{"id":17,"date":"2019-01-31T21:36:44","date_gmt":"2019-01-31T21:36:44","guid":{"rendered":"http:\/\/groups.cs.umass.edu\/equate\/?p=17"},"modified":"2019-04-03T17:20:58","modified_gmt":"2019-04-03T17:20:58","slug":"data-diversity","status":"publish","type":"post","link":"https:\/\/groups.cs.umass.edu\/equate\/research\/data-diversity","title":{"rendered":"Data Diversity"},"content":{"rendered":"

The big data revolution and advancements in machine learning technologies have\u00a0<\/span>revolutionized decision making, advertising, medicine, and even election\u00a0<\/span>campaigns. Yet, data is an imperfect medium, often tainted by skews and\u00a0<\/span>biases. Learning systems and analysis software learn and amplify these biases.\u00a0<\/span>As a result, discrimination shows up in many data-driven applications, such as\u00a0<\/span>advertisements, hotel bookings, image search, and vendor services. Since data\u00a0<\/span>skew is often a cause of algorithmic bias, the ability to retrieve balanced,\u00a0<\/span>diverse datasets can mitigate the underlying problem. Diversification also has\u00a0<\/span>usability implications, as it allows us to produce representative samples of a\u00a0<\/span>dataset that are small enough for human consumption. Our research focuses on\u00a0<\/span>developing methods for producing appropriately diverse subsets of given\u00a0<\/span>datasets efficiently and scalably, aiming to alleviate biases in the\u00a0<\/span>underlying data and to facilitate user-facing data exploration systems.<\/span><\/p>\n

Publications<\/h4>\n