The Human Behaviour-Change Project: Building the science of behaviour change for complex intervention development

Every day, roughly 2,000 new studies on behaviour change are published, but the scientific community lacks an efficient way to learn from them. This matters because behaviour change is central to tackling major health problems—from smoking cessation to vaccine uptake—yet current methods for synthesising evidence are slow and fragmented. Researchers typically ask whether a single intervention “works” without building the deeper understanding needed to answer: *What works, how well, for whom, in what setting, for what behaviours, and why?* This project will build a system that automates that process. A team of behavioural, information, and computer scientists will develop software that machine-reads intervention evaluations, extracts key data, and feeds it into a computational model. That model will link intervention characteristics with context, mechanisms of action, and outcomes. Researchers, practitioners, and policymakers will then interrogate the model through a user interface, and the system will learn from their feedback. If successful, this infrastructure could transform how evidence is used in public health, social policy, and clinical practice—replacing slow manual reviews with a continuously updated, queryable knowledge base. The impact is not a new drug or device, but a smarter way to design and evaluate interventions that shape everyday behaviours.