Using relational agents to increase engagement in computer-based interventions: preliminary outcomes

Abstract

Introduction. Multimedia computer-tailored (Expert System; ES) interventions represent a low cost, easily disseminated method of changing behaviors in a general population. Effectiveness has been established. However, these interventions have produced only small effect sizes, partially due to low participant engagement. Relational Agents (RA), computer-based virtual characters who can establish a personal relationship, have the potential to increase engagement. This paper reports the Month 12 outcomes for a RCT involving Relational Agents. Method. The design was 3 groups (Control, ES, RA) by 3 Occasions (Months 0, 12, 24). A representative USA sample (N = 1639) of individuals at risk for both exercise and UV protection was recruited. Results. Month 12 data demonstrated that the RA Intervention increased participant engagement. The average number of sessions viewed/week was 0.048 (ES) compared to 0.142 (RA). The RA also increased satisfaction and was viewed as more helpful. Proportion at criteria for exercise was: Control (13.6%), ES (15.7%) and RA (16.6%). Conclusions. The Relational Agent Intervention has the potential to outperform the previous best practice for computer-based interventions.