ELIZA, but cleverer: Designing persuasive artificial agents

This extended abstract introduces research into persuasive dialogue modelling which is to be carried out as part of PIPS,1 one of the five Integrated Projects funded from the first call of the EU 6th Framework Programme in the field of e-Health. One of PIPS’ aims is to design and build a personal computational advisory system to support European Citizens in taking informed health-related decisions about lifestyle. PIPS’ advice will be based on an extensive knowledge engineering effort, aimed at integrating many heterogeneous data sources. The PIPS artificial advisory agent can be simplistically viewed as a kind of online health advisor whom a user might typically chat with when he needs practical, health-related advice concerning some particular change of lifestyle. It is envisaged that the artificial advisory agent will be used largely for the purposes of preventative medicine. The agent will be able to offer individually personalised counselling on nutrition, exercise, and other health-related practices, such as stopping smoking, and practising hygiene in the kitchen. Research in health promotion has long proved that persuasive skills are needed for such advisor–advisee scenarios, as they involve changing well rooted attitudes and behaviours (e.g., (Prochaska et al., 1992, 1994, 1997)). Unlike ELIZA, then, the PIPS advisory agent will not only be extremely knowledgeable (thanks to its comprehensive knowledge base, medical ontologies, and knowledge extraction techniques), but it will be skilled in the art of persuasion. This paper concerns the ‘art of persuasion’ part of this undeniably large problem.