Research
Six lines of work
Each theme below states its computing contribution first. Education is the domain in which we evaluate the systems we build, not a substitute for building them.
How we work
Three commitments shape every project. Separate the variables: adaptive systems and generative AI report contradictory effects largely because adaptation is rarely isolated from content quality and time on task, so our designs hold those apart and measure them independently. Benchmark the artefact, not the impression: fluent output is not correct output, so we build open benchmarks of factual accuracy rather than trusting plausibility. Publish the whole thing: protocols preregistered before data collection, datasets deposited with DOIs and bilingual metadata, instruments and benchmarks released under open licences.
Open by default
Datasets and software deposited on Zenodo with DOIs and open licences · ORCID for people and ROR for organisations · GPL 3.0+ for released tools · Diamond Open Access publication with no author fees. Preregistration on OSF is committed in the projects funded from 2026 and will be linked when the registrations exist.