Constructing a behaviour change ontology for extracting and synthesising evidence in the Human Behaviour-Change Project
AbstractBackground Developments in machine learning and artificial intelligence make it increasingly possible to identify the very large published literature on behaviour change and interventions. Natural Language Processing methods have made the selection of included studies more efficient but the process of data extraction is typically done manually and is highly labour intensive. However, for machines to contribute to the extraction of data from papers we need an ontology of behaviour change interventions i.e. an unambiguous organisation of the elements to identify these elements in published reports. Methods Three methods were used: intensive discussion amongst behaviour change experts to develop a preliminary ontology; a combination of literature review and consensus methods to populate elements in the ontology; and a review of existing relevant ontologies and annotation methods. Findings The preliminary ontology included the elements: intervention (content; delivery), context, exposure, mechanisms of action, behaviour and effect on outcome. The consensus–based Behaviour Change Technique Taxonomy populates intervention content with 93 techniques. Based on literature reviews, a hierarchical taxonomy of delivery methods and a large number of mechanisms of action have been identified. Discussion While some parts of the ontology have been developed work is needed on unpopulated elements. The Behaviour Change ontology will be linked to existing ontologies and provide a basis for machine learning. Further stages of the Human Behaviour-Change Project will develop an artificial intelligence system capable of extracting and interpreting evidence in the large literature on behaviour change interventions.
Copyright (c) 2017 M. Johnston, R. West, M. Kelly, S. Michie
This work is licensed under a Creative Commons Attribution 4.0 International License.