Adaptive Control of Working Memory

人类工作记忆（WM）是一个高度灵活的系统，用于指导目标导向的（认知）行为。然而，WM的灵活性是以严格的容量限制为代价的，要求WM的内容不断适应当前的检索需求。本项目研究如何通过适应性认知控制机制实现这一目标。我们关注认知控制文献中三个众所周知的现象：频率一致性效应、一致性序列效应和误差后调整。这三种效应以前被用来描述适应性认知控制如何平衡对环境中相关和不相关刺激的注意。在这里，我们考虑声明性WM任务中的等效效应，以研究WM内容如何持续适应未来的检索需求。通过行为和脑电图研究，我们试图检验(1)在编码和维持阶段，表征水平、声明性WM内容是如何自适应调整的；(2)主动和追溯控制过程是如何实现这种适应性的；(3)对错误和冲突的监测是否以与注意力的自适应控制相同的方式促进了这种能力。最后，该项目旨在通过包含和验证基于冲突的自适应控制机制来推进现有的WM计算模型。

Human working memory (WM) is a highly flexible system optimized for guiding goal-directed (cognitive) actions. However, the flexibility of WM comes at the cost of a strict capacity limitation, requiring that the content of WM is continuously adapted to ongoing retrieval demands. The present project investigates how this is achieved by mechanisms of adaptive cognitive control. We focus on three well-known phenomena from the cognitive control literature: frequency congruency effects, congruency sequence effects, and post-error adjustments. These three effects have previously been used to describe how adaptive cognitive control balances attention to relevant and irrelevant stimuli in the environment. Here, we consider equivalent effects in declarative WM tasks to investigate how WM content is continuously adapted to future retrieval demands. Using behavioral and EEG studies, we seek to examine (1) on which representational levels, declarative WM content is adaptively adjusted during phases of encoding and maintenance, (2) how proactive and retroactive control processes achieve this adaptiveness, and (3) whether the monitoring of errors and conflicts contribute to this ability in the same way as in adaptive control of attention. Finally, the project aims to advance existing computational models of WM by including and validating conflict-based adaptive control mechanisms.