Responsibility in Multi-Agent Sequential Decision-Making: Comparing Human Judgments to Formal Models of Causal Attribution

2026-08-04

AI Summary

In multi-agent sequential decision-making, aligning formal responsibility attribution with human intuition remains an open challenge. This study presents the first systematic comparison of multiple actual-causality-based responsibility attribution methods against human judgments. Through large-scale behavioral experiments on a modified version of the Goofspiel card game, the authors collected human assessments of responsibility for failure outcomes and employed statistical analyses to evaluate the alignment of each method with these judgments. The findings reveal that no single existing method fully captures human intuitions about responsibility; however, the results highlight agent bias and information accessibility as critical factors shaping responsibility attributions. These insights provide an empirical foundation for developing multi-agent responsibility mechanisms that better reflect human cognitive patterns.

Abstract

With the growing adoption of artificial intelligence in high-stakes decision-making, identifying the causes of outcomes--particularly failures--and determining who is responsible has become a critical concern. In this work, we examine how well formal definitions of \textit{responsibility attribution}, grounded in the framework of \textit{actual causality}, align with human judgments of responsibility. To this end, we conduct a large-scale survey to elicit human judgments of responsibility in multi-agent sequential decision-making scenarios, using a modified version of the card game Goofspiel. We evaluate multiple responsibility attribution methods, assess their alignment with human judgments about responsibility, and identify factors that significantly shape responsibility judgments. While no single responsibility attribution method consistently aligns with human responses, our findings highlight key factors that influence human responsibility judgments, including agent-specific biases and amount of information available to agents during decision-making.

Problem

• responsibility attribution
• actual causality
• multi-agent sequential decision-making
• human judgment
• causal attribution

Innovation

• responsibility attribution
• actual causality
• multi-agent sequential decision-making
• human judgment alignment
• Goofspiel