Understanding AI Trustworthiness: A Scoping Review of AIES & FAccT Articles

Technical focus dominates but misses crucial context: Current AI trustworthiness research prioritizes technical attributes like reliability and fairness, but largely overlooks the sociotechnical dimensions—the interaction between technology, people, and social systems—that determine whether AI actually works in real-world settings.

Trustworthiness means different things to different people: The paper identifies trustworthiness as a "contested concept shaped by those with the power to define it," suggesting that researchers, companies, policymakers, and affected communities may have fundamentally different understandings of what makes AI trustworthy, yet these differences aren't adequately addressed in current research.

Measurement gaps exist between theory and practice: While significant progress has been made in defining technical attributes like transparency and accountability, the paper reveals critical gaps in how trustworthiness is actually measured and validated, indicating a disconnect between what researchers say matters and how they verify it works.

Interdisciplinary collaboration is essential but underutilized: The conclusion calls for an interdisciplinary approach, implying that current research—concentrated in technical and computer science communities—needs input from social scientists, ethicists, domain experts, and affected communities to develop more holistic trustworthiness frameworks.

Social and ethical considerations are being deprioritized: The research reveals that technical precision is often pursued "at the expense of social and ethical considerations," suggesting that responsible AI development requires rebalancing priorities to give equal weight to how systems affect people and society.