment (a large class of simple algorithms will achieve only constant regret). A more interesting objective, then, is to understand under what conditions a recommendation algorithm can ensure stationarity of user's preferences. We show how to design a content recommendations which can achieve approximate stationarity, under mild conditions on the set of available content, when a user's preferences are known, and how one can learn enough about a user's preferences to implement such a strategy even when user preferences are initially unknown."]]></description>
<dc:subject>polarization recommender_systems low-regret_learning to_teach:data-mining in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:228777ce0410/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:polarization"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:low-regret_learning"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2206.13102">
    <title>[2206.13102] Modeling Content Creator Incentives on Algorithm-Curated Platforms</title>
    <dc:date>2023-03-22T02:57:24+00:00</dc:date>
    <link>https://arxiv.org/abs/2206.13102</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Content creators compete for user attention. Their reach crucially depends on algorithmic choices made by developers on online platforms. To maximize exposure, many creators adapt strategically, as evidenced by examples like the sprawling search engine optimization industry. This begets competition for the finite user attention pool. We formalize these dynamics in what we call an exposure game, a model of incentives induced by algorithms including modern factorization and (deep) two-tower architectures. We prove that seemingly innocuous algorithmic choices -- e.g., non-negative vs. unconstrained factorization -- significantly affect the existence and character of (Nash) equilibria in exposure games. We proffer use of creator behavior models like ours for an (ex-ante) pre-deployment audit. Such an audit can identify misalignment between desirable and incentivized content, and thus complement post-hoc measures like content filtering and moderation. To this end, we propose tools for numerically finding equilibria in exposure games, and illustrate results of an audit on the MovieLens and LastFM datasets. Among else, we find that the strategically produced content exhibits strong dependence between algorithmic exploration and content diversity, and between model expressivity and bias towards gender-based user and creator groups."]]></description>
<dc:subject>game_theory recommender_systems re:actually-dr-internet-is-the-name-of-the-monsters-creator jordan.michael_i. to_teach:data-mining social_media networked_life philip_k_dick_and_the_fake_humans_rules_everything_around_me in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:73a5583bf609/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:game_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:actually-dr-internet-is-the-name-of-the-monsters-creator"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:jordan.michael_i."/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_media"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networked_life"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:philip_k_dick_and_the_fake_humans_rules_everything_around_me"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.wired.com/story/welfare-state-algorithms/">
    <title>Inside the Suspicion Machine | WIRED</title>
    <dc:date>2023-03-21T15:43:38+00:00</dc:date>
    <link>https://www.wired.com/story/welfare-state-algorithms/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- Last tag for the underlying analysis.
--- The bit about coding _any_ comment from the social worker as a flag for trouble is mind-blowing, and not in a good way.]]></description>
<dc:subject>classifiers risk_assessment welfare_state algorithmic_fairness have_read to_teach:data-mining track_down_references bad_data_analysis</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:102b2cb016b2/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:classifiers"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:risk_assessment"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:welfare_state"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:track_down_references"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bad_data_analysis"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2206.02667">
    <title>[2206.02667] Multi-learner risk reduction under endogenous participation dynamics</title>
    <dc:date>2023-03-18T14:35:55+00:00</dc:date>
    <link>https://arxiv.org/abs/2206.02667</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Prediction systems face exogenous and endogenous distribution shift -- the world constantly changes, and the predictions the system makes change the environment in which it operates. For example, a music recommender observes exogeneous changes in the user distribution as different communities have increased access to high speed internet. If users under the age of 18 enjoy their recommendations, the proportion of the user base comprised of those under 18 may endogeneously increase. Most of the study of endogenous shifts has focused on the single decision-maker setting, where there is one learner that users either choose to use or not.
"This paper studies participation dynamics between sub-populations and possibly many learners. We study the behavior of systems with \emph{risk-reducing} learners and sub-populations. A risk-reducing learner updates their decision upon observing a mixture distribution of the sub-populations  in such a way that it decreases the risk of the learner on that mixture. A risk reducing sub-population updates its apportionment amongst learners in a way which reduces its overall loss.
"Previous work on the single learner case shows that myopic risk minimization can result in high overall loss~\citep{perdomo2020performative, miller2021outside} and representation disparity~\citep{hashimoto2018fairness, zhang2019group}. Our work analyzes the outcomes of multiple myopic learners and market forces, often leading to better global loss and less representation disparity."]]></description>
<dc:subject>to:NB algorithmic_fairness machine_learning recommender_systems to_read to_teach:data-mining distributed_systems re:actually-dr-internet-is-the-name-of-the-monsters-creator social_life_of_the_mind re:democratic_cognition diversity</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ae99da06e897/</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:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:machine_learning"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:distributed_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:actually-dr-internet-is-the-name-of-the-monsters-creator"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_life_of_the_mind"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:democratic_cognition"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:diversity"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.amacad.org/publication/moral-economy-high-tech-modernism">
    <title>The Moral Economy of High-Tech Modernism | American Academy of Arts and Sciences</title>
    <dc:date>2023-03-07T16:04:22+00:00</dc:date>
    <link>https://www.amacad.org/publication/moral-economy-high-tech-modernism</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["At the end of the day, the relationship between high modernism and high-tech modernism is a struggle between two elites: a new elite of coders, who claim to mediate the wisdom of crowds, and an older elite who based their claims to legitimacy on specialized professional, scientific, or bureaucratic knowledge.32 Both elites draw on rhetorical resources to justify their positions; neither is disinterested.
"The robust offense and disbelief that many people feel about algorithmic judgments suggests that the old high modernist moral political economy, faults and all, is not quite dead. The new moral political economy that will replace it has not yet matured, but is being bred from within. Articulated by technologists and their financial backers, it feeds in a kind of matriphagy on the enfeebled body (and the critique) of its progenitor. Just as high modernist bureaucracies did before, high-tech modernist tools and their designers categorize and order things, people, and situations. But they do so in distinctive ways. By embedding surveillance into everything, they have made us stop worrying about it, and perhaps even come to love it.33 By producing incomprehensible bespoke categorizations, they have made it harder for people to identify their common fate. By relying on opaque and automated feedback loops, they have reshaped the possible pathways to political reaction and resistance. By increasing the efficiency of online coordination, they have made mobilization more emotional, ad hoc, and collectively unstable. And by insisting on market fairness and the wisdom of crowds as organizing social concepts, they have fundamentally transformed our moral intuitions about authority, truth, objectivity, and deservingness."

