Statistical Methods in Adaptive Learning

This aim of this project is to enable development of small-scale adaptive learning systems by creating new statistical methodology for analysis and design of such systems. Recent years have seen an upsurge in the use of Artificial Intelligence (AI) in Education, including large-scale AI adaptive learning systems. However, such systems are not yet routinely used at smaller scales, for instance in individual higher education courses.

To enable the development of adaptive learning systems across a range of scales, we need to answer three research questions:

1. How can well-calibrated predictions of a learner's future performance be made to avoid bias in subsequent decisions and summaries?

2. How can interpretable summaries of each learner's abilities be provided to enable each person to understand where they need to improve? 3. How can good choices be made about which content to offer a learner next to help people learn most effectively?

This project will develop new methods to improve the design of adaptive learning systems. Developing flexible models for learning systems will enable us to make well-calibrated predictions of a learner's future performance across a range of scales. Creating summaries of the fit from a complex model will provide us with interpretable summaries of each learner's abilities.

Developing methods for design of adaptive systems will enable us to make good choices about which content a learner should be offered next.

Together, our methodological developments will lay the ground for the creation of general-purpose software to provide adaptive learning systems at small-medium scales.