Using systems factorial technology for global model analysis of ACT-R’s core architectural assumptions

Cognitive architectures (CAs) are unified theories of cognition which describe invariant properties in the structure and function of cognition, including how sub-systems (e.g., memory, vision) interact as a coherent system. One problem stemming from the size and flexibility of CAs is deriving critical tests of their core architectural assumptions. To address this issue, we combine systems factorial technology (SFT) and global model analysis (GMA) into a unified framework called SFT-GMA. In the framework, the prediction space is defined in terms of qualitative classes of SFT models, and GMA identifies constraints on this space based on core architectural assumptions. Critical tests are then derived and tested with SFT. Our application of SFT-GMA to ACT-R revealed two key insights: (1) we identified critical tests despite many degrees of freedom in model specification, and (2) ACT-R requires serial processing of perceptual stimuli under most conditions. These processing constraints on perception are at odds with data reported in several published experiments.