--- Now add in HF's thoughts from "Philip K. Dick and the Fake Humans", and the fact that the emergent behavior of these systems is opaque to their designers and operators...]]></description>
<dc:subject>have_read kith_and_kin farrell.henry political_philosophy networked_life in_NB to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:2865aef699d1/</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:kith_and_kin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:farrell.henry"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:political_philosophy"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networked_life"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.jstor.org/stable/27765899#metadata_info_tab_contents">
    <title>Leveling with Lagrange: An Alternate View of Constrained Optimization on JSTOR</title>
    <dc:date>2022-12-29T03:50:50+00:00</dc:date>
    <link>https://www.jstor.org/stable/27765899#metadata_info_tab_contents</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- I find the description of "leveling" unilluminating, and borderline incomprehensible.  I am going to keep telling The Kids that Lagrange is a way of turning constrained problems into unconstrained problems over a larger space.  (The lie-told-to-children there is that the new problem is not "find a minimum" but "find a saddle point".)]]></description>
<dc:subject>to_teach:data-mining optimization have_read lagrange_multipliers</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:8bc5a2b74eb5/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:optimization"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:lagrange_multipliers"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.aiweirdness.com/galactica/">
    <title>Galactica: the AI knowledge base that makes stuff up</title>
    <dc:date>2022-12-28T04:48:04+00:00</dc:date>
    <link>https://www.aiweirdness.com/galactica/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- How the [EXPLETIVE] did anyone think this was a good idea?
--- As I keep pointing out, you can "invert" a probability-of-next-token predictor from any source compression scheme.  (Roughly speaking, if $L(x)$ is the number of bits used to encode finite string $x$, then the compressor assigns the string a probability $\approx 2^{-L(x})}$, hence $L(x_a) - L(x)$ is pretty much $-\log_2{Pr(a|x)}$. [Yes, yes, qualifications and details apply.])  And we have long known techniques, like Lempel-Ziv, which approach the limit of source coding as they see asymptotically long training sets.  So when you are considering an application of a large language model, you should ask yourself if you would feel comfortable if the next-word-predicting neural network was replaced with a really big implementation of gzip, with a dictionary built from scraping a large fraction of the Web...]]></description>
<dc:subject>neural_networks natural_language_processing utter_stupidity to_teach:data-mining large_language_models_(so_called) in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:0e7642f03a88/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neural_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:natural_language_processing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:utter_stupidity"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:large_language_models_(so_called)"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2201.07203">
    <title>[2201.07203] Emergent Instabilities in Algorithmic Feedback Loops</title>
    <dc:date>2022-12-02T16:01:14+00:00</dc:date>
    <link>https://arxiv.org/abs/2201.07203</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Algorithms that aid human tasks, such as recommendation systems, are ubiquitous. They appear in everything from social media to streaming videos to online shopping. However, the feedback loop between people and algorithms is poorly understood and can amplify cognitive and social biases (algorithmic confounding), leading to unexpected outcomes. In this work, we explore algorithmic confounding in collaborative filtering-based recommendation algorithms through teacher-student learning simulations. Namely, a student collaborative filtering-based model, trained on simulated choices, is used by the recommendation algorithm to recommend items to agents. Agents might choose some of these items, according to an underlying teacher model, with new choices then fed back into the student model as new training data (approximating online machine learning). These simulations demonstrate how algorithmic confounding produces erroneous recommendations which in turn lead to instability, i.e., wide variations in an item's popularity between each simulation realization. We use the simulations to demonstrate a novel approach to training collaborative filtering models that can create more stable and accurate recommendations. Our methodology is general enough that it can be extended to other socio-technical systems in order to better quantify and improve the stability of algorithms. These results highlight the need to account for emergent behaviors from interactions between people and algorithms."]]></description>
<dc:subject>recommender_systems re:actually-dr-internet-is-the-name-of-the-monsters-creator lerman.kristina via:? to_teach:data-mining to_read in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:cad25f273fb9/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:actually-dr-internet-is-the-name-of-the-monsters-creator"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:lerman.kristina"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:?"/>
	<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_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2208.08489">
    <title>[2208.08489] Understanding Scaling Laws for Recommendation Models</title>
    <dc:date>2022-09-03T19:30:05+00:00</dc:date>
    <link>https://arxiv.org/abs/2208.08489</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Scale has been a major driving force in improving machine learning performance, and understanding scaling laws is essential for strategic planning for a sustainable model quality performance growth, long-term resource planning and developing efficient system infrastructures to support large-scale models. In this paper, we study empirical scaling laws for DLRM style recommendation models, in particular Click-Through Rate (CTR). We observe that model quality scales with power law plus constant in model size, data size and amount of compute used for training. We characterize scaling efficiency along three different resource dimensions, namely data, parameters and compute by comparing the different scaling schemes along these axes. We show that parameter scaling is out of steam for the model architecture under study, and until a higher-performing model architecture emerges, data scaling is the path forward. The key research questions addressed by this study include: Does a recommendation model scale sustainably as predicted by the scaling laws? Or are we far off from the scaling law predictions? What are the limits of scaling? What are the implications of the scaling laws on long-term hardware/system development?"]]></description>
<dc:subject>recommender_systems to_teach:data-mining your_favorite_deep_neural_network_sucks in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:4cf0f6d5f5f3/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:your_favorite_deep_neural_network_sucks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.foreignaffairs.com/world/spirals-delusion-artificial-intelligence-decision-making">
    <title>Spirals of Delusion: How AI Distorts Decision-Making and Makes Dictators More Dangerous</title>
    <dc:date>2022-08-31T21:56:27+00:00</dc:date>
    <link>https://www.foreignaffairs.com/world/spirals-delusion-artificial-intelligence-decision-making</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>have_read in_NB data_mining algorithmic_fairness kith_and_kin farrell.henry to_teach:data-mining re:democratic_cognition seeing_like_a_finite_state_machine</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:bf239783c563/</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:in_NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kith_and_kin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:farrell.henry"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:democratic_cognition"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:seeing_like_a_finite_state_machine"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2207.08815">
    <title>[2207.08815] Why do tree-based models still outperform deep learning on tabular data?</title>
    <dc:date>2022-08-25T16:03:35+00:00</dc:date>
    <link>https://arxiv.org/abs/2207.08815</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["While deep learning has enabled tremendous progress on text and image datasets, its superiority on tabular data is not clear. We contribute extensive benchmarks of standard and novel deep learning methods as well as tree-based models such as XGBoost and Random Forests, across a large number of datasets and hyperparameter combinations. We define a standard set of 45 datasets from varied domains with clear characteristics of tabular data and a benchmarking methodology accounting for both fitting models and finding good hyperparameters. Results show that tree-based models remain state-of-the-art on medium-sized data (∼10K samples) even without accounting for their superior speed. To understand this gap, we conduct an empirical investigation into the differing inductive biases of tree-based models and Neural Networks (NNs). This leads to a series of challenges which should guide researchers aiming to build tabular-specific NNs: 1. be robust to uninformative features, 2. preserve the orientation of the data, and 3. be able to easily learn irregular functions. To stimulate research on tabular architectures, we contribute a standard benchmark and raw data for baselines: every point of a 20 000 compute hours hyperparameter search for each learner."]]></description>
<dc:subject>to:NB to_read your_favorite_deep_neural_network_sucks ensemble_methods decision_trees to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:2b5e2a0c03ab/</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:your_favorite_deep_neural_network_sucks"/>
	<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:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.nytimes.com/2022/06/25/technology/china-surveillance-police.html">
    <title>How China Is Policing the Future - The New York Times</title>
    <dc:date>2022-07-22T14:57:34+00:00</dc:date>
    <link>https://www.nytimes.com/2022/06/25/technology/china-surveillance-police.html</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- Remarkably little about whether these systems actually work, either in the sense of predicting accurately, _or_ in the sense of getting people to do what the authorities want.]]></description>
<dc:subject>have_read to_teach:data-mining seeing_like_a_finite_state_machine surveillance china:prc prediction data_mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:b19320bc850b/</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_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:seeing_like_a_finite_state_machine"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:surveillance"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:china:prc"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:prediction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://jmlr.org/papers/v23/21-1427.html">
    <title>Inherent Tradeoffs in Learning Fair Representations</title>
    <dc:date>2022-07-19T14:03:30+00:00</dc:date>
    <link>https://jmlr.org/papers/v23/21-1427.html</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Real-world applications of machine learning tools in high-stakes domains are often regulated to be fair, in the sense that the predicted target should satisfy some quantitative notion of parity with respect to a protected attribute. However, the exact tradeoff between fairness and accuracy is not entirely clear, even for the basic paradigm of classification problems. In this paper, we characterize an inherent tradeoff between statistical parity and accuracy in the classification setting by providing a lower bound on the sum of group-wise errors of any fair classifiers. Our impossibility theorem could be interpreted as a certain uncertainty principle in fairness: if the base rates differ among groups, then any fair classifier satisfying statistical parity has to incur a large error on at least one of the groups. We further extend this result to give a lower bound on the joint error of any (approximately) fair classifiers, from the perspective of learning fair representations. To show that our lower bound is tight, assuming oracle access to Bayes (potentially unfair) classifiers, we also construct an algorithm that returns a randomized classifier which is both optimal (in terms of accuracy) and fair. Interestingly, when the protected attribute can take more than two values, an extension of this lower bound does not admit an analytic solution. Nevertheless, in this case, we show that the lower bound can be efficiently computed by solving a linear program, which we term as the TV-Barycenter problem, a barycenter problem under the TV-distance. On the upside, we prove that if the group-wise Bayes optimal classifiers are close, then learning fair representations leads to an alternative notion of fairness, known as the accuracy parity, which states that the error rates are close between groups. Finally, we also conduct experiments on real-world datasets to confirm our theoretical findings."

