The temporal dynamics of attentional allocation during counterfactual learning

Counterfactual learning, which refers to learning from what could have happened under different circumstances, is crucial for understanding decision-making in humans. Recent studies have well characterized the computational and neural processes underlying this aspect of learning. However, little is known about the temporal dynamics of attentional allocation towards the different alternatives faced by individuals, and their consequences. The present study explores the role of gaze and pupil dilation as potential indicators of counterfactual thinking during reinforcement learning tasks. By examining how these physiological measures correlate with the cognitive processes involved in evaluating both experienced and forgone outcomes, the research provides insights into the complex dynamics of human counterfactual learning. We recorded the gaze and pupil size of 36 participants engaged in a computer-based 2-arm bandit task with full information. The participants faced scenarios where they made choices and were then shown both the actual outcomes and hypothetical outcomes of unchosen options. While the overall pattern of allocation typically involved a first fixation on the factual outcome followed by a subsequent one on the counterfactual outcome, this pattern was affected differently by both outcomes’ value. On the other hand, pupil size reflected the outcome valence of the chosen option as well as the similarity between outcomes. These findings suggest that gaze dynamics and pupil dilation serve as effective biomarkers of counterfactual processing, reflecting the cognitive load and confirmation bias associated with such processes.