Mapping Neural Theories of Consciousness onto the Common Model of Cognition
The paper "Mapping Neural Theories of Consciousness onto the Common Model of Cognition" by Rosenbloom et al., presents an analytical approach to understanding neural theories of consciousness through the lens of the Common Model of Cognition (CMC). It systematically assesses how four prominent neