Individual research project · 2026
Recursive theory of mind in an abstract epistemic environment
Bounded reasoning about what agents know, observe and believe about one another.
Conceptual illustration; not an experimental result.
Research question
How can an agent represent another agent’s beliefs when their observations differ? The project explores recursive theory of mind in a controlled environment with asymmetric access to events.
Methodology
A simplified framework inspired by dynamic epistemic logic represents possible worlds and accessibility relations. Bounded recursive queries and deterministic scenarios test first-, second- and third-order false beliefs, including agents leaving and re-entering the environment.
Findings
The model supports the core false-belief scenarios. Observability-consistent revision remains only partially supported: higher-order belief repair after re-entry can propagate more broadly than intended.
Limitations
The evaluation uses abstract, deterministic scenarios with bounded reasoning depth. It does not establish performance in noisy, embodied or large-scale social environments.
Related research
Multi-agent social simulation