Exploring the Role of Theory of Mind in Human Decision Making: Cognitive, Spatial, and Emotional Influences in the Adversarial Rock-Paper-Scissors Game

TL;DR

This study investigates how Theory of Mind (ToM) components—cognitive, spatial, and emotional perceptiveness—influence human decision making in repeated adversarial Rock-Paper-Scissors games. Using a three-study design with static bots, dynamic bots, and human opponents, the authors administered a comprehensive ToM survey and analyzed results with Exploratory Factor Analysis (EFA) and Structural Equation Modeling (SEM). They identify two latent ToM factors: Factor 1, encompassing recursive thinking, spatial reasoning, and social perceptiveness, which positively affects decision effectiveness against dynamic bots and humans, and Factor 2, tied to interpersonal skills and rationality, which negatively impacts performance. Although no single ToM metric robustly predicts performance across all conditions, the findings demonstrate strong interrelations among ToM facets and suggest that integrated, multi-dimensional ToM models are needed to anticipate and adapt to human behavior in adversarial settings. These insights inform the design of adaptive, human-centered AI systems and advance understanding of ToM’s role in human-machine collaboration.

Abstract

Understanding how humans attribute beliefs, goals, and intentions to others, known as theory of mind (ToM), is critical in the context of human-computer interaction. Despite various metrics used to assess ToM, the interplay between cognitive, spatial, and emotional factors in influencing human decision making during adversarial interactions remains underexplored. This paper investigates these relationships using the Rock-Paper-Scissors (RPS) game as a testbed. Through established ToM tests, we analyze how cognitive reasoning, spatial awareness, and emotional perceptiveness affect human performance when interacting with bots and human opponents in repeated RPS settings. Our findings reveal significant correlations among certain ToM metrics and highlight humans' ability to detect patterns in opponents' actions. However, most individual ToM metrics proved insufficient for predicting performance variations, with recursive thinking being the only metric moderately associated with decision effectiveness. Through exploratory factor analysis (EFA) and structural equation modeling (SEM), we identified two latent factors influencing decision effectiveness: Factor 1, characterized by recursive thinking, emotional perceptiveness, and spatial reasoning, positively affects decision-making against dynamic bots and human players, while Factor 2, linked to interpersonal skills and rational ability, has a negative impact. These insights lay the groundwork for further research on ToM metrics and for designing more intuitive, adaptive systems that better anticipate and adapt to human behavior, ultimately enhancing human-machine collaboration.

Figures (24)

• Figure 1: Schematic illustration of the study design.
• Figure 2: Pearson correlations between ToM metrics across the three studies. Significant correlations, with $p < 0.05$, are marked with an asterisk (*).
• Figure 3: RPS performance of participants in terms of (a) effectiveness and (b) prediction accuracy when playing with static bots, dynamic bots, and humans. Error bars indicate 95% confidence intervals. Note that the two plots have different ranges on the $y-$axis with percent score in $[0.5, 0.65]$ and prediction accuracy in $[0.25, 0.6]$
• Figure 4: Estimated coefficients from linear regression models for predicting the player's effectiveness and prediction accuracy when facing static bots, dynamic bots, and human players. The error bars indicate the 95% confidence intervals.
• Figure 5: A structural equation model (SEM) for estimating the player's effectiveness and prediction accuracy when facing static bots, dynamic bots, and human players. The numbers on the arrows represent the coefficients. Solid lines indicate paths significant at the 0.05 level.