"Learning, memory, development, and evolution"

点击翻译按钮获取中文摘要

This proposal focuses on building and testing computational models of learning, memory, development, and evolution to better understand how these processes work in humans. For learning and development, we use artificial neural networks inspired by principles of how brains work. Such models contribute to understanding of a variety of issues in psychological development: why children go through stages in their thinking, why children make the mistakes that they do, and the causes and remedies for disorders in children's cognitive development. Good models implement a concrete, easy-to-study version of phenomena, in which unimportant details are stripped away so that the basic principles can be more easily understood. In this proposal, we concentrate more specifically on simulating the development of children's ideas about other people's minds, particularly the beliefs and desires that underlie behaviour. Children go through an early stage in which they assume that other people share the same beliefs as they themselves possess. Then later children come to appreciate that other people may sometimes harbour false beliefs. We also propose to model how memory retrieval puts those memories in an unstable state allowing them to be reconsolidated. In the presence of new, relevant information, retrieved memories can be changed. This has implications for faulty eyewitness testimony as well as treating post-traumatic stress. We will also simulate the ability of infants to take an active role in their own learning, e.g., by quitting tasks that are too difficult for them to learn. This suggests that they might have the ability to monitor their own progress in learning and the sense to move on to something else when there is lack of progress. Finally, we also simulate the biological evolution of different kinds of learning ability in populations of agents that interact, solve problems, reproduce according to their fitness, and eventually die. We are interested in the interplay between individual learning, imitation, and the passing of theories about the world from one agent to another. We aim to discover how these different learning techniques are created by natural selection and how they, in turn, affect the course of biological evolution.