BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Haiyi Zhu (CMU) DTSTART:20210901T160000Z DTEND:20210901T170000Z DTSTAMP:20260417T093433Z UID:Metagov/74 DESCRIPTION:Title: Community-Centered AI: Identifying and Navigating Trade-offs Across Multi ple Community Goals in AI Design\nby Haiyi Zhu (CMU) as part of Metago vernance Seminar\n\n\nAbstract\nAI technologies are increasingly impacting a wide variety of communities\, both online and offline\, in complex and important ways. However\, AI tools and systems that appear to provide effi cient solutions to address communities’ problems can fail in practice. T o address the challenge\, researchers are increasingly arguing for the imp ortance of engaging relevant community stakeholders in the design of the A I systems that will be deployed in these communities. However\, the compli cated nature of a community's goals and needs\, and the complexity of AI ’s development procedure\, outputs\, and potential impacts\, often preve nt community stakeholders from meaningful participation in the decision-ma king of AI system design. In this talk\, we propose a community-centered A I design approach and argue for the importance of collective decision-maki ng via deliberation\, mediated through technical innovations. Specifically \, we develop and use a suite of innovative tools\, techniques\, and meth ods to capture and explain the trade-offs across multiple community goals in AI design\, to engage community stakeholders to explore\, discuss\, and negotiate the trade-offs\, and make collective and informed decisions. In the talk\, I will discuss our ongoing work\, focusing on conducting commu nity-centered AI design in two high-impact online community contexts\, Wik ipedia and 7 Cups.\n LOCATION:https://researchseminars.org/talk/Metagov/74/ END:VEVENT END:VCALENDAR