Learning and using models of causality for human-aligned reinforcement learning

Project Title:

Learning and using models of causality for human-aligned reinforcement learning

Supervisor(s):

Assoc Prof Peter Vamplew, Dr Cameron Foale; Co-supervisor Assoc Prof Richard Dazeley (Deakin)

Contact person and email address:

Assoc Prof Peter Vamplew (p.vamplew@federation.edu.au)

A brief description of the project:

The ability to understand the causes of events within its environment is of great value to an artificially intelligent agent, particularly in the context of addressing concerns of human-alignment such as the ability to explain its actions, or in being able to identify whether adverse outcomes are caused by the agent itself or by other factors within the environment, such as humans or other agents.

This project will investigate methods for learning a model of causality based on observations of environmental state transitions, and examine how such a model might be used to develop novel approaches to agent explainability and AI safety.