Task attentiveness is driven by observations of external rewards rather than internal performance estimates

Attention is a cognitive process that directs mental resources toward relevant information, facilitating learning and decision-making. Although individuals can monitor and report their own attentional state, it remains unclear whether these self-reports are primarily driven by estimation of performance or external cues. In this study, sixty-one participants completed a probabilistic reinforcement learning task in which they made binary choices to maximize monetary rewards. Due to the probabilistic nature of the task, participants could make accurate choices, such as selecting the option most likely to yield a reward, yet still observe a negative outcome. During task performance, participants rated their attentiveness at random intervals using a visual analogue scale. Using Bayesian regression analysis and computational reinforcement learning model, we found that self-reported attention was more strongly influenced by external outcomes (reward delivery) than by task performance estimates. Furthermore, we examined the causal association between task performance and the experience of attentiveness. Time-lagged analysis suggested that attentiveness ratings did not predict future performance, and rather were only influenced by reward history. These findings suggest that individuals may rely more on external cues, such as rewards, rather than internal assessments of performance when reporting their level of attention.