Human behaviour change



The need

Human behaviour needs radical change to protect our individual and collective well-being. To achieve this, we need to develop more effective behaviour change interventions, tailored to the behaviour, population and setting. The Human Behaviour-Change Project will build an Artificial Intelligence system to continually scan the world literature on behaviour change, extract key information, and use this to build and update a model of human behaviour to answer the big question:

‘What behaviour change interventions work, how well, for whom, in what setting, for what behaviours and why?’

The vision

Many of the global threats we face, for example, to health and environmental sustainability, can only be solved by people, organisations and governments changing their behaviour. To achieve this we need to rapidly improve our understanding of behaviour and how to change it.

The scientific literature on behaviour change is vast and accumulating at an accelerating rate. However, this literature is fragmented, and is inconsistently and incompletely reported. The result is that most of it is wasted. Current attempts to synthesise this evidence can take years, miss much that is relevant and fail to detect patterns which can generate knowledge allowing generalisation beyond the contexts within which the evidence was generated.

Advances in computing such as the IBM Watson platform demonstrate how it is possible to apply Natural Language Processing and Machine Learning technologies to reveal insights from large volumes of unstructured text. For example, Watson for Oncology recommends interventions for individual cancer patients based on clinical evidence extracted and organised from research publications, medical reports and clinical trials.

It is now possible to create Artificial Intelligence systems to identify relevant information in the world literature, extract it into an organised knowledge base (‘ontology’) created by behavioural scientists, and generate new insights about behaviour change.

This knowledge base can then be interrogated on demand to answer questions about behaviour change, big and small. It will provide answers drawing on knowledge integrated from a broader literature than humans can review, and will also be able to point to relevant references as well as estimate the confidence with which statements can be made.

The project

The Human Behaviour-Change Project, funded by the Wellcome Trust and led by Professor Susan Michie, is a collaboration between behavioural scientists (Michie, Johnston, Kelly, West), computer scientists (Shawe-Taylor and Mac Aonghusa) and system architects (Thomas) based at University College London and Universities of Cambridge and Aberdeen, UK, and the IBM Research Laboratory in Ireland. A multi-disciplinary team of 12 will work over four years to revolutionise current practices of evidence synthesis and our ability to generate new knowledge. The work will depend on a close interplay between the scientific disciplines involved.

The behavioural scientists will develop an “ontology” of behaviour change interventions that will organise the fragmented knowledge in the scientific literature into a form that enables the efficient accumulation of knowledge, as shown below:

Top level of the Behaviour Change Intervention Ontology


The computer scientists will build an Artificial Intelligence system, trained by behavioural scientists, to apply Natural Language Processing to extract relevant information from scientific reports and to organise that information into the Ontology using reasoning and machine learning. The Artificial Intelligence system will furthermore infer new knowledge while continually learning from new information fed into it.

The system architects will build and evaluate a sophisticated online user interface to interact with the Artificial Intelligence system to enable users to readily access the breadth and depth of up-to-date evidence, and get answers to their questions, with explanations of these answers that people can understand and trust.