Understanding the workings of human cognition remains a fundamental scientific challenge: while the basic building blocks of the brain are well known, as is their general organization, there is no agreement on how cognition emerges from their interaction. Unified theories of cognition consisting of invariant mechanisms and representations, implemented computationally as cognitive architectures, have been proposed as a way to organize the empirical findings and master the complexity of neural systems. If successful, they would also represent the most promising way of engineering artificial systems capable of general intelligence, a.k.a. Strong AI or AGI.
ACT-R is an integrated computational cognitive architecture resulting from decades of cumulative effort by an international community of cognitive researchers. It consists of a modular framework (see figure 1) with the following components: a) procedural and declarative memory modules, including both symbolic and subsymbolic (i.e., statistical) representation and learning mechanisms; b) perceptual and motor modules that incorporate many known human factors parameters and provide principled limitations on the interaction with an external environment; c) a constrained modular framework for incorporating additional factors such as fatigue and emotions that are not currently part of the architecture; and d) asynchronous interaction between modules that assemble small, sub-second cognitive steps into complex streams of cognition to accomplish high-level functionality.
Models built using the architecture can learn to perform complex dynamic tasks while interacting directly with the same environment as human users. ACT-R can account for all quantitative measures of human performance, from behavioral measures such as response time, percent of correct responses and eye movements, to fine-grained neural measurements such as EEG and fMRI. Hundred of cognitive models have been validated against experimental data, for tasks ranging from performing simple psychology experiments to controlling complex information systems such as air traffic control.