Papers
- FAccT
The Doctor Will (Still) See You Now: On the Structural Limits of Agentic AI in Healthcare
A qualitative study based on interviews with 20 stakeholders examining how agentic AI is defined, evaluated, and constrained in healthcare, identifying three mutually reinforcing tensions: conceptual fragmentation, an autonomy contradiction, and an evaluation blind spot.
Agentic AIHealthcare AIResponsible AI - FAccT
Expert Evaluation and the Limits of Human Feedback in Mental Health AI Safety Testing
A mixed-methods study examining inter-rater reliability among three psychiatrists evaluating 360 LLM-generated mental health responses, revealing systematic expert disagreement driven by incompatible clinical frameworks rather than measurement error.
AI SafetyRLHFMental Health AI - arXiv
How Language Models Fail: Token-Level Signatures of Committed and Persistent Reasoning Failures
Characterizing language model reasoning failures through token-level uncertainty signals, and showing they arise through two empirically distinguishable processes: committed failure, where a model locks onto a wrong path early, and persistent uncertainty, where uncertainty accumulates throughout the trace.
LLMReasoningUncertainty QuantificationFailure Detection - ECAI
LeRAAT: LLM-Enabled Real-Time Aviation Advisory Tool
A real-time advisory system leveraging large language models to assist aviation professionals with decision-making during complex operational scenarios.
LLMAviationNLPReal-Time Systems - AIES
An Adaptive Responsible AI Governance Framework for Decentralized Organizations
A case study of a responsible AI assessment run across more than 50 semi-autonomous business units of a multinational enterprise, identifying four patterns that shape implementation and proposing the ARGO framework to balance central coordination with local autonomy.
Responsible AIAI GovernanceOrganizational Practice - arXiv
The Measurement Imbalance in Agentic AI Evaluation Undermines Industry Productivity Claims
A systematic review of 84 papers (2023–2025) exposing an evaluation imbalance in agentic AI, where technical metrics dominate (83%) while human-centered, safety, and economic dimensions remain peripheral, with a proposed four-axis evaluation framework.
Agentic AIAI EvaluationBenchmarking - FAccT
Responsible AI in the Global Context
A global survey-based study exploring responsible AI practices across 1000 organizations in 20 industries and 19 regions, defining a conceptual RAI maturity model.
Responsible AIAI Governance - ICHMS
A Trust-Aware Reinforcement Learning Approach to Enhance Human-Agent Teaming: An Overcooked-AI Study
Embedding a real-time trust prediction model into a reinforcement learning agent's reward function, so that shaped rewards and a bonus mechanism respond to predicted trust dynamics. The trust-aware agent outperformed non-trust-aware baselines when collaborating with a human proxy.
Reinforcement LearningHuman-Agent TeamingTrustMulti-Agent RL - IUI
More than Marketing? On the Information Value of AI Benchmarks for Practitioners
A qualitative interview study with 19 practitioners in academia, product, and policy examining how AI benchmarks are used to inform decision-making, finding that benchmarks serve as relative performance indicators but often lack the real-world relevance needed for substantive deployment decisions.
AI BenchmarksModel Evaluation - RO-MAN
Action Over Words: Predicting Human Trust in AI Partners Through Gameplay Behaviors
Inferring trust in human-AI teams from gameplay behavior rather than questionnaires. Participants collaborated with AI agents of varying competence in Overcooked-AI, and behavioral models matched the predictive accuracy of self-report without asking players a single question.
TrustHuman-AI InteractionBehavioral Measurement - TechRxiv
Human-Agent Teaming: A System-Theoretic Overview
A systems-theoretic and interdisciplinary account of human-agent teaming, written to bridge the gap between the AI and human-machine interaction communities and to establish a common language between them.
Human-Agent TeamingSystems TheoryInterdisciplinary Review - AAAI Symposium
RL-HAT: A New Framework for Understanding Human-Agent Teaming
A framework for human-agent teaming grounded in reinforcement learning, offering a shared language across the disciplines that study it. It extends standard RL constructs with belief states, prior knowledge, social considerations, situational awareness and mental models.
Human-Agent TeamingReinforcement LearningConceptual Frameworks