Parisa, Nathan, and Peter present at AAAI2025 Workshops

Peter Lewis and Nathan Lloyd travelled to Philadelphia to present recent work at the AAAI 2025 Workshops.

Peter, on behalf of Parisa, presented their work on Adding Reflective Governance to LLMs at the 2nd International Workshop on AI Governance (AIGOV), a workshop focusing on the critical aspects of AI Governance, with particular note to ethical and responsible practice. Parisa’s work demonstrates how a reflective architecture can be used to add self-governance to an LLM-based AI system, enabling it to adaptively behave in line with social expectations.

Nathan presented co-authored work with Peter entitled Why was I sanctioned? Perspective-taking for causal explanations at the 1st International Workshop on Theory of Mind for AI (ToM4AI), an interdisciplinary workshop co-organized by Faculty Affiliate Ştefan Sarkadi that aims to bridge human cognition with AI. Nathan’s work proposed an extension to the Event Calculus, and domain extensions like the Expectation Event Calculus, to enable the modeling of other’s beliefs by recording their narratives. Which, when supported by domain dependent axioms, may provide explanations of the sort as to why an individual may have been sanctioned by another.