Implementation and experimentation of a computational cognitive architecture

At the beginning, we introduce the cognitive architecture: CLARION (Connectionist Learning Adaptive Rule Induction On-liNe): its history of development and current research works briefly. Then, we discuss CLARION on the implementation and experimentation aspects which are the focus of this dissertation. For this purpose, we first give a detailed discussion on the architecture from the theoretical aspects, more specifically, its three comprising computational subsystems: ACS (Action-Centered Subsystem), NACS (Non-Action Centered Subsystem) and MS/MCS (Motivational/Meta-Cognitive Subsystem). Both the computational representations and the learning processes in each individual subsystem and the coordination between them are studied. Second, for implementation aspect, we discuss briefly the overall implementation structure of CLARION and then extend it to the issue of implementation of each computational subsystem: ACS, NACS and MS/MCS in great details. Following this, third, we addressed the experimentation aspect by introducing simulations of three typical human learning/reasoning tasks using CLARION: Process Control Task in which only ACS is involved; TOH (Tower of Hanoi) task which studied goal structure, top-down and bottom-up learning processes; AGL (Artificial Grammar Learning) in which ACS, NACS and their coordination are studied; At the end, comparisons with other cognitive models (ACT-R and SOAR) are discussed in details and the future work is pointed.