RL-based ‘gamification’ may mitigate the impact of anxiety in learning tasks by reducing maladaptive over-weighing of negative descriptive feedback

Anxiety is associated with a variety of learning deficits and poorer outcomes in academic settings. Recent years have seen a push for gamification of classroom learning content—i.e. reframing learning tasks to have a reinforcement learning (RL) based structure—as a promising way to mitigate the impact of anxiety in student learning. However, the mechanisms by which gamification impacts RL processes and the way in which it acts in naturalistic academic learning settings (which rely on descriptive narrative structure not usually present in classical RL tasks) are not yet known. We tested 152 participants in an online narrative-based, choose-your-own-adventure task to test the potential impact of gamification on RL processes in a naturalistic learning context. Model-free learning performance showed deficits in accuracy, confidence, and memory in high-anxiety compared to low-anxiety participants. Computational model fits showed significant differences in subjective weighing of negative descriptive feedback (but not in weighing points-based feedback), and in the correlation between descriptive feedback weighing and post-task confidence and memory. We conclude that gamification may specifically inhibit maladaptive overweighing of negative descriptive feedback by providing an additional / alternative feedback source as point-based reward, which in turn may mitigate deficits in accuracy, confidence, and memory in anxious learners.