Developing Theoretically and Computationally Informed Learning Curricula

My project will explore how well humans and algorithms learn and generalise from different sources of information depending on how that information is presented. My principal goal is to develop and test theoretically informed computational models that provide a principled account of how learning in one domain can transfer to another. In particular, I will focus on investigating transfer between conceptual and sensorimotor abilities, building on existing research conducted by the Immersive Cognition (ICON) Lab at the University of Leeds (e.g. evidence from my supervision team demonstrating a striking relationship between sensorimotor abilities and mathematical ability, as measured on standardised national academic attainment scores; Giles et al., 2018, Psych Science).

Over the course of this project I will conduct a series of experiments to systematically investigate the extent and limits of transfer of learning between artificially generated domains of skills and conceptual understanding (referred to here as "learning domains"). Using a VR-based experimental paradigm built upon previous work by the ICON lab (Brookes et al., 2019, Behaviour Research Methods), I will gather data on the relationship between artificial learning domain similarity and the efficiency with which the domains can be learned when presented in certain sequences. Computationally generated learning domains will be used, allowing the degree of relatedness between any pair of domains to be quantified and controlled.

The models developed in this project may subsequently be applied to the design of educational curricula for human learners, and the development of more robust machine learning algorithms.