Prof. Chris Summerfield
Our research is concerned with how humans and artificial agents learn about the world, and make decisions.
Much of our work studies the mechanisms by which humans categorise stimuli into two or more discrete classes (e.g. categorising an object as "large" or "small", or as "good" or "bad"). We study this process by asking participants to perform categorisation tasks and measuring their choices and response times. This allows us to build mathematical models of how decisions might occur in neural circuits, and compate these to human data. A key focus of this work is to understand how decisions about sensory signals differ from those made about economic (or "value-based") signals, such as when deciding among consumer items. In some recent projects, we hae also attempted to understand how humans make sequential decisions, i.e. how they plan over possible future states.
Another line of work investigates how we learn about the world. Here, our focus is to understand how neural representations are formed during human learning, and how these representations allow humans to make good decisions in novel environments. To do this, we ask people to learn to categorise stimuli into different groups, and then present them with new items they have never seen before. This allows us to understand how initial learning can help deal with novelty. We compare the performance of humans to artificial systems, such as neural networks, that are faced with the same challenge.
More information can be found on our lab website