Between architecture and model: Strategies for cognitive control

One major limitation of current cognitive architectures is that models are typically constructed in an “empty” architecture, and that the knowledge specifications (typically production rules) are specific to the particular task. This means that general cognitive control strategies have to be implemented for each specific model, which means a lack of consistency and constraint. Alternatively, cognitive control can be implemented as a part of the architecture itself, which is often implausible, because many strategies are learned and differ among individuals. A third solution is to assume cognitive control consists of learned strategies that can be used for many different tasks. The PRIMs theory (Taatgen, 2013) provides a modeling framework for this type of reuse. The approach is discussed using the example of working memory control in which I show that three different working memory tasks share the same strategic components to store and recall items. The broader impact of the work is that it shows that general cognitive skills may play a much more important role in understanding many aspects of cognition that are traditionally considered to be part of either the cognitive architecture or specific task knowledge, and therefore provides an important stepping stone towards the larger goal of unified theories of cognition.