Bluetooth-sensed social presence is associated with immediate vigor and delayed fatigue: A multi-method time series analysis

Abstract

Social interactions affect emotional well-being, yet their temporal dynamics remain understudied in natural settings. We examined these patterns through an experience sampling method study with 80 participants over two weeks (k = 3,716 assessments), combining Bluetooth device counts (n = 123,574) as a social presence proxy with self-reported emotional states. Using linear mixed-effects models, continuous-time structural equation models, and multi-state Markov analyses, we uncovered complex temporal relationships between social presence and emotions. Our findings revealed that social presence was associated with increased immediate vigor and reduced dejection, followed by delayed fatigue. Bidirectional relationships emerged between social presence and negative mood, with peak effects occurring 3-4 h after initial contact. State-transition analyses demonstrated longer persistence of vigorous states compared to fatigued states, with social presence influencing these transitions. These results demonstrate how social-emotional processes unfold across multiple timescales, suggesting both immediate benefits and delayed costs of social presence.

Keywords: Psychology; Research methodology social sciences; Social sciences.

Graphical abstract

Figure 1

Valid responses across days with continuous spacing of time intervals

Figure 2

Temporal dynamics of social presence and negative mood revealed by continuous-time structural equation modeling (A) Social presence is having an increasing effect on negative mood, while negative mood has a decreasing effect on social presence. (B) Situation control has an instantaneous negative effect on social presence and negative mood. Parameter values are presented as mean and 95% confidence interval. (C) The lag of the peak effect of the cross-effect social presence on negative mood was 3 h and 7 min, white the lag of the peak effect of negative mood on social presence was 3 h and 50 min. Shaded areas represent the 95% credible intervals. (D) The half-life of the auto-effect of negative mood was 2 h 38 min, while the half-life of social presence was 1 h and 55 min. Shaded areas represent the 95% credible intervals.

Figure 3

States recovered by the two-state model (A) Distribution of the emotional states, suggesting a zero-inflated negative binomial. (B) Individual recovered emotional states and median (bold) across group. (C) Summary of the emotional state profiles across the group. Shows that state 1 is vigorous, in state 2 fatigue is prevalent. (D) Time courses of vigorous and fatigued state across the time of study.

Figure 4

Transition probabilities of the two-state model of vigorous and fatigued states (A) On average, the transition probability at any day to a fatigued state is 8%, in contrast, from fatigued to vigorous state is 14%. (B) Bluetooth devices and situation control both have effects on state transitions, with hazard ratios showing increased likelihood of transitions between states (approximately 1.2–1.4 times more likely) when these factors are present. Parameter values are presented as mean and 95% confidence interval, the red line at 1.0 represents the baseline hazard ratio.