BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Sabre//Sabre VObject 4.5.8//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Europe/Zurich X-LIC-LOCATION:Europe/Zurich TZURL:http://tzurl.org/zoneinfo/Europe/Zurich BEGIN:DAYLIGHT TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST DTSTART:19810329T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET DTSTART:19961027T030000 RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT UID:news2009@dmi.unibas.ch DTSTAMP;TZID=Europe/Zurich:20260323T160841 DTSTART;TZID=Europe/Zurich:20260324T161500 SUMMARY:Rethinking Fairness for LLMs and Agentic AI DESCRIPTION:Abstract: Artificial intelligence systems increasingly shape ev eryday life\, influencing decisions in domains such as hiring\, education\ , healthcare\, and content generation. A growing body of evidence shows th at these systems do not perform equally across populations: they often ach ieve lower predictive accuracy for sensitive groups\, amplify historical b iases present in data\, and\, in the case of generative models\, produce s tereotypical or skewed representations. In this talk\, I will briefly cov er how fairness evolved from predictive ML to generative AI and then move to a new class of fairness problems that emerge with large language models (LLMs) and increasingly agentic AI systems. These new settings rely on co mmunication and dialogue in natural language\, and sequential decisions th at shape user experience and downstream outcomes. I will illustrate how th is perspective opens new directions for evaluating and regulating fairness in AI\, by extending existing notions of distributional and procedural fa irness with the concept of interactional fairness.Dr. Ruta Binkyte [https: //www.rutabinkyte.com/] X-ALT-DESC:

Abstract: Artificial intelligence systems increasingly shape everyday life\, influencing decisions in domains such as hiring\, educatio n\, healthcare\, and content generation. A growing body of evidence shows that these systems do not perform equally across populations: they often a chieve lower predictive accuracy for sensitive groups\, amplify historical biases present in data\, and\, in the case of generative models\, produce stereotypical or skewed representations. \;
In this talk\, I wil l briefly cover how fairness evolved from predictive ML to generative AI a nd then move to a new class of fairness problems that emerge with large la nguage models (LLMs) and increasingly agentic AI systems. These new settin gs rely on communication and dialogue in natural language\, and sequential decisions that shape user experience and downstream outcomes. I will illu strate how this perspective opens new directions for evaluating and regula ting fairness in AI\, by extending existing notions of distributional and procedural fairness with the concept of interactional fairness.

Dr. Ruta Binkyte

END:VEVENT END:VCALENDAR