Reasoning about Unforeseen Possibilities During Policy Learning
This addresses the challenge of policy learning in autonomous agents for scenarios where domain knowledge is incomplete or evolves, offering a solution that is incremental by building on existing methods to handle unforeseen possibilities.
The paper tackles the problem of learning optimal policies when agents initially lack knowledge of all possible states and actions, using a model that combines probabilistic and symbolic reasoning to discover and exploit unforeseen possibilities through interaction and expert communication. The agent simulations demonstrate convergence to optimal policies even when starting unaware of critical factors.
Methods for learning optimal policies in autonomous agents often assume that the way the domain is conceptualised---its possible states and actions and their causal structure---is known in advance and does not change during learning. This is an unrealistic assumption in many scenarios, because new evidence can reveal important information about what is possible, possibilities that the agent was not aware existed prior to learning. We present a model of an agent which both discovers and learns to exploit unforeseen possibilities using two sources of evidence: direct interaction with the world and communication with a domain expert. We use a combination of probabilistic and symbolic reasoning to estimate all components of the decision problem, including its set of random variables and their causal dependencies. Agent simulations show that the agent converges on optimal polices even when it starts out unaware of factors that are critical to behaving optimally.