Focusing on Individual Risk: Personality-Based Content Adaptation in Virtual Agents for Safety Education
Year: 2026
Type: conference-paper
Abstract: In intelligent virtual agents (IVAs), adaptation strategies based on individual user characteristics are important for enhancing the effectiveness of educational and behavior change interventions. However, prior research on personality-based adaptation in IVAs has predominantly focused on adapting how information is delivered—such as persuasion strategies and communication styles—while the approach of adapting what to focus on, that is, shifting the content focus itself according to individual characteristics, remains underexplored. In this study, we propose a personality-adapted IVA for safety education that shifts the focus of accident analysis information according to users’ personality-based risk profiles. The dialogue structure incorporates user-driven reasoning about accident causes and action planning to support active processing of the focused information. We conducted a field experiment with 74 employees of a power company, assigning participants to either an adaptive condition, which presented personality-focused information, or a control condition, which presented general safety information, over six dialogue sessions spanning three weeks. Comparing pre- and post-intervention importance ratings of accident causes corresponding to each participant’s targeted personality trait, we found that only the importance ratings for the targeted trait selectively increased in the adaptive condition, while no changes were observed for non-targeted traits or in the control condition. This selective pattern demonstrates that shifting content focus according to individual risk factors is an effective IVA adaptation strategy for promoting users’ recognition of accident risks linked to their own personality traits. Furthermore, these findings suggest that content focusing in IVAs can serve as a means of helping users identify and become aware of their own internal states—cognitive and behavioral tendencies—that are relevant to specific outcomes.
Institutions: Gifu University, Chubu Electric Power (Japan), Waseda University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7207743389