On the Relation between Accuracy and Fairness in Binary Classification

<!DOCTYPE html> <html lang="en" class="no-js"> <head> <meta charset="utf-8"> <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> <link rel="icon" href="/static/img/favicon.ico" /> <link rel="apple-touch-icon" href="/static/img/apple-touch-icon.png"> <title>On the Relation between Accuracy and Fairness in Binary Classification :: FAT ML</title> <meta name="viewport" content="width=device-width, initial-scale=1.0"> <link href="https://fonts.googleapis.com/css?family=Open+Sans:400,400i,700|Roboto+Slab:400,700" rel="stylesheet"> <link href="/static/sass/main.css" rel="stylesheet" type="text/css" />  <script> (function(H){ H.className = H.className.replace(/\bno-js\b/,'js') })(document.documentElement); </script> </head> <body class="standard-page"> <div class="site-bg"> <nav class="site-header navbar navbar-default"> <div class="container"> <div class="navbar-header"> <button type="button" class="navbar-toggle collapsed" data-toggle="collapse" data-target="#site-header-nav" aria-expanded="false"> <span class="sr-only">Toggle navigation</span> <span class="icon-bar"></span> <span class="icon-bar"></span> <span class="icon-bar"></span> </button> <a class="navbar-brand" href="/"> <img src="/static/img/fatml-header-logo.png" srcset="/static/img/fatml-header-logo.svg 1x" alt="FAT ML 2016" height="20" width="90"> </a> </div>  <div class="collapse navbar-collapse" id="site-header-nav"> <ul class="nav navbar-nav" role="navigation"> <li ><a href="/schedule/2018">2018</a></li> <li ><a href="/schedule/2017">2017</a></li> <li ><a href="/schedule/2016">2016</a></li> <li class="active"><a href="/schedule/2015">2015</a></li> <li ><a href="/schedule/2014">2014</a></li> <li ><a href="/organization">Organization</a></li> <li ><a href="/resources/relevant-scholarship">Resources</a></li> <li><a href="https://lists.princeton.edu/cgi-bin/wa?SUBED1=fatml&A=1">Mailing list</a></li> </ul> <ul class="nav navbar-nav navbar-right">  </ul> </div> </nav> <nav class="workshop-navbar navbar navbar-inverse"> <div class="container"> <div class="navbar-header"> <button type="button" class="navbar-toggle collapsed" data-toggle="collapse" data-target="#workshop-header-nav" aria-expanded="false"> <span class="sr-only">Toggle navigation</span> <span class="icon-bar"></span> <span class="icon-bar"></span> <span class="icon-bar"></span> </button> <a class="navbar-brand" href="/schedule/2015">2015</a> </div> <div class="collapse navbar-collapse" id="workshop-header-nav"> <ul class="nav navbar-nav" role="navigation"> <li><a href="/schedule/2015/page/scope-2015">Scope</a></li> <li><a href="/schedule/2015/page/call-for-papers-2015">Call for Papers</a></li> <li><a href="/schedule/2015/page/attend-2015">Attend</a></li> <li><a href="/schedule/2015">Schedule</a></li> <li><a href="/schedule/2015/speakers">Speakers</a></li> <li><a href="/schedule/2015/page/papers-2015">Papers</a></li> <li><a href="/schedule/2015/organizers">Organizers</a></li>  </ul> </div> </div> </nav> <div class="general-content"> <div class="container"> <div class="presentation-detail"> <h1 class="presentation-detail__title">On the Relation between Accuracy and Fairness in Binary Classification</h1> <p class="presentation-detail__speaker"> <a href="/schedule/2015/speaker/indre-zliobaite"> <img src="/media/cache/4a/14/4a14d457d71f15fda71eca027337f9dd.jpg" class="presentation-detail__speaker-thumb"> </a> <a href="/schedule/2015/speaker/indre-zliobaite"><span class="presentation-detail__speaker-name">Indr臈沤liobait臈</span></a>  </p> <p class="presentation-detail__description"><p><strong>Indr臈沤liobait臈</strong></p> <p>Our study revisits the problem of accuracy/fairness tradeoff in binary classification. We argue that comparison of non-discriminatory classifiers needs to account for different rates of positive predictions, otherwise conclusions about performance may be misleading, because accuracy and discrimination of naive baselines on the same dataset vary with different rates of positive predictions. We provide methodological recommendations for sound comparison of nondiscriminatory classifiers, and present a brief theoretical and empirical analysis of tradeoffs between accuracy and non-discrimination.</p></p> </div> </div> </div> </div>  <script type="text/javascript" src="/static/javascripts/jquery.js" charset="utf-8"></script> <script type="text/javascript" src="/static/javascripts/bootstrap.js" charset="utf-8"></script> <script type="text/javascript" src="/static/javascripts/js.cookie.js" charset="utf-8"></script> <script type="text/javascript" src="/static/conference/javascripts/profile-photo.js" charset="utf-8"></script> </body> </html>

CINXE.COM

On the Relation between Accuracy and Fairness in Binary Classification :: FAT ML