{"id":24,"date":"2019-01-31T21:43:01","date_gmt":"2019-01-31T21:43:01","guid":{"rendered":"http:\/\/groups.cs.umass.edu\/equate\/?p=24"},"modified":"2020-10-05T21:15:49","modified_gmt":"2020-10-05T21:15:49","slug":"safe-and-fair-machine-learning","status":"publish","type":"post","link":"https:\/\/groups.cs.umass.edu\/equate\/research\/safe-and-fair-machine-learning","title":{"rendered":"Safe and Fair Machine Learning"},"content":{"rendered":"<p><span style=\"font-weight: 400\">In this project we study how the user of a machine learning<\/span>\u00a0<span style=\"font-weight: 400\">(ML) algorithm (method) can place constraints on the algorithm&#8217;s behavior. We contend that standard ML algorithms are not user-friendly, in that they can require ML and data science expertise to apply responsibly to real-world applications. We present a new type of ML algorithm that shifts, from the user of the algorithm to the researcher who designs the algorithm, many of the challenges associated with ensuring that the ML method is safe to use. The resulting algorithms provide a simple interface for specifying what constitutes undesirable behavior of the ML algorithm, and provide high-probability guarantees that it will not produce this undesirable behavior.<\/span><\/p>\n<p><!--more--><\/p>\n<h4>Publications<\/h4>\n<ul>\n<li>Yair Zick, Reza Shokri, Martin Stobel. <a href=\"https:\/\/arxiv.org\/abs\/1907.00164\" target=\"_blank\" rel=\"noopener noreferrer\">Privacy Risks of Explaining Machine Learning Models<\/a>.<\/li>\n<li id=\"contrib-1\" class=\"contributor\"><span class=\"name\">Philip S. Thomas,\u00a0<\/span><a id=\"xref-award-group-5-1\" class=\"xref-award\" href=\"https:\/\/science.sciencemag.org\/content\/366\/6468\/999#award-group-5\"><\/a><span class=\"name\">Bruno Castro da Silva,\u00a0<\/span><span class=\"name\">Andrew G. Barto,\u00a0<\/span><span class=\"name\">Stephen Giguere,\u00a0<\/span><span class=\"name\">Yuriy Brun,\u00a0<\/span><a id=\"xref-award-group-2-1\" class=\"xref-award\" href=\"https:\/\/science.sciencemag.org\/content\/366\/6468\/999#award-group-2\"><\/a><a id=\"xref-award-group-3-1\" class=\"xref-award\" href=\"https:\/\/science.sciencemag.org\/content\/366\/6468\/999#award-group-3\"><\/a><span class=\"name\"><span class=\"name\">Emma Brunskill.\u00a0<\/span><\/span><a id=\"xref-award-group-1-1\" class=\"xref-award\" href=\"https:\/\/science.sciencemag.org\/content\/366\/6468\/999#award-group-1\"><\/a><a href=\"https:\/\/science.sciencemag.org\/content\/366\/6468\/999\" target=\"_blank\" rel=\"noopener noreferrer\">Preventing undesirable behavior of intelligent machines<\/a>.<a id=\"xref-award-group-4-1\" class=\"xref-award\" href=\"https:\/\/science.sciencemag.org\/content\/366\/6468\/999#award-group-4\"><\/a><\/li>\n<li>Philip S. Thomas, Bruno Castro da Silva, Andrew G. Barto, Emma Brunskill. <a href=\"https:\/\/arxiv.org\/abs\/1708.05448\">On Ensuring that Intelligent Machines Are Well-Behaved<\/a>.<\/li>\n<li>Przemyslaw A. Grabowicz, Kenta Takatsu, Luis F. Lafuerza.\u00a0<span style=\"font-size: inherit\"><a href=\"https:\/\/arxiv.org\/abs\/1912.08189\" target=\"_blank\" rel=\"noopener noreferrer\">Supervised learning algorithms resilient to discriminatory data perturbations<\/a>.<\/span><\/li>\n<li>B. Metevier, S. Giguere, S. Brockman, A. Kobren, Y. Brun, E. Brunskill, and P. S. Thomas. <a href=\"https:\/\/people.cs.umass.edu\/~pthomas\/papers\/Metevier2019.pdf\" target=\"_blank\" rel=\"noopener noreferrer\">Offline Contextual Bandits with High Probability Fairness Guarantees.<\/a> In <i>Advances in Neural Information Processing Systems<\/i>, 2019.<\/li>\n<\/ul>\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>In this project we study how the user of a machine learning\u00a0(ML) algorithm (method) can place constraints on the algorithm&#8217;s behavior. We contend that standard ML algorithms are not user-friendly, in that they can require ML and data science expertise to apply responsibly to real-world applications. We present a new type of ML algorithm that [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[4,2],"tags":[],"class_list":["post-24","post","type-post","status-publish","format-standard","hentry","category-featured","category-research"],"_links":{"self":[{"href":"https:\/\/groups.cs.umass.edu\/equate\/wp-json\/wp\/v2\/posts\/24","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/groups.cs.umass.edu\/equate\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/groups.cs.umass.edu\/equate\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/groups.cs.umass.edu\/equate\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/groups.cs.umass.edu\/equate\/wp-json\/wp\/v2\/comments?post=24"}],"version-history":[{"count":11,"href":"https:\/\/groups.cs.umass.edu\/equate\/wp-json\/wp\/v2\/posts\/24\/revisions"}],"predecessor-version":[{"id":190,"href":"https:\/\/groups.cs.umass.edu\/equate\/wp-json\/wp\/v2\/posts\/24\/revisions\/190"}],"wp:attachment":[{"href":"https:\/\/groups.cs.umass.edu\/equate\/wp-json\/wp\/v2\/media?parent=24"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/groups.cs.umass.edu\/equate\/wp-json\/wp\/v2\/categories?post=24"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/groups.cs.umass.edu\/equate\/wp-json\/wp\/v2\/tags?post=24"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}