Embedded Files
AI CHAOS!
As AI systems are increasingly deployed in high-stakes domains, ranging from healthcare and autonomous driving to platform governance and automated legal compliance, their failures directly threaten human safety and fundamental rights.
Central to HCOMP’s mission around Human-AI Complementarity and Alignment, human oversight serves as a vital socio-technical safeguard. Drawing on recent interdisciplinary roadmaps, effective oversight is conceptualized not as a solitary interface task, but as a deliberate, risk-mitigating activity structured around the two operational pillars of continuous monitoring and timely intervention. This responsibility is distributed across diverse organizational stakeholders, including frontline operators, domain experts, and compliance officers.
However, emerging regulatory mandates like the European AI Act and the Digital Services Act have outpaced our conceptual clarity and empirical methodologies. Poorly designed oversight architectures risk creating a dangerous illusion of control, obscuring accountability without providing humans with genuine causal agency or systemic support.
Together, we hope to advance human empowerment in the debate on oversight of AI systems by identifying research gaps, open challenges, and general guidelines.
Yutong Liu & Digit / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/
Embracing the HCOMP 2026 theme of "Connections,'' this workshop establishes vital links across AI, human computation, HCI, psychology, law, and policy to bridge this critical gap.
The outcomes of our activities will be published in a blog post on this site and organized into an ACM Interactions submission on oversight problems.
Page updated
Google Sites
Report abuse