Accountability in Algorithmic Systems: From Principles to Practice

Abstract

Growing concerns over the societal implications of artificial intelligence has motivated an interdisciplinary push towards mechanisms and tools that hold algorithmic systems accountable. Although there have been considerable strides around defining what it means to hold AI systems accountable, operationalizing those principles have created a barrage of challenges. Researchers, practitioners, and regulators have all raised concerns about the completeness of accountability methods and observed spikes in anxiousness about the potential risk of these tools being manipulated as rubber stamps of approval while harms continue to slip through the cracks. This interactive panel gathers researchers and practitioners with expertise in HCI, Responsible AI, Machine Learning, and Public Policy to critically discuss issues regarding accountability in algorithmic systems to reflect on potential opportunities for re-imagining scalable directions for accountability within these systems.