CLARION architecture
Explanation
CLARION (Connectionist Learning with Adaptive Rule Induction ON-line) is a cognitive architecture developed by Ron Sun (Rensselaer Polytechnic Institute) since the 1990s. Its distinctive feature is the systematic attempt to model the distinction between conscious/explicit and unconscious/implicit processes in human cognition, combining subsymbolic representations (connectionist) with symbolic representations (rules).
The CLARION architecture has two levels in each subsystem: a bottom level (subsymbolic, distributed representations of neural-network type, implicit processing) and a top level (symbolic, explicit rules, conscious processing). The two levels constantly interact: implicit knowledge can be extracted as explicit rules (modelling how humans verbally articulate skills initially tacit); explicit rules can be internalised at the implicit level (modelling how practice converts deliberate skills into automatic ones).
CLARION has four main subsystems: action-centred subsystem (skills, action rules), non-action-centred subsystem (general knowledge), motivational subsystem (drives, goals, objectives), metacognitive subsystem (control, monitoring, regulation). Each with its dual implicit-explicit architecture. This organisation aims to capture the complexity of real human cognition, where multiple systems cooperate and compete.
One of CLARION's contributions is the quantitative modelling of the transition from conscious to automatic processes (motor skill acquisition, expertise in professional domains). Ron Sun and collaborators have published numerous studies comparing model predictions with human experimental data (problem-solving, implicit vs. explicit learning, decision-making), with good fits in many tasks.
CLARION has been applied in diverse domains: modelling complex task learning, operator-aid systems in industrial contexts, expertise and decision-making analysis, simulation of social processes (CLARION-architecture agents interacting). It is one of the few architectures that explicitly theorises the conscious/unconscious distinction as fundamental architectural organisation.
For the theory of consciousness, CLARION offers a concrete proposal on how the consciousness/unconsciousness dichotomy could be organised in a cognitive system. Consciousness would be associated with the explicit level (rules, symbols deliberately manipulable) and unconsciousness with the implicit level (distributed activations, parallel processing). This is simplifying compared with the complexity of human consciousness, but offers a testable architectural hypothesis and a computational testbed. CLARION does not solve the hard problem nor pretends to: it focuses on the functional and structural aspects of dual cognition. As a cognitive architecture with explicit focus on the conscious/unconscious distinction, it is a significant contribution to the panorama of computational cognitive science.
Strengths
- Explicit model of the conscious/non-conscious duality.
- Psychologically plausible bottom-up learning.
- Integration of motivation and metacognition.
- Applications in social dynamics and human learning.
- Principled symbolic-connectionist hybrid.
Main critiques
- Duality perhaps too rigid compared with the real continuum.
- Metacognition implemented in a limited way.
- Does not address qualia or the hard problem.
- Parameters often fitted post hoc.