--- I am sure this is not _just_ Chouldechova (2016), because Geoff wouldn't do that.
(Also, to keep repeating a point, suppose your sibling or your spouse was having their fate determined by a _randomized_ classifier, with the judge [or loan officer, etc.] rolling the d20 in front of you so there's no hiding what's going on.  Would you really think they'd been treated _fairly_?!?)]]></description>
<dc:subject>in_NB algorithmic_fairness classifiers gordon.geoffrey 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:8e8eb5d2fc4d/</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:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:classifiers"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:gordon.geoffrey"/>
	<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://jmlr.org/papers/v23/20-874.html">
    <title>Model Averaging Is Asymptotically Better Than Model Selection For Prediction</title>
    <dc:date>2022-07-19T13:59:00+00:00</dc:date>
    <link>https://jmlr.org/papers/v23/20-874.html</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We compare the performance of six model average predictors---Mallows' model averaging, stacking, Bayes model averaging, bagging, random forests, and boosting---to the components used to form them.In all six cases we identify conditions under which the model average predictor is consistent for its intended limit and performs as well or better than any of its components asymptotically. This is well known empirically, especially for complex problems, although theoretical results do not seem to have been formally established. We have focused our attention on the regression context since that is where model averaging techniques differ most often from current practice."

--- Could've sworn I bookmarked this already!]]></description>
<dc:subject>in_NB model_selection ensemble_methods regression to_teach:data-mining to_teach:childs_garden_of_statistical_learning_theory</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:9781f6f1983c/</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:model_selection"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:ensemble_methods"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:regression"/>
	<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:childs_garden_of_statistical_learning_theory"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.commerce.senate.gov/services/files/62102355-DC26-4909-BF90-8FB068145F18">
    <title>Testimony before the Senate Subcommittee on Communications, Media, and Broadband Hearing of December 9th, 2021 Algorithmic transparency and assessing effects of algorithmic ranking (Dean Eckles)</title>
    <dc:date>2022-07-11T02:00:52+00:00</dc:date>
    <link>https://www.commerce.senate.gov/services/files/62102355-DC26-4909-BF90-8FB068145F18</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Testimony before the Senate Subcommittee on
Communications, Media, and Broadband
Hearing of December 9th, 2021"]]></description>
<dc:subject>to_read eckles.dean social_media recommender_systems to_teach:data-mining re:actually-dr-internet-is-the-name-of-the-monsters-creator via:henry_farrell</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:687a063d5c63/</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:eckles.dean"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:social_media"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:actually-dr-internet-is-the-name-of-the-monsters-creator"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:henry_farrell"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/1906.04493">
    <title>[1906.04493] Generative Adversarial Networks are Special Cases of Artificial Curiosity (1990) and also Closely Related to Predictability Minimization (1991)</title>
    <dc:date>2022-07-08T13:40:00+00:00</dc:date>
    <link>https://arxiv.org/abs/1906.04493</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["I review unsupervised or self-supervised neural networks playing minimax games in game-theoretic settings: (i) Artificial Curiosity (AC, 1990) is based on two such networks. One network learns to generate a probability distribution over outputs, the other learns to predict effects of the outputs. Each network minimizes the objective function maximized by the other. (ii) Generative Adversarial Networks (GANs, 2010-2014) are an application of AC where the effect of an output is 1 if the output is in a given set, and 0 otherwise. (iii) Predictability Minimization (PM, 1990s) models data distributions through a neural encoder that maximizes the objective function minimized by a neural predictor of the code components. I correct a previously published claim that PM is not based on a minimax game."

--- Last tag is really "to mention in the neural network notes".]]></description>
<dc:subject>neural_networks generative_adversarial_networks your_favorite_deep_neural_network_sucks to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:107a922bade3/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:neural_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:generative_adversarial_networks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:your_favorite_deep_neural_network_sucks"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://logicmag.io/play/my-stepdad's-huge-data-set/">
    <title>My Stepdad’s Huge Data Set</title>
    <dc:date>2022-07-03T04:11:52+00:00</dc:date>
    <link>https://logicmag.io/play/my-stepdad's-huge-data-set/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[--- A well-written and accessible article which I will _not_ be teaching in the data-mining class, because there's just no way that could go well.]]></description>
<dc:subject>have_read pr0n advertising data_mining epidemiology_of_representations practices_relating_to_the_transmission_of_genetic_information to_teach:data-mining sfw</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:9a15a8940499/</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:pr0n"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:advertising"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:epidemiology_of_representations"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:practices_relating_to_the_transmission_of_genetic_information"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:sfw"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.nytimes.com/2022/06/25/technology/china-surveillance-police.html?action=click&amp;module=Well&amp;pgtype=Homepage&amp;section=Business">
    <title>How China Is Policing the Future - The New York Times</title>
    <dc:date>2022-06-28T18:24:38+00:00</dc:date>
    <link>https://www.nytimes.com/2022/06/25/technology/china-surveillance-police.html?action=click&amp;module=Well&amp;pgtype=Homepage&amp;section=Business</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>surveillance data_mining prediction china:prc to_teach:data-mining police</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a70ccc43e029/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:surveillance"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:prediction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:china:prc"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:police"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.aiweirdness.com/interview-with-a-squirrel/">
    <title>Interview with a squirrel</title>
    <dc:date>2022-06-25T17:21:21+00:00</dc:date>
    <link>https://www.aiweirdness.com/interview-with-a-squirrel/</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>funny:geeky funny:malicious artificial_intelligence philip_k_dick_and_the_fake_humans_rules_everything_around_me text_mining natural_language_processing to_teach:data-mining via:?</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:ef35cd9c4100/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:funny:geeky"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:funny:malicious"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:artificial_intelligence"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:philip_k_dick_and_the_fake_humans_rules_everything_around_me"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:text_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:natural_language_processing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:?"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.annualreviews.org/doi/abs/10.1146/annurev-criminol-030920-112506">
    <title>The Impact of Incarceration on Recidivism | Annual Review of Criminology</title>
    <dc:date>2022-06-09T07:54:01+00:00</dc:date>
    <link>https://www.annualreviews.org/doi/abs/10.1146/annurev-criminol-030920-112506</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The US prison population stands at 1.43 million persons, with an additional 740,000 persons in local jails. Nearly all will eventually return to society. This review examines the available evidence on how the experience of incarceration is likely to impact the probability that formerly incarcerated individuals will reoffend. Our focus is on two types of studies, those based on the random assignments of cases to judges, called judge instrumental-variable studies, and those based on discontinuities in sentence severity in sentencing grids, called regression discontinuity studies. Both types of studies are designed to account for selection bias in nonexperimental estimates of the impact of incarceration on reoffending. Most such studies find that the experience of postconviction imprisonment has little impact on the probability of recidivism. A smaller number of studies do, however, find significant effects, both positive and negative. The negative, recidivism-reducing effects are mostly in settings in which rehabilitative programming is emphasized and the positive, criminogenic effects are found in settings in which such programming is not emphasized. The findings of studies of pretrial incarceration are more consistent—most find a deleterious effect on postrelease reoffending. We also conclude that additional work is needed to better understand the heterogeneous effects of incarceration as well as the mechanisms through which incarceration effects, when observed, are generated. For policy, our conclusion of the generally deleterious effect of pretrial detention adds to a larger body of evidence pointing to the social value of limiting its use."

