JST_ICAITD_2025_2

Abstract:

This study explores how empathetic reframing can be realized in human–AI dialogue through a structured abductive process. We propose a multi-agent dialogue architecture that coordinates personal context, external information, and nudge-based response design to support reinterpretation while preserving user agency. Using a tourism consultation scenario—where users express strong preferences for "authentic" travel  experiences—we examine how the system externalizes and contrastively presents abductive hypotheses in the final response. Rather than evaluating internal reasoning alone, the system focuses on the structural presentation of interpretations. Experimental results show that reframing depends not only on hypothesis selection, but on how personal meaning and situational context are grounded and exposed in dialogue. These findings suggest that within the examined settings, reframing appears to emerge as a property of dialogue structure rather than isolated reasoning mechanisms.