Similarity attracts, or does it? Studying personality-based convergence and sense of engagement with a digital health assistant

• Digital services aim to keep the users engaged by personalizing their offer. • Personalization is mainly via similarity with users’ characteristics/past choices. • Digital Health Assistants are chatbots implemented in health applications. • We created three personality-based linguistic variants of a Digital Health Assistant. • We tested the effect of user-DHA personality convergence on engagement. According to the similarity-attraction hypothesis in social psychology, users prefer people like them (or “convergent” with them) in attitudes and behaviors. In this study, we test the similarity-attraction hypothesis on conversational agents (CAs) by measuring if they are more engaging when their perceived personality converges with the users’. We simulated on the Qualtrics platform a textual Digital Health Assistant (DHA) whose style could vary thanks to incorporating specific linguistic cues. The study participants interacted with the DHA variants, assessed their personality, and then rated their engagement. They also assessed their own personality. After correlating engagement and personality scores for each CA variant, we found no effect of convergence. These results contribute to a better understanding of CA personalization, questioning similarity as a widespread strategy whose high privacy toll might not be compensated by improving the user experience.