--- The "to_teach" tags are really "to mention, when the inevitable questions about what we're doing modeling the COMPAS/ProPublica data come up".]]></description>
<dc:subject>to:NB crime causal_inference to_teach:data-mining to_teach:statistics_of_inequality_and_discrimination nagin.daniel</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:f5c024c3bf2d/</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:crime"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:causal_inference"/>
	<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:li rdf:resource="https://pinboard.in/u:cshalizi/t:nagin.daniel"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2205.03009">
    <title>[2205.03009] Watching the watchers: bias and vulnerability in remote proctoring software</title>
    <dc:date>2022-05-23T15:01:27+00:00</dc:date>
    <link>https://arxiv.org/abs/2205.03009</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Educators are rapidly switching to remote proctoring and examination software for their testing needs, both due to the COVID-19 pandemic and the expanding virtualization of the education sector. State boards are increasingly utilizing these software for high stakes legal and medical licensing exams. Three key concerns arise with the use of these complex software: exam integrity, exam procedural fairness, and exam-taker security and privacy. We conduct the first technical analysis of each of these concerns through a case study of four primary proctoring suites used in U.S. law school and state attorney licensing exams. We reverse engineer these proctoring suites and find that despite promises of high-security, all their anti-cheating measures can be trivially bypassed and can pose significant user security risks. We evaluate current facial recognition classifiers alongside the classifier used by Examplify, the legal exam proctoring suite with the largest market share, to ascertain their accuracy and determine whether faces with certain skin tones are more readily flagged for cheating. Finally, we offer recommendations to improve the integrity and fairness of the remotely proctored exam experience."

--- As yorksranter says, the fact that in some conditions all of these give error rates above 20% for all groups says that the big problem here isn't _unfairness_, it's _not working well enough to be reliable for anyone_.]]></description>
<dc:subject>to:NB pattern_recognition classifiers algorithmic_fairness to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:697655fc032e/</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:pattern_recognition"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:classifiers"/>
	<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:Bag></taxo:topics>
</item>
<item rdf:about="https://www.nature.com/articles/s41467-019-10933-3/">
    <title>Estimating the success of re-identifications in incomplete datasets using generative models | Nature Communications</title>
    <dc:date>2022-04-13T03:01:09+00:00</dc:date>
    <link>https://www.nature.com/articles/s41467-019-10933-3/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["While rich medical, behavioral, and socio-demographic data are key to modern data-driven research, their collection and use raise legitimate privacy concerns. Anonymizing datasets through de-identification and sampling before sharing them has been the main tool used to address those concerns. We here propose a generative copula-based method that can accurately estimate the likelihood of a specific person to be correctly re-identified, even in a heavily incomplete dataset. On 210 populations, our method obtains AUC scores for predicting individual uniqueness ranging from 0.84 to 0.97, with low false-discovery rate. Using our model, we find that 99.98% of Americans would be correctly re-identified in any dataset using 15 demographic attributes. Our results suggest that even heavily sampled anonymized datasets are unlikely to satisfy the modern standards for anonymization set forth by GDPR and seriously challenge the technical and legal adequacy of the de-identification release-and-forget model."]]></description>
<dc:subject>to:NB privacy data_mining to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:6a95196c6bc1/</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:privacy"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2203.06498">
    <title>[2203.06498] The worst of both worlds: A comparative analysis of errors in learning from data in psychology and machine learning</title>
    <dc:date>2022-03-31T23:35:44+00:00</dc:date>
    <link>https://arxiv.org/abs/2203.06498</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>data_analysis bad_data_analysis psychology data_mining gelman.andrew to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:6c960a97eea4/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_analysis"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bad_data_analysis"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:psychology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:gelman.andrew"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.jmlr.org/papers/v23/20-874.html">
    <title>Model Averaging Is Asymptotically Better Than Model Selection For Prediction</title>
    <dc:date>2022-03-27T15:53:13+00:00</dc:date>
    <link>https://www.jmlr.org/papers/v23/20-874.html</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We compare the performance of six model average predictors---Mallows' model averaging, stacking, Bayes model averaging, bagging, random forests, and boosting---to the components used to form them.In all six cases we identify conditions under which the model average predictor is consistent for its intended limit and performs as well or better than any of its components asymptotically. This is well known empirically, especially for complex problems, although theoretical results do not seem to have been formally established. We have focused our attention on the regression context since that is wheremodel averaging techniques differ most often from current practice."

--- Of course I find this weeks after I teach model averaging.]]></description>
<dc:subject>to:NB ensemble_methods regression to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:5b7a2e7ec1aa/</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:ensemble_methods"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:regression"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2201.10408">
    <title>[2201.10408] Beyond the Frontier: Fairness Without Accuracy Loss</title>
    <dc:date>2022-03-14T18:12:52+00:00</dc:date>
    <link>https://arxiv.org/abs/2201.10408</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Notions of fair machine learning that seek to control various kinds of error across protected groups generally are cast as constrained optimization problems over a fixed model class. For such problems, tradeoffs arise: asking for various kinds of technical fairness requires compromising on overall error, and adding more protected groups increases error rates across all groups. Our goal is to break though such accuracy-fairness tradeoffs.
"We develop a simple algorithmic framework that allows us to deploy models and then revise them dynamically when groups are discovered on which the error rate is suboptimal. Protected groups don't need to be pre-specified: At any point, if it is discovered that there is some group on which our current model performs substantially worse than optimally, then there is a simple update operation that improves the error on that group without increasing either overall error or the error on previously identified groups. We do not restrict the complexity of the groups that can be identified, and they can intersect in arbitrary ways. The key insight that allows us to break through the tradeoff barrier is to dynamically expand the model class as new groups are identified. The result is provably fast convergence to a model that can't be distinguished from the Bayes optimal predictor, at least by those tasked with finding high error groups.
"We explore two instantiations of this framework: as a "bias bug bounty" design in which external auditors are invited to discover groups on which our current model's error is suboptimal, and as an algorithmic paradigm in which the discovery of groups on which the error is suboptimal is posed as an optimization problem. In the bias bounty case, when we say that a model cannot be distinguished from Bayes optimal, we mean by any participant in the bounty program. We provide both theoretical analysis and experimental validation."

--- ETA: Comments http://bactra.org/notebooks/ethics-politics-data-mining.html#beyond-the-frontier]]></description>
<dc:subject>algorithmic_fairness learning_theory roth.aaron to_teach:data-mining kearns.michael in_NB blogged</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a6689f803747/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:learning_theory"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:roth.aaron"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kearns.michael"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:blogged"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.sciencedirect.com/science/article/pii/S016792361400061X?casa_token=8mKjKuwIF58AAAAA:DjRmaJuDoBZjjHb2kA3iEoCckvybsakE7Ww6qdBRABxULXlOuE8FIvmSbgMgYO0ZwLasjyow">
    <title>A data-driven approach to predict the success of bank telemarketing - ScienceDirect</title>
    <dc:date>2022-03-12T13:25:01+00:00</dc:date>
    <link>https://www.sciencedirect.com/science/article/pii/S016792361400061X?casa_token=8mKjKuwIF58AAAAA:DjRmaJuDoBZjjHb2kA3iEoCckvybsakE7Ww6qdBRABxULXlOuE8FIvmSbgMgYO0ZwLasjyow</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[Eye-balling it, the random forest I did for the solutions looks pretty competitive here.

--- Link to my homework assignment:
http://www.stat.cmu.edu/~cshalizi/dm/22/hw/07/hw-07.pdf]]></description>
<dc:subject>data_mining marketing to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:3c63a5c22abb/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:marketing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://archive.ics.uci.edu/ml/datasets/Bank+Marketing">
    <title>UCI Machine Learning Repository: Bank Marketing Data Set</title>
    <dc:date>2022-03-12T13:20:21+00:00</dc:date>
    <link>https://archive.ics.uci.edu/ml/datasets/Bank+Marketing</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[In which I decide to have The Kids act as decision-support for telemarketers.
Notes to self (already in 2022HW7):
- European "term deposit" \approx American "certificate of deposit"
- "nr.employed` is apparently # of employed persons in Portugal, in thousands.
- `euribor3m` = Euro Inter Bank Offer Rate for 3 month loans (not sure if deposit interest rates are formally pegged to this but they should certainly co-vary)
- "variation rate" or "rate of variation" is apparently how you say "percentage growth rate" or "percentage rate of change" in Portuguese and Spanish.
- A little Googling suggests that telemarketers in Lisbon (currently) make in the range of 7--9  Euros.

