Akihiro Miyamoto

Background: Sense of agency (SoA) is the subjective experience of controlling one’s actions and their effects in the world. SoA is important for behavioral change and goal achievement, and its measurement via digital devices has gained attention in health promotion and therapeutic interventions. Objective: This study describes longitudinal changes and individual variations in general explicit sense of agency (GenExp-SoA) and local implicit sense of agency (LocalImp-SoA), measured in daily life using a mobile app, and examines their associations with demographic and personality traits. Methods: One hundred adults in Japan participated in a 3-month, prospective observational survey, completing daily mobile app-based assessments of GenExp-SoA (using the Sense of Agency Scale) and LocalImp-SoA (using effect binding with the Libet clock method). Personality traits were measured via the Big Five questionnaire. Linear mixed-effects models were used to evaluate longitudinal changes and interactions with personal characteristics for both types of SoA and their relationship. We also examined state persistence and cross-lagged effects using continuous-time structural equation modeling. A total of 13,391 paired observations were analyzed. Results: GenExp-SoA decreased over time (β=−0.008, P <.001). The decrease was notable for women (β=−0.019, P <.001) and participants with extraversion scores higher than the median (β=−0.019, P <.001). An increase was observed in those aged 50‐59 years (β=0.023, P <.001). LocalImp-SoA also decreased (β=−0.003, P =.02), but its association with personal characteristics was limited. No significant association was found between the 2 forms of SoA. The variance patterns differed between measures, with a high intraclass correlation coefficient of 0.77 for GenExp-SoA and a very low intraclass correlation coefficient of 0.02 for LocalImp-SoA, indicating predominant within-individual variability for the latter. Continuous-time structural equation modeling showed stronger self-persistence for GenExp-SoA, minimal persistence for LocalImp-SoA, and negligible cross-lagged effects. Findings were robust in bootstrap and sensitivity analyses. Conclusions: The 2 forms of SoA measured in daily life settings displayed distinct temporal patterns, as well as different associations with individual characteristics. These findings may inform the design and evaluation of epidemiological studies and digital health interventions targeting the SoA, including measurement strategies. The results support lower-frequency, between-individual assessments for explicit agency and high-frequency, within-individual designs for implicit agency.

User-adaptive systems increasingly rely on latent psychological constructs to personalize interaction and support user well-being. Among these states is the sense of agency (SoA), which is the subjective experience of being in control of one’s actions and their effects. While it has received growing attention, it remains underexplored as a user-modelable construct. This work investigates how SoA relates to affective states and autobiographical memory during music listening, and whether these latent constructs can be inferred from observable music signals. We conduct a set of empirical analyses examining correlations, mediation, and moderation pathways between SoA, affect, and memory-related processes. We further evaluate the feasibility of predicting these psychological constructs from music features using supervised learning models. Our results show that SoA is systematically embedded within affective and autobiographical memory contexts, exhibiting structured statistical relationships and feasibility-level predictability from music features. Rather than proposing a deployed adaptive system, this work positions SoA as a latent user mental state that can be empirically characterized and modeled, providing a foundation for future personalized and agency-aware interactive systems.