From Interaction to Attitude: Exploring the Impact of Human-AI Cooperation on Mental Illness Stigma

TL;DR

The paper investigates how human-AI cooperation design in chatbots influences attitudes toward mental illness. Using a two-week longitudinal mixed-methods study with three chatbot variants—information-only, cooperative with mental health content, and cooperative with unrelated content—the authors show that two-way cooperation generally improves perceived chatbot intelligence, likability, and elicited empathy, contributing to stigma reduction. Content alignment matters: when cooperative tasks relate to mental illness and promote equal status and shared goals, stigma decreases more consistently, whereas mismatches between the chatbot’s role and content can increase coercive attitudes. The findings offer design principles for scalable anti-stigma interventions and illuminate the nuances of trust and relationship-building in human-AI collaboration for social impact.

Abstract

AI conversational agents have demonstrated efficacy in social contact interventions for stigma reduction at a low cost. However, the underlying mechanisms of how interaction designs contribute to these effects remain unclear. This study investigates how participating in three human-chatbot interactions affects attitudes toward mental illness. We developed three chatbots capable of engaging in either one-way information dissemination from chatbot to a human or two-way cooperation where the chatbot and a human exchange thoughts and work together on a cooperation task. We then conducted a two-week mixed-methods study to investigate variations over time and across different group memberships. The results indicate that human-AI cooperation can effectively reduce stigma toward individuals with mental illness by fostering relationships between humans and AI through social contact. Additionally, compared to a one-way chatbot, interacting with a cooperative chatbot led participants to perceive it as more competent and likable, promoting greater empathy during the conversation. However, despite the success in reducing stigma, inconsistencies between the chatbot's role and the mental health context raised concerns. We discuss the implications of our findings for human-chatbot interaction designs aimed at changing human attitudes.

Figures (11)

• Figure 1: Recent studies leverage digital technologies for social contact interventions. In one scenario, technology serves as a medium facilitating human-to-human contact. In another case, technology functions as a virtual agent representing the stigmatized group, engaging directly with participants.
• Figure 2: Experimental procedure. Throughout the two-week study period, all three groups completed daily tasks using a Telegram chatbot, Holly. On odd-numbered days, participants engaged in small talk, were presented with a vignette about Holly's experiences related to mental illness, and then engaged in a learning task. For Group 1, this task consisted solely of reading a message aimed at boosting their mental illness knowledge. Groups 2 and 3, on the other hand, engaged in a cooperation task to absorb the study content. The difference between Groups 2 and 3 is that Group 2's learning content was related to mental illness, while Group 3's learning content was not. On even-numbered days, participants in all three groups engaged in small talk and answered questions posed by the chatbot.
• Figure 3: Illustrative example of content topic designs.
• Figure 4: Illustrative example of interaction mode designs.
• Figure 5: Each cooperation task was comprised of two rounds. Within each, the chatbot and the human user alternated between producing a summary of the learning material and correcting the summary provided by the other party. Following the completion of two rounds, the chatbot presented the user with the final summary.