Forum

There is substantial evidence that humans engage their own decision-making mechanisms when predicting choices of others. According to one account, such predictions are implemented by the observer running a single simulation of the other person’s decision process in their mind and selecting the simulation outcome as the choice option. However, such an implementation would result in a large degree of stochasticity in predicted choices, thereby resulting in relatively low prediction accuracy.
In this talk, I will present an alternative idea - that observers could reduce uncertainty and increase prediction accuracy by simulating the decision between the same two options multiple times, sampling the outcomes of these simulations and then selecting the option corresponding to the most frequent outcome.

I will present results from a behavioural decision study employing the modified dictator game paradigm, aimed at testing whether we can identify behavioural indicators of sampling multiple simulation outcomes when participants predict other’s choices: specifically, whether participants show lower stochasticity and higher response times in predictions than self-decisions, after controlling for decision difficulty across these two conditions. I will complement these behavioural findings with analyses based on comparing the Drift-Diffusion model and a “sampling-of-decision-outcomes” model of choice. Together, these findings will shed light on potential differences between the mechanisms involved in making choices for ourselves and predicting the choices of others.