Workshop on Trustworthy Algorithmic Decision-Making

点击翻译按钮获取中文摘要

Algorithms are increasingly being used in systems that make decisions that affect people, including filtering news and updates we see, setting prices for items, scoring resumes and credit applications, recommending routes and places, and controlling smart homes and autonomous cars. As algorithms make these decisions, it is natural to ask whether, when, and why we should trust them, and to consider ways to improve their trustworthiness. To this end, researchers, regulators, and industry have increasingly called for methods to audit and account for algorithmic decisions to ensure their transparency and fairness. This project's goal is to integrate these perspectives, bringing together people who design and study these algorithms for a two-day workshop to lay out the problem space and possible ways forward for algorithmic trustworthiness. The workshop organizers will emphasize diversity of perspectives, recruiting participants from a wide variety of disciplines, jobs, and backgrounds with a particular focus on including people from underrepresented groups in computing and PhD students. The focus will be on developing interdisciplinary and convergent perspectives on algorithmic decision-making by bringing together researchers, designers, and policy-makers working on different aspects of the problem and different domains, with the goal of integrating their ideas to get a bigger picture of the larger question of trustworthiness of these algorithms in society. The main output will be a report describing the problem space of trustworthy algorithmic decision-making, outlining the major issues that need research to move these algorithms and the decisions they make toward being trustworthy, and identifying promising opportunities and approaches for studying these issues. It will also identify some of the major challenges and constraints that will make progress on these issues difficult, and will further the discussion of moving past these challenges. In addition to the report, the participants will produce a set of whitepapers that will illuminate important aspects of this problem, along with responses and additional reflections by attendees.