Evaluating the Effect of Adapting Virtual Humans Based on Individual Differences in Users

This paper investigates the effects of adapting a virtual human’s persuasion strategy based on users’ personalities and prior beliefs regarding recommended behavior in the context of promoting mental health coping skills among college students. The paper uses the Theory of Planned Behavior (TPB) as the theoretical model to study how a virtual human’s persuasion strategies impact behavior change. The paper also employs Cialdini’s six persuasion strategies - Reciprocity, Scarcity, Authority, Commitment, Likability, and Consensus - to manipulate the virtual human’s dialog. The paper develops a user model that predicts the effectiveness of different persuasion strategies based on user data from a previous study. The paper then evaluates the user model in an empirical study with 292 undergraduate students, comparing three experimental conditions - a matched condition where the virtual human used a more effective persuasion strategy, a mismatched condition where the virtual human used a less effective persuasion strategy, and a control condition where the virtual human did not use any persuasion strategy. The paper finds that adapting the virtual human’s persuasion strategy can positively influence users who have low self-efficacy to perform the recommended behavior, but can negatively influence users who already have high self-efficacy. The paper also finds that persuasion strategies may not be sufficient to induce behavior change, and suggests accounting for users’ perceived barriers and benefits of the recommended behavior. The paper contributes to the Human-Computer Interaction research by providing evidence for the importance of individual differences in designing virtual human health interventions.