Blog

A collaborative effort between behavioural scientists, computer scientists and systems architects is busy creating an online ‘Knowledge System’ for the Human Behaviour-Change Project (HBCP). Emma Norris and Susan Michie introduce us to the project and share a sneak preview of the research that will be published in the new HBCP collection on Wellcome Open Research, on behalf of their team, which includes James Thomas, Marie Johnston, Pol Mac Aonghusa, John Shawe-Taylor, Michael P. Kelly & Robert West.

Behaviour change is key to addressing many of the challenges facing the human population (e.g. reducing carbon emissions, preventing overuse of antibiotics, stopping tobacco use and reducing transmission of infectious diseases). A huge amount of information is being gathered on how best to achieve this in different situations, but we have very limited capacity to collate it, synthesise it and use it to make recommendations.

The Human Behaviour-Change Project (HBCP) is a Wellcome-funded project aiming to support decisions about behaviour change interventions using cutting-edge Artificial Intelligence (AI). The project aims to largely automate the process of collating, synthesise and interpreting evidence from the vast and growing literature on behaviour change intervention evaluations.

Working together

The project is a collaboration between behavioural and computer scientists and system architects that aims to create an AI-based Knowledge System that will scan the world’s published reports of behavioural intervention evaluations. This system will extract and analyse relevant information on interventions and their effectiveness organised using a ‘Behaviour Change Intervention Ontology’ (BCIO), developed as part of the project. You can read more on what ontologies are and how they can be used to structure knowledge here.

The Knowledge System will answer user queries and make recommendations as to what interventions are likely to work in a given scenario. It will also outline the level of confidence in its answers and explain the process behind its answers. The first behaviour we are investigating is smoking cessation, drawing on published reports of randomised controlled trials.

 The key activities involved in the project are to develop:

1. An ontology of behaviour change interventions and evaluation reports: the Behaviour Change Intervention Ontology (BCIO).
2. An automated system to extract information from behaviour change intervention evaluation reports using Natural Language Processing.
3. A representation of that information structured according to the BCIO.
4. Reasoning and Machine Learning algorithms to synthesise this information and make inferences in response to user queries.
5. An interface for computers and human users to interact with the system.

For all to benefit

In behaviour change, open access to knowledge is essential to enable the development of effective interventions by researchers, policy-makers and practitioners. The establishment of effective health interventions benefits all. We want to publish our key findings in one easily accessible place, providing free access to all the outputs from the project.

We are also making our methods, working papers and resources available via Open Science Framework. We would like to receive feedback on our papers via Wellcome Open Research. The HBCP is a huge undertaking and it will require involvement of much of the research community working together to advance it at the speed required.

What to expect

Our initial launch of papers in the collection contains five papers:

• Editorial – introducing the project.
• Methodology paper – explaining the methods we used for ontology development.
• Upper-level Ontology paper – specifying the overarching structure of the Behaviour Change Intervention Ontology.
• Mode of Delivery Ontology paper – describing a part of the BCIO that characterises ways that behaviour change interventions are delivered (e.g. by face-to-face contact, websites, video)
• Setting Ontology paper – describing a part of the BCIO that characterises the locations in which interventions are delivered (e.g. what country they are in, whether they are in hospitals or primary care)

We will continue to publish papers in the collection as other parts of the project are completed, with several currently in the pipeline.

You can find more information on the Human Behaviour-Change Project on our website and Twitter.