Rapport-building dialogue strategy: a comprehensive analysis of human subjective, behavioral and facial emotional engagement
Purpose Rapport is a conversational aspect focused on relationship building. Because rapport can influence outcomes in collaborative tasks, this study proposes a concrete methodology for introducing rapport-building into dialogue systems, namely a rapport-building dialogue strategy, and evaluates its effectiveness. Design/methodology/approach We implemented a rapport-building dialogue strategy for virtual agents by prompting a large language model (LLM) based on psychological theories of rapport. We employed two dialogue approaches, a predefined sequence and a free-form approach, to guide dialogue generation and compared their performance through human evaluations. To analyze the human aspects of rapport further, we conducted subjective, behavioral and facial emotion assessments. Findings Our experimental results show that the system based on a free-form dialogue strategy, which allows the adaptive use of rapport-building strategies, achieves the highest subjective scores. Our human-centered analyses reveal strong associations between rapport perception and user experience, indicating that perceived rapport is closely aligned with engagement and experiential quality. We observe directional tendencies in behavioral and emotional patterns, with more expressive participants tending to report higher rapport scores, suggesting potential associations that warrant further investigation. Research limitations/implications While these findings provide valuable insights, the study's scope was limited to 20 Japanese graduate students; future work could benefit from extending this study with larger participant samples and cross-cultural investigations. Originality/value This study introduces: (1) a rapport-building dialogue strategy; and (2) a comprehensive analysis that distinguishes statistically-supported drivers of rapport from promising signals and non-influential factors. Together, they offer structured guidance for the development and evaluation of future rapport-building dialogue systems.
Authors
- Seiya Kawano (ORCID: https://orcid.org/0000-0002-5830-8169)
- Muhammad Yeza Baihaqi (ORCID: https://orcid.org/0000-0003-0235-3398)
- Ángel M. García (ORCID: https://orcid.org/0000-0002-1946-0520)
- Koichiro Yoshino (ORCID: https://orcid.org/0000-0001-6579-361X)
Institutions
- Kyoto Institute of Technology (JP)
- Nara Institute of Science and Technology (JP)
Publication Details
- Journal
- APSIPA Transactions on Signal and Information Processing
- Published
- 2026-09-24
- DOI
- https://doi.org/10.1108/atsip-02-2025-0115
- Primary Topic
- Social Robot Interaction and HRI
- Type
- article
- Field-Weighted Citation Impact
- 0.00