Personalized persuasion in ambient intelligence

This thesis examines the use of personalized persuasion in ambient intelligence. Persuasive technologies—systems and services intentionally designed to influence user behavior—are emergent. Applications range from iPhone apps which help users stop smoking to distributed networks of smart sensors that persuade users to reduce their energy consumption. Many of these systems apply social science knowledge about influence strategies to increase the effectiveness of their persuasion attempts. This thesis first examines user responses to influence strategies. The work shows that while most influence strategies are effective on average, large individual differences exist. The responses to some strategies—e.g. authority arguments—are negative for a large proportion of users despite replicating the average positive effect of the use of this strategy. This heterogeneity in responses to social influence strategies proves stable over both time and contexts. After showing that a proper selection of a single influence strategy leads to more persuasion than combinations of strategies, the thesis develops the idea of creating and utilizing persuasion profiles: collections of the estimates, and their associated certainty, of the effects of influence strategies on individuals. The thesis describes how these profiles can be build both via meta-judgmental measures as well as operative measures. The Susceptibility to Persuasive Strategies Scales (STPS) is presented for the purpose of creating persuasion profiles based on meta-judgmental measures. The thesis next examines the applied value of persuasion profiles. Via several designs the thesis shows that persuasive systems in which the influence strategy that is used is adapted to individual users outperform non-personalized systems. These ideas are further advanced by proposing and evaluating a method for building personalized persuasive technologies. Persuasion Profiles as presented in this thesis will be a core component of persuasive technologies to come. The ambient intelligence scenario makes it possible to build dynamic profiles based on unobtrusive measurements. This in turn will help researchers and designers to measure, predict, influence, and ultimately understand human responses to persuasion.