--- Link to my homework assignment: http://www.stat.cmu.edu/~cshalizi/dm/22/hw/07/hw-07.pdf]]></description>
<dc:subject>data_sets to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:00f7ce3b643a/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_sets"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.nytimes.com/2022/02/19/technology/qanon-messages-authors.html">
    <title>Who Is Behind QAnon? Linguistic Detectives Find Fingerprints - The New York Times</title>
    <dc:date>2022-02-27T03:28:42+00:00</dc:date>
    <link>https://www.nytimes.com/2022/02/19/technology/qanon-messages-authors.html</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[Last tag is probably asking for trouble.]]></description>
<dc:subject>qanon conspiracy_theories psychoceramics stylometrics text_mining natural_language_processing classifiers to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:cf3f4ef95f36/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:qanon"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:conspiracy_theories"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:psychoceramics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:stylometrics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:text_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:natural_language_processing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:classifiers"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://hci.stanford.edu/publications/2019/streetlevelalgorithms/streetlevelalgorithms-chi2019.pdf">
    <title>Street–Level Algorithms: A Theory at the Gaps Between Policy and Decisions</title>
    <dc:date>2022-02-26T19:03:55+00:00</dc:date>
    <link>https://hci.stanford.edu/publications/2019/streetlevelalgorithms/streetlevelalgorithms-chi2019.pdf</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Errors and biases are earning algorithms increasingly malignant reputations in society. A central challenge is that
algorithms must bridge the gap between high–level policy
and on–the–ground decisions, making inferences in novel
situations where the policy or training data do not readily
apply. In this paper, we draw on the theory of street–level
bureaucracies, how human bureaucrats such as police and
judges interpret policy to make on–the–ground decisions.
We present by analogy a theory of street–level algorithms,
the algorithms that bridge the gaps between policy and decisions about people in a socio-technical system. We argue that
unlike street–level bureaucrats, who reflexively refine their
decision criteria as they reason through a novel situation,
street–level algorithms at best rene their criteria only after
the decision is made. This loop–and–a–half delay results in
illogical decisions when handling new or extenuating circumstances. This theory suggests designs for street–level
algorithms that draw on historical design patterns for street–
level bureaucracies, including mechanisms for self–policing
and recourse in the case of error."

--- ETA after reading: I like the framing, mostly, and I appreciate the point about "reflexively" considering the decision boundary before making a decision vs. at best updating with feedback after the fact.  But even then, what humans really do, _sometimes_, is alter the rule based on what they _imagine_ the consequences of a decision would be, in light of what they _conceive_ the purpose of the behavior to be.  I emphasize the subjective, mental terms, because matching this, or coming close, is perilously close to AI-complete.

Further unfair complaints:
- A persistent conflation of "marginal", in the sense of lower-status social categories, with "marginal", in the sense of "marginal case", i.e., one near a decision boundary.
- Somewhat uncritical presentations of the case studies:
    + No references are given for the algorithmic reasons why videos with "transgender" in the title flipped YouTube's de-monetization switch.  It _could_ be that those algorithms somehow encodeed "gender = sex (noun) = sex (verb)".  But, well, there's a hell of a lot of TG porn online (proof: omitted), and while this would still be a kind of stupidity on the part of the YouTube algorithm, and indeed an instance of data-set shift, it'd be one with a rather different valence.  (If this conjecture is even close to right, people making videos about dealing with the trauma of incest were probably also de-monetized.)
    + Similarly, in disputes about crowd-sourcing, they very plainly take the side of the workers, not those paying for work.  Now my bias, too, is to always side with the workers, but if you're really going to do critical social science, you need to at least peer into that bias.  (It's quite possible that existing practices are bad for both sides of the worker-employer divide!)
- I do like the idea that when people ask for recourse, the system should provide the representations (they say "embeddings", but I forgive them) for similar cases.  I'd amend this to a selection of the most similar cases with the same outcome as the one being appealed, and the most similar cases with different outcomes.  ("Similar" needs specification here, yes.)  However, the presentation of the appeal/recourse process seems to presume that recourse will be granted --- it would instead make more sense to allow for the possibility that the system got it right the first time, and whatever the complainant alleges as distinguishing special features of their case are _properly_ ignored.]]></description>
<dc:subject>via:henry_farrell classifiers algorithmic_fairness to_teach:data-mining decision-making bureaucracy have_read in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:4aa7f6ce05aa/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:henry_farrell"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:classifiers"/>
	<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:decision-making"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bureaucracy"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://press.princeton.edu/books/paperback/9780691177311/experiments-of-the-mind">
    <title>Experiments of the Mind | Princeton University Press</title>
    <dc:date>2022-02-05T21:05:34+00:00</dc:date>
    <link>https://press.princeton.edu/books/paperback/9780691177311/experiments-of-the-mind</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>books:noted to_teach:data-mining downloaded</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:eb137eaf7346/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:books:noted"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:downloaded"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.pnas.org/content/118/22/e2018340118">
    <title>Algorithmic monoculture and social welfare | PNAS</title>
    <dc:date>2022-02-01T15:21:45+00:00</dc:date>
    <link>https://www.pnas.org/content/118/22/e2018340118</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["As algorithms are increasingly applied to screen applicants for high-stakes decisions in employment, lending, and other domains, concerns have been raised about the effects of algorithmic monoculture, in which many decision-makers all rely on the same algorithm. This concern invokes analogies to agriculture, where a monocultural system runs the risk of severe harm from unexpected shocks. Here, we show that the dangers of algorithmic monoculture run much deeper, in that monocultural convergence on a single algorithm by a group of decision-making agents, even when the algorithm is more accurate for any one agent in isolation, can reduce the overall quality of the decisions being made by the full collection of agents. Unexpected shocks are therefore not needed to expose the risks of monoculture; it can hurt accuracy even under “normal” operations and even for algorithms that are more accurate when used by only a single decision-maker. Our results rely on minimal assumptions and involve the development of a probabilistic framework for analyzing systems that use multiple noisy estimates of a set of alternatives."

--- Unfair reaction before reading: Ain't this just Hong and Page (2004), also PNAS?]]></description>
<dc:subject>to:NB to_read kleinberg.jon diversity algorithmic_fairness via:? to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:28f02d0bd157/</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:kleinberg.jon"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:diversity"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:?"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2104.04103">
    <title>[2104.04103] Causal Decision Making and Causal Effect Estimation Are Not the Same... and Why It Matters</title>
    <dc:date>2022-01-24T17:27:24+00:00</dc:date>
    <link>https://arxiv.org/abs/2104.04103</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Causal decision making (CDM) based on machine learning has become a routine part of business. Businesses algorithmically target offers, incentives, and recommendations to affect consumer behavior. Recently, we have seen an acceleration of research related to CDM and causal effect estimation (CEE) using machine-learned models. This article highlights an important perspective: CDM is not the same as CEE, and counterintuitively, accurate CEE is not necessary for accurate CDM. Our experience is that this is not well understood by practitioners or most researchers. Technically, the estimand of interest is different, and this has important implications both for modeling and for the use of statistical models for CDM. We draw on prior research to highlight three implications. (1) We should consider carefully the objective function of the causal machine learning, and if possible, optimize for accurate treatment assignment rather than for accurate effect-size estimation. (2) Confounding does not have the same effect on CDM as it does on CEE. The upshot is that for supporting CDM it may be just as good or even better to learn with confounded data as with unconfounded data. Finally, (3) causal statistical modeling may not be necessary to support CDM because a proxy target for statistical modeling might do as well or better. This third observation helps to explain at least one broad common CDM practice that seems wrong at first blush: the widespread use of non-causal models for targeting interventions. The last two implications are particularly important in practice, as acquiring (unconfounded) data on all counterfactuals can be costly and often impracticable. These observations open substantial research ground. We hope to facilitate research in this area by pointing to related articles from multiple contributing fields, including two dozen articles published the last three to four years."

--- ETA after reading: Think of decision-making as a classification problem, rather than estimation.  If your classifier mis-estimates Pr(Y|X=x), but you're nonetheless on the correct side of 1/2 (or whatever your optimal boundary might be), it doesn't matter for classification accuracy!  So if you over-estimate the benefits of treatment for those you decide to treat, well, you're still treating them...]]></description>
<dc:subject>causal_inference decision-making via:vaguery to_teach:data-mining provost.foster have_read blogged in_NB</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:4546e3e7d5f3/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:causal_inference"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:decision-making"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:vaguery"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:provost.foster"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:blogged"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.quantamagazine.org/computer-scientists-discover-limits-of-major-research-algorithm-20210817/">
    <title>Computer Scientists Discover Limits of Major Research Algorithm | Quanta Magazine</title>
    <dc:date>2021-12-13T06:49:36+00:00</dc:date>
    <link>https://www.quantamagazine.org/computer-scientists-discover-limits-of-major-research-algorithm-20210817/</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA[Computational complexity of gradient descent.]]></description>
<dc:subject>optimization popular_science to_teach:data-mining have_read to_teach:childs_garden_of_statistical_learning_theory</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:c65c29d817a3/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:optimization"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:popular_science"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:childs_garden_of_statistical_learning_theory"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.washingtonpost.com/technology/interactive/2021/how-facebook-algorithm-works?no_nav=true">
    <title>Here’s how Facebook’s algorithm works - Washington Post</title>
    <dc:date>2021-10-27T13:52:30+00:00</dc:date>
    <link>https://www.washingtonpost.com/technology/interactive/2021/how-facebook-algorithm-works?no_nav=true</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>to_teach:data-mining recommender_systems 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:ddda55fbd673/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<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.pnas.org/content/118/32/e2101967118.short">
    <title>Examining the consumption of radical content on YouTube | PNAS</title>
    <dc:date>2021-08-12T00:29:29+00:00</dc:date>
    <link>https://www.pnas.org/content/118/32/e2101967118.short</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Although it is under-studied relative to other social media platforms, YouTube is arguably the largest and most engaging online media consumption platform in the world. Recently, YouTube’s scale has fueled concerns that YouTube users are being radicalized via a combination of biased recommendations and ostensibly apolitical “anti-woke” channels, both of which have been claimed to direct attention to radical political content. Here we test this hypothesis using a representative panel of more than 300,000 Americans and their individual-level browsing behavior, on and off YouTube, from January 2016 through December 2019. Using a labeled set of political news channels, we find that news consumption on YouTube is dominated by mainstream and largely centrist sources. Consumers of far-right content, while more engaged than average, represent a small and stable percentage of news consumers. However, consumption of “anti-woke” content, defined in terms of its opposition to progressive intellectual and political agendas, grew steadily in popularity and is correlated with consumption of far-right content off-platform. We find no evidence that engagement with far-right content is caused by YouTube recommendations systematically, nor do we find clear evidence that anti-woke channels serve as a gateway to the far right. Rather, consumption of political content on YouTube appears to reflect individual preferences that extend across the web as a whole."

--- Same as [https://arxiv.org/abs/2011.12843] or subtly different?]]></description>
<dc:subject>to:NB to_read re:actually-dr-internet-is-the-name-of-the-monsters-creator recommender_systems to_teach:data-mining kith_and_kin clauset.aaron watts.duncan</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:d6a1015a7bb5/</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:re:actually-dr-internet-is-the-name-of-the-monsters-creator"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<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:watts.duncan"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.annualreviews.org/doi/abs/10.1146/annurev-soc-090820-020800">
    <title>The Society of Algorithms | Annual Review of Sociology</title>
    <dc:date>2021-08-03T04:35:55+00:00</dc:date>
    <link>https://www.annualreviews.org/doi/abs/10.1146/annurev-soc-090820-020800</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The pairing of massive data sets with processes—or algorithms—written in computer code to sort through, organize, extract, or mine them has made inroads in almost every major social institution. This article proposes a reading of the scholarly literature concerned with the social implications of this transformation. First, we discuss the rise of a new occupational class, which we call the coding elite. This group has consolidated power through their technical control over the digital means of production and by extracting labor from a newly marginalized or unpaid workforce, the cybertariat. Second, we show that the implementation of techniques of mathematical optimization across domains as varied as education, medicine, credit and finance, and criminal justice has intensified the dominance of actuarial logics of decision-making, potentially transforming pathways to social reproduction and mobility but also generating a pushback by those so governed. Third, we explore how the same pervasive algorithmic intermediation in digital communication is transforming the way people interact, associate, and think. We conclude by cautioning against the wildest promises of artificial intelligence but acknowledging the increasingly tight coupling between algorithmic processes, social structures, and subjectivities."]]></description>
<dc:subject>to:NB data_mining networked_life to_teach:data-mining fourcade.marion</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:08a0ff6c918b/</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:data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networked_life"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:fourcade.marion"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.technologyreview.com/2021/07/07/1027916/we-tested-ai-interview-tools/?_hsmi=140522126&amp;_hsenc=p2ANqtz-8vjTOrf7wPGyBHzGvNM6HAsRT9_ivd6OAndeGlgd7q_DeKQzL8wzhIWUxeUBctrU37fL6-4CUvXLmfyBLOvUsDCnCCddi0lNRwTMkkbRP71Z9mu90">
    <title>We tested AI interview tools. Here’s what we found. | MIT Technology Review</title>
    <dc:date>2021-07-26T15:07:26+00:00</dc:date>
    <link>https://www.technologyreview.com/2021/07/07/1027916/we-tested-ai-interview-tools/?_hsmi=140522126&amp;_hsenc=p2ANqtz-8vjTOrf7wPGyBHzGvNM6HAsRT9_ivd6OAndeGlgd7q_DeKQzL8wzhIWUxeUBctrU37fL6-4CUvXLmfyBLOvUsDCnCCddi0lNRwTMkkbRP71Z9mu90</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Our candidate turned to MyInterview and repeated the experiment. She read the same Wikipedia entry aloud in German. The algorithm not only returned a personality assessment, but it also predicted our candidate to be a 73% match for the fake job, putting her in the top half of all the applicants we had asked to apply."]]></description>
<dc:subject>to_teach:data-mining utter_stupidity via:yorksranter bad_data_analysis</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:7a8fd8909696/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:utter_stupidity"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:via:yorksranter"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bad_data_analysis"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://nymag.com/intelligencer/2020/09/inside-palantir-technologies-peter-thiel-alex-karp.html">
    <title>Inside Palantir, Silicon Valley’s Most Secretive Unicorn</title>
    <dc:date>2021-07-15T17:49:26+00:00</dc:date>
    <link>https://nymag.com/intelligencer/2020/09/inside-palantir-technologies-peter-thiel-alex-karp.html</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Rooms Full of People" is good.]]></description>
<dc:subject>data_mining national_surveillance_state to_teach:data-mining have_read</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a0a558670d10/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:national_surveillance_state"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.nature.com/articles/s41386-021-01020-7">
    <title>Systematic misestimation of machine learning performance in neuroimaging studies of depression | Neuropsychopharmacology</title>
    <dc:date>2021-06-11T18:03:12+00:00</dc:date>
    <link>https://www.nature.com/articles/s41386-021-01020-7</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["We currently observe a disconcerting phenomenon in machine learning studies in psychiatry: While we would expect larger samples to yield better results due to the availability of more data, larger machine learning studies consistently show much weaker performance than the numerous small-scale studies. Here, we systematically investigated this effect focusing on one of the most heavily studied questions in the field, namely the classification of patients suffering from Major Depressive Disorder (MDD) and healthy controls based on neuroimaging data. Drawing upon structural MRI data from a balanced sample of N = 1868 MDD patients and healthy controls from our recent international Predictive Analytics Competition (PAC), we first trained and tested a classification model on the full dataset which yielded an accuracy of 61%. Next, we mimicked the process by which researchers would draw samples of various sizes (N = 4 to N = 150) from the population and showed a strong risk of misestimation. Specifically, for small sample sizes (N = 20), we observe accuracies of up to 95%. For medium sample sizes (N = 100) accuracies up to 75% were found. Importantly, further investigation showed that sufficiently large test sets effectively protect against performance misestimation whereas larger datasets per se do not. While these results question the validity of a substantial part of the current literature, we outline the relatively low-cost remedy of larger test sets, which is readily available in most cases."

--- I haven't read the paper yet so there might be alternative explanations, but I can't help noting that this is 100% consistent with the most cynical possible interpretation of [http://bactra.org/weblog/698.html].]]></description>
<dc:subject>to:NB neural_data_analysis statistics classifiers to_teach:data-mining re:neutral_model_of_inquiry data_mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:a31d8ee83a33/</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_data_analysis"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:classifiers"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:neutral_model_of_inquiry"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://bigthink.com/laurie-vazquez/are-you-a-geek-or-a-nerd">
    <title>What's the Difference Between a Geek and a Nerd? Not Much, According to the Data - Big Think</title>
    <dc:date>2021-06-04T04:01:37+00:00</dc:date>
    <link>https://bigthink.com/laurie-vazquez/are-you-a-geek-or-a-nerd</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>text_mining geekdom nerdworld to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:17d81b4fb072/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:text_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:geekdom"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:nerdworld"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://doxa.substack.com/p/phrenology-insurance-claims-and-digital?token=eyJ1c2VyX2lkIjozMTk2MjUwOSwicG9zdF9pZCI6MzcxODg4MDYsIl8iOiI5RlR3eiIsImlhdCI6MTYyMjc3NzYxOSwiZXhwIjoxNjIyNzgxMjE5LCJpc3MiOiJwdWItMjM5NjUzIiwic3ViIjoicG9zdC1yZWFjdGlvbiJ9.r6AUl5BSCgCU4aMl6zL1Pt8xcC3cnNU5E7J6LDGlQbs">
    <title>Phrenology, insurance claims, and digital gaydar - doxa</title>
    <dc:date>2021-06-04T03:34:23+00:00</dc:date>
    <link>https://doxa.substack.com/p/phrenology-insurance-claims-and-digital?token=eyJ1c2VyX2lkIjozMTk2MjUwOSwicG9zdF9pZCI6MzcxODg4MDYsIl8iOiI5RlR3eiIsImlhdCI6MTYyMjc3NzYxOSwiZXhwIjoxNjIyNzgxMjE5LCJpc3MiOiJwdWItMjM5NjUzIiwic3ViIjoicG9zdC1yZWFjdGlvbiJ9.r6AUl5BSCgCU4aMl6zL1Pt8xcC3cnNU5E7J6LDGlQbs</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>classifiers algorithmic_fairness to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:903e4f804855/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:classifiers"/>
	<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:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2105.09065">
    <title>[2105.09065] Statistical Learning for Best Practices in Tattoo Removal</title>
    <dc:date>2021-05-20T21:22:32+00:00</dc:date>
    <link>https://arxiv.org/abs/2105.09065</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["The causes behind complications in laser-assisted tattoo removal are currently not well understood, and in the literature relating to tattoo removal the emphasis on removal treatment is on removal technologies and tools, not best parameters involved in the treatment process. Additionally, the very challenge of determining best practices is difficult given the complexity of interactions between factors that may correlate to these complications. In this paper we apply a battery of classical statistical methods and techniques to identify features that may be closely correlated to causes of complication during the tattoo removal process, and report quantitative evidence for potential best practices. We develop elementary statistical descriptions of tattoo data collected by the largest gang rehabilitation and reentry organization in the world, Homeboy Industries; perform parametric and nonparametric tests of significance; and finally, produce a statistical model explaining treatment parameter interactions, as well as develop a ranking system for treatment parameters utilizing bootstrapping and gradient boosting."]]></description>
<dc:subject>to:NB statistics tattoos to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:182f73b992e8/</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:statistics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:tattoos"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://doi.org/10.1111/rssb.12425">
    <title>AMF: Aggregated Mondrian forests for online learning - Mourtada - - Journal of the Royal Statistical Society: Series B (Statistical Methodology) - Wiley Online Library</title>
    <dc:date>2021-05-20T13:53:05+00:00</dc:date>
    <link>https://doi.org/10.1111/rssb.12425</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Random forest (RF) is one of the algorithms of choice in many supervised learning applications, be it classification or regression. The appeal of such tree-ensemble methods comes from a combination of several characteristics: a remarkable accuracy in a variety of tasks, a small number of parameters to tune, robustness with respect to features scaling, a reasonable computational cost for training and prediction, and their suitability in high-dimensional settings. The most commonly used RF variants, however, are ‘offline’ algorithms, which require the availability of the whole dataset at once. In this paper, we introduce AMF, an online RF algorithm based on Mondrian Forests. Using a variant of the context tree weighting algorithm, we show that it is possible to efficiently perform an exact aggregation over all prunings of the trees; in particular, this enables to obtain a truly online parameter-free algorithm which is competitive with the optimal pruning of the Mondrian tree, and thus adaptive to the unknown regularity of the regression function. Numerical experiments show that AMF is competitive with respect to several strong baselines on a large number of datasets for multi-class classification."]]></description>
<dc:subject>to:NB to_read ensemble_methods random_forests regression classifiers to_teach:data-mining online_learning statistics</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:42bd56d40bd2/</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:ensemble_methods"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:random_forests"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:regression"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:classifiers"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:online_learning"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:statistics"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.nber.org/papers/w28811">
    <title>AI Adoption and System-Wide Change | NBER</title>
    <dc:date>2021-05-17T14:40:30+00:00</dc:date>
    <link>https://www.nber.org/papers/w28811</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Analyses of AI adoption focus on its adoption at the individual task level. What has received significantly less attention is how AI adoption is shaped by the fact that organisations are composed of many interacting tasks. AI adoption may, therefore, require system-wide change which is both a constraint and an opportunity. We provide the first formal analysis where multiple tasks may be part of a modular or non-modular system. We find that reliance on AI, a prediction tool, increases decision variation which, in turn, raises challenges if decisions across the organisation interact. Modularity, which leads to task independence rather than system-level inter-dependencies, softens that impact. Thus, modularity can facilitate AI adoption. However, it does this at the expense of synergies. By contrast, when there are mechanisms for inter-decision coordination, AI adoption is enhanced when there is a non-modular environment. Consequently, we show that there are important cases where AI adoption will be enhanced when it can be adopted beyond tasks but as part of a designed organisational system."

--- Deliberately not tagged "artificial_intelligence".]]></description>
<dc:subject>to:NB economics data_mining to_teach:data-mining organizations</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:29d9f6d466a9/</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:economics"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:organizations"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.sup.org/books/title/?id=32597">
    <title>The AI Marketing Canvas: A Five-Stage Road Map to Implementing Artificial Intelligence in Marketing | Raj Venkatesan and Jim Lecinski</title>
    <dc:date>2021-05-17T13:01:26+00:00</dc:date>
    <link>https://www.sup.org/books/title/?id=32597</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["This book offers a direct, actionable plan CMOs can use to map out initiatives that are properly sequenced and designed for success—regardless of where their marketing organization is in the process.
"The authors pose the following critical questions to marketers: (1) How should modern marketers be thinking about artificial intelligence and machine learning? and (2) How should marketers be developing a strategy and plan to implement AI into their marketing toolkit?
"The opening chapters provide marketing leaders with an overview of what exactly AI is and how is it different than traditional computer science approaches. Venkatesan and Lecinski, then, propose a best-practice, five-stage framework for implementing what they term the "AI Marketing Canvas." Their approach is based on research and interviews they conducted with leading marketers, and offers many tangible examples of what brands are doing at each stage of the AI Marketing Canvas. By way of guidance, Venkatesan and Lecinski provide examples of brands—including Google, Lyft, Ancestry.com, and Coca-Cola—that have successfully woven AI into their marketing strategies. The book concludes with a discussion of important implications for marketing leaders—for your team and culture."

--- I am quite sure that I would find this book equally horrifying in aims, content and style, but I am also equally sure that a fair chunk of my students will end up working for these people.]]></description>
<dc:subject>in_NB books:noted marketing data_mining deceiving_us_has_become_an_industrial_process to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:be76d8874382/</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:marketing"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:data_mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:deceiving_us_has_become_an_industrial_process"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2105.04648">
    <title>[2105.04648] Joint Fairness Model with Applications to Risk Predictions for Under-represented Populations</title>
    <dc:date>2021-05-12T18:29:05+00:00</dc:date>
    <link>https://arxiv.org/abs/2105.04648</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Under-representation of certain populations, based on gender, race/ethnicity, and age, in data collection for predictive modeling may yield less-accurate predictions for the under-represented groups. Recently, this issue of fairness in predictions has attracted significant attention, as data-driven models are increasingly utilized to perform crucial decision-making tasks. Methods to achieve fairness in the machine learning literature typically build a single prediction model subject to some fairness criteria in a manner that encourages fair prediction performances for all groups. These approaches have two major limitations: i) fairness is often achieved by compromising accuracy for some groups; ii) the underlying relationship between dependent and independent variables may not be the same across groups. We propose a Joint Fairness Model (JFM) approach for binary outcomes that estimates group-specific classifiers using a joint modeling objective function that incorporates fairness criteria for prediction. We introduce an Accelerated Smoothing Proximal Gradient Algorithm to solve the convex objective function, and demonstrate the properties of the proposed JFM estimates. Next, we presented the key asymptotic properties for the JFM parameter estimates. We examined the efficacy of the JFM approach in achieving prediction performances and parities, in comparison with the Single Fairness Model, group-separate model, and group-ignorant model through extensive simulations. Finally, we demonstrated the utility of the JFM method in the motivating example to obtain fair risk predictions for under-represented older patients diagnosed with coronavirus disease 2019 (COVID-19)."]]></description>
<dc:subject>prediction classifiers algorithmic_fairness smyth.padhraic to_teach:data-mining in_NB re:codename:one_law_for_the_lion_and_ox_is_oppression</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:9778e5e6bcab/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:prediction"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:classifiers"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:algorithmic_fairness"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:smyth.padhraic"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:in_NB"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:re:codename:one_law_for_the_lion_and_ox_is_oppression"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2105.04134">
    <title>[2105.04134] Bagging cross-validated bandwidth selection in nonparametric regression estimation with applications to large-sized samples</title>
    <dc:date>2021-05-12T18:15:40+00:00</dc:date>
    <link>https://arxiv.org/abs/2105.04134</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Cross-validation is a well-known and widely used bandwidth selection method in nonparametric regression estimation. However, this technique has two remarkable drawbacks: (i) the large variability of the selected bandwidths, and (ii) the inability to provide results in a reasonable time for very large sample sizes. To overcome these problems, bagging cross-validation bandwidths are analyzed in this paper. This approach consists in computing the cross-validation bandwidths for a finite number of subsamples and then rescaling the averaged smoothing parameters to the original sample size. Under a random-design regression model, asymptotic expressions up to a second-order for the bias and variance of the leave-one-out cross-validation bandwidth for the Nadaraya--Watson estimator are obtained. Subsequently, the asymptotic bias and variance and the limit distribution are derived for the bagged cross-validation selector. Suitable choices of the number of subsamples and the subsample size lead to an n−1/2 rate for the convergence in distribution of the bagging cross-validation selector, outperforming the rate n−3/10 of leave-one-out cross-validation. Several simulations and an illustration on a real dataset related to the COVID-19 pandemic show the behavior of our proposal and its better performance, in terms of statistical efficiency and computing time, when compared to leave-one-out cross-validation."]]></description>
<dc:subject>to:NB cross-validation ensemble_methods bootstrap to_teach:data-mining kernel_smoothing</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:f2a3494ec6a3/</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:cross-validation"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:ensemble_methods"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:bootstrap"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kernel_smoothing"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="http://simondedeo.com/?p=705">
    <title>The 11th Reason to Delete your Social Media Account: the Algorithm will Find You – Axiom of Chance</title>
    <dc:date>2021-05-03T19:56:22+00:00</dc:date>
    <link>http://simondedeo.com/?p=705</link>
    <dc:creator>cshalizi</dc:creator><dc:subject>networked_life moral_psychology recommender_systems kith_and_kin dedeo.simon have_read to_teach:data-mining</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:63530f9ce085/</dc:identifier>
<taxo:topics><rdf:Bag>	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:networked_life"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:moral_psychology"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:kith_and_kin"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:dedeo.simon"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:have_read"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://www.ucpress.edu/book/9780520382046/netflix-recommends">
    <title>Netflix Recommends by Mattias Frey - Paperback - University of California Press</title>
    <dc:date>2021-05-02T12:41:49+00:00</dc:date>
    <link>https://www.ucpress.edu/book/9780520382046/netflix-recommends</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Algorithmic recommender systems, deployed by media companies to suggest content based on users’ viewing histories, have inspired hopes for personalized, curated media but also dire warnings of filter bubbles and media homogeneity. Curiously, both proponents and detractors assume that recommender systems are novel, effective, and widely used methods to choose films and series. Scrutinizing the world’s most subscribed streaming service, Netflix, this book challenges that consensus. Investigating real-life users, marketing rhetoric, technical processes, business models, and historical antecedents, Mattias Frey demonstrates that these choice aids are neither as revolutionary nor alarming as their celebrants and critics maintain—and neither as trusted nor widely used. Netflix Recommends illustrates the constellations of sources that real viewers use to choose films and series in the digital age and argues that although some lament AI’s hostile takeover of humanistic cultures, the thirst for filters, curators, and critics is stronger than ever."]]></description>
<dc:subject>to:NB books:noted recommender_systems to_teach:data-mining books:in_library books:have_suggested_to_library</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:cab062a9cab3/</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:recommender_systems"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:books:in_library"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:books:have_suggested_to_library"/>
</rdf:Bag></taxo:topics>
</item>
<item rdf:about="https://arxiv.org/abs/2003.00381">
    <title>[2003.00381] Statistical power for cluster analysis</title>
    <dc:date>2021-04-22T15:32:01+00:00</dc:date>
    <link>https://arxiv.org/abs/2003.00381</link>
    <dc:creator>cshalizi</dc:creator><description><![CDATA["Cluster algorithms are gaining in popularity due to their compelling ability to identify discrete subgroups in data, and their increasing accessibility in mainstream programming languages and statistical software. While researchers can follow guidelines to choose the right algorithms, and to determine what constitutes convincing clustering, there are no firmly established ways of computing a priori statistical power for cluster analysis. Here, we take a simulation approach to estimate power and classification accuracy for popular analysis pipelines. We systematically varied cluster size, number of clusters, number of different features between clusters, effect size within each different feature, and cluster covariance structure in generated datasets. We then subjected these datasets to common dimensionality reduction approaches (none, multi-dimensional scaling, or uniform manifold approximation and projection) and cluster algorithms (k-means, hierarchical agglomerative clustering with Ward linkage and Euclidean distance, or average linkage and cosine distance, HDBSCAN). Furthermore, we simulated additional datasets to explore the effect of sample size and cluster separation on statistical power and classification accuracy. We found that clustering outcomes were driven by large effect sizes or the accumulation of many smaller effects across features, and were mostly unaffected by differences in covariance structure. Sufficient statistical power can be achieved with relatively small samples (N=20 per subgroup), provided cluster separation is large ({\Delta}=4). Finally, we discuss whether fuzzy clustering (c-means) could provide a more parsimonious alternative for identifying separable multivariate normal distributions, particularly those with lower centroid separation."

--- I'll be interested to see if they look at what happens when there are no clusters,...]]></description>
<dc:subject>to:NB clustering to_teach:data-mining color_me_skeptical</dc:subject>
<dc:source>https://pinboard.in/</dc:source>
<dc:identifier>https://pinboard.in/u:cshalizi/b:6df8375fa239/</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:clustering"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:to_teach:data-mining"/>
	<rdf:li rdf:resource="https://pinboard.in/u:cshalizi/t:color_me_skeptical"/>
</rdf:Bag></taxo:topics>
</item>
</rdf:RDF>