The ability to exert cognitive control is central to human brain function, facilitating goal-directed task performance. However, humans exhibit limitations in the duration over which they can exert cognitive control---a phenomenon referred to as cognitive fatigue. This study explores a computational rationale for cognitive fatigue in continual learning scenarios: cognitive fatigue serves to limit the extended performance of one task to avoid the forgetting of previously learned tasks. Our study employs a meta-learning framework, wherein cognitive control is optimally allocated to balance immediate task performance with forgetting of other tasks. We demonstrate that this model replicates common patterns of cognitive fatigue, such as performance degradation over time and sensitivity to reward. Furthermore, we discuss novel predictions, including variations in cognitive fatigue based on task representation overlap. This approach offers a novel perspective on the computational role of cognitive fatigue in neural systems.

One possibility that may facilitate learning difficult tasks is building a specific curriculum, i.e., to start with simpler tasks and continue with increasing task difficulty. While such curriculum learning has long been considered and investigated in educational settings, little research has been conducted on curriculum learning in well-controlled experimental settings. To fill this gap, we investigated curriculum learning in a perceptual decision-making task. Across two experiments, we tested whether different training conditions (ascending difficulty, descending difficulty, random difficulty, easy difficulty, and difficult difficulty) affected participants' performance in a subsequent testing phase. We found that participants trained on a sequence of varying difficulty (i.e., ascending difficulty, descending difficulty, and random difficulty) outperformed others from easy-only and difficult-only training. However, we found no differences during testing between the conditions with varying difficulty. These results suggest that varying the difficulty during training yields advantages when learning the random-dot-motion task.

In light of constraints inherent to empirical research, such as finite time and resources, there has been growing interest in using artificial intelligence to streamline the scientific process. However, despite advancements in automating scientific discovery, the implementation of strategies for sampling useful experiments remains a challenge. This metascientific study evaluates different experimental sampling strategies based on their effectiveness in advancing the discovery of linear models of human cognition based on synthetic data. We investigate the hypothesis put forth by Dubova et al. (2022) that random sampling of experiments is more effective than model-driven sampling. Indeed, the results of this study indicate that random sampling is more effective in a majority of cases, and that the underperformance of model-driven strategies can be attributed to a narrow sampling of the design space. Despite limitations in our approach, the work presented offers a novel framework for the metascientific study of autonomous empirical research.