During Michaelmas Term, OxTalks will be moving to a new platform (full details are available on the Staff Gateway).
For now, continue using the current page and event submission process (freeze period dates to be advised).
If you have any questions, please contact halo@digital.ox.ac.uk
Multiple complimentary approaches are available for modelling the adaptive behaviour of individual agents in complex systems, and in this work reinforcement learning is the focus. A core problem here is that unambiguous identification of rewards driving the behaviour of entities operating in complex (open-ended) real-world environments is at least difficult, if not impossible. In part this is because the true goals of agents are not observable; also, reward-driven behaviours emerge endogenously over longer timescales and are dynamically updated as environments change. Defining a reliable reward function to use in models therefore remains a challenge. Reproducing the emergence of rewards is a potential solution, and would be have application in many domains. Simulation experiments will be described which assess a candidate algorithm for the dynamic updating of rewards, RULE: Reward Updating through Learning and Expectation. The approach is tested in a simplified ecosystem-like setting where manipulated conditions challenge the survival of an entity population, calling for significant behavioural change.