Better Policy: How can Emergent Digital Technologies (EDTs) be harnessed to ensure transparency, fairness, and inclusivity in informing, shaping, and improving policy-making processes?
Current projects:
PREDICT – Predictive algorithm for assessing external validity and identifying contextual tailored interventions
Leveraging large language models for predictive insights in food policy and behavioural interventions. Pre-print available here; published in Discover Food here (Aug 2025).
PREDICT II – Extending predictive algorithms to food and energy interventions
Cross-domain generalisation, model benchmarking and system evaluation for real-world policy use.
Usability study with policymakers
Investigating the usability and trustworthiness of LLM-assisted tools in policy contexts.
On preferring people to algorithms
Cross-national evidence on attitudes toward algorithmic versus human decision-making. Working paper available here; published in the Journal of Risk and Uncertainty here (Oct 2025).
IMPACT-AI – Interpretable model for predicting AI cognitive tendencies
Toward a multi-factor theory of bias in large language models.
Scaling behavioural interventions with LLMs
Using large language models to estimate the contextual scalability of behavioural policies.
Synthetic participants in nudge research
Exploring the reliability and boundaries of using AI-generated data for early-stage behavioural research
Collaborations:
Do LLM architecture and prompt parsimony drive alignment between synthetic and human responses?
Evaluating LLMs as proxies for public opinion on institutional trust and nudge approval.
SCAIA – Scale of confidence and aversion in AI applications
Developing and validating a measurement instrument for algorithm aversion and appreciation.