Groundbreaking Study Reveals How Mental Simulation Shapes Predictions of Others' Choices
A new open-access paper published in iScience examines the cognitive processes people use when forecasting the decisions of others. Titled "Deciding to simulate: Cognitive mechanisms of predicting the decisions of others," the work was led by Erik Stuchlý of the University of Hamburg, with co-authors Sophie Bavard from the Paris Brain Institute and Sebastian Gluth, also at the University of Hamburg. The study, available at the ScienceDirect link provided by the authors, provides direct experimental evidence supporting the idea that individuals often rely on their own biased decision-making systems to simulate and predict what others will choose.
The research addresses a core question in cognitive science and decision neuroscience: do people truly engage in mental simulation when anticipating someone else's choices, or do they rely on more abstract rules or stereotypes? By adapting participants to different risk contexts and then asking them to predict choices for various hypothetical agents, the team demonstrated that prediction biases mirror participants' own adapted valuation processes in many cases, but not universally.
Background on Theory of Mind and Decision Simulation
Theory of Mind refers to the ability to attribute mental states such as beliefs, desires, and intentions to others. This capacity underpins social interactions, from negotiating contracts to choosing lunch spots based on a friend's preferences. Contemporary models suggest that predicting behavior combines general knowledge about human tendencies with mental simulation of the other person's perspective. Neuroimaging studies have previously hinted at overlap between self-decision and other-prediction mechanisms, particularly in value encoding areas of the brain.
The current work builds on this foundation by testing whether simulation occurs at the level of the valuation system itself. If predictions engage the same processes used for personal choices, then any context-induced bias in one's own decisions should transfer to predictions about others. The authors focused on risky choice, a domain where adaptation effects are well-documented: exposure to predominantly high- or low-probability gambles shifts how people evaluate subsequent options.
Experimental Design and Methods
The researchers conducted three pre-registered online experiments involving risky choices between a safe option with a guaranteed small payoff and a risky option offering a higher payoff with varying probabilities. In the adaptation phase, participants were divided into high-probability and low-probability groups. The high-probability group encountered gambles where the risky option was often advantageous on average, while the low-probability group faced the opposite context.
Following adaptation, participants entered a post-adaptation phase with balanced expected values. In the control experiment, they continued making choices for themselves. Experiment 1 asked them to predict choices for a vaguely defined "average" other person. Experiment 2 introduced two specific agents: one risk-averse and one risk-seeking, with predictions made in counterbalanced order. Drift diffusion modeling was applied to decompose decisions into parameters such as risk valuation, caution, and initial bias.
This design allowed the team to isolate whether adaptation effects on valuation carried over to predictions and whether the effect depended on the characteristics of the target agent.
Photo by National Cancer Institute on Unsplash
Key Findings from the Behavioral and Modeling Analyses
Results showed that the adaptation manipulation successfully biased participants' own choices in the control condition, consistent with shifts in risk perception. The same bias transferred to predictions for the risk-average agent in Experiment 1 and for the risk-averse agent in Experiment 2. However, no adaptation effect appeared when predicting the risk-seeking agent's choices. Participants' own risk tendencies also correlated with predictions for the average and risk-averse agents but not the risk-seeking one.
Drift diffusion modeling confirmed that the bias stemmed primarily from changes in the risk valuation parameter rather than alterations in decision caution or starting point bias. These patterns indicate that people do engage their own decision-making machinery when simulating others, but the extent of this simulation is modulated by perceived similarity or characteristics of the other person.
The findings align with the highlights reported in the paper: biased decision processes lead to biased predictions, risk perception underlies the effect, and the bias does not transfer equally to all agents.
Implications for Cognitive Science and Neuroscience Research
This work strengthens the simulation account of social prediction while highlighting its boundaries. For researchers in psychology and cognitive modeling, it offers a methodological template for probing mechanistic overlap between self and other decisions using adaptation paradigms and computational modeling. The results suggest that Theory of Mind processes are not purely abstract but grounded in embodied simulation of one's own valuation system.
In neuroscience contexts, the findings complement existing fMRI evidence on value representation during prediction tasks. Future studies could extend the approach to other decision domains such as intertemporal choice or social dilemmas, or incorporate neuroimaging to track valuation signals during prediction phases.
For academics and PhD students exploring decision neuroscience, the study underscores the value of combining behavioral adaptation with formal modeling to dissect cognitive mechanisms. It also raises questions about how individual differences in risk attitudes or social cognition might influence simulation tendencies.
Relevance to Higher Education and Academic Careers
Research of this nature contributes directly to training the next generation of scholars in interdisciplinary fields spanning psychology, neuroscience, and computational modeling. University departments focused on cognitive science can draw on these methods for graduate seminars and lab rotations. The open-access publication facilitates broad dissemination, allowing instructors worldwide to incorporate the findings into curricula on decision-making and social cognition.
Early-career researchers may find inspiration in the collaborative model demonstrated here, with authors spanning institutions in Germany and France. The pre-registered design and use of drift diffusion modeling exemplify rigorous practices valued in academic hiring and grant applications.
Future Directions and Broader Applications
The modulating role of agent characteristics points to several avenues for follow-up research. Investigating predictions for real rather than hypothetical agents, or incorporating cultural or demographic variables, could refine understanding of when simulation dominates versus when stereotype-based or rule-based strategies take precedence.
Practical implications extend to areas such as negotiation training, marketing, and artificial intelligence systems designed to model human behavior. Understanding the conditions under which people project their own biases onto others can inform interventions aimed at improving predictive accuracy in social and professional settings.
Longer-term, integrating these insights with developmental studies or clinical populations could illuminate how simulation abilities emerge and how they might be altered in conditions affecting social cognition.
Accessing the Full Publication
The complete study, including detailed methods, supplementary analyses, and figures, is freely available via the open-access license. Readers interested in the full experimental protocols, data, or modeling code are encouraged to consult the primary source for comprehensive details and to explore potential extensions of the paradigm.
