Dr Alan Pickering BA PhD - Research project

Using neurally-inspired computational models of learning under reward to understand the ways in which humans learn to categorise stimuli

Neuroscience studies have uncovered strong evidence that dopamine cells in the brain fire in response to errors in our ability to predict upcoming rewards. If we are pessimistic about future rewards then when the pleasant outcome arrives there is a short-lived increase in dopamine cell firing; if we overestimate a reward then we get a short depression in the firing rate. It is further believed that these phasic changes in dopamine cell firing are critical for "cementing" our learning about the responses which we executed just before the under- or over-predicted reward was delivered. If we make a response and get exactly the reward we expected, then no further learning occurs. We have a biologically-inspired computer model of these ideas and we have used it successfully to simulate performance on a variety of simple learning tasks.

The PhD project will develop this model further and test predictions about learning, under rewarding conditions, made by the model. For example, we propose that impulsive / sensation-seeking personality traits may correspond to a larger than normal response of the dopamine system to reward prediction errors. By using the model we can therefore work out the circumstances under which impulsive individuals will learn better or worse than nonimpulsive people. We can test this prediction by relating participants' scores on impulsivity personality scales to their performance on laboratory learning tasks.

The project will thus involve testing people on computerised learning tasks and relating their performance to simulated data generated by the model. Part of the PhD will therefore involve programming variations to the model; training in this will be provided and no previous programming experience is required (although it would be an advantage). A current PhD student, who has worked on the model very successfully, had no programming experience before she started.