800 Section - Condition-Specific and Trait-Like Reproducibility of Valence Ratings in Healthy Adults

Objectives : In this study, we examined the reproducibility of valence ratings derived from facial expression analysis using Kokoro Sensor (ver. 1.2.2.0), focusing on both condition-specific and overall consistency. Our previous study transformed valence into three categories—positive, neutral, and negative—and assessed their validity using the content validity index (CVI) and κ statistics. However, agreement between the negative and neutral conditions was notably low, highlighting limitations of the categorical approach. To address this, the present study treated valence as a continuous variable and evaluated its trait-like stability.

Design: Forty healthy young adults rated their valence responses twice under three different emotional stimulus conditions. Intraclass correlation coefficients (ICC [3,1]) and 95% confidence intervals (CIs) were calculated for each condition and for the combined scores. The study was approved by the institutional ethics committee, and written informed consent was obtained from all participants.

Results: Results showed ICCs of 0.82 (CI: 0.68–0.90) for the positive, 0.61 (CI: 0.38–0.77) for the negative, and 0.05 (CI: -0.26–0.35) for the neutral condition. Although ICC for the neutral condition was markedly low, Bland–Altman plots indicated small mean differences, suggesting stable absolute ratings. When all conditions were combined, the valence scores showed excellent reproducibility with an ICC of 0.94 (CI: 0.92–0.96).

Conclusion: These findings suggest that treating valence as a continuous, trait-like construct may better capture stable intra-individual patterns that are obscured under condition-specific categorical frameworks. This approach could inform future emotion models that incorporate trait–state dynamics and contribute to the development of individualized affective computing systems.