Explainable Agency within Team Cognition for Regulated Domains

AAAI 2026 Spring Symposium Series
Multi-Agent System Safety and Teamwork (MASST): Building Better AI

Azam Khan and Simon Breslav

April 2026
12 Pages (In Press)

In regulated domains, humans remain legally accountable for AI system outcomes, making AI design a fundamentally coactive problem. Such systems must support human–agent interdependence, explainability, and interpretability for non-AI experts, while enabling governance and regulatory traceability. In the construction sector, despite growing AI adoption, regulatory requirements remain largely unintegrated, contributing to rising compliance costs and declining productivity. This work examines policy-based multi-agent systems as a foundation for effective human–agent teaming in regulated environments such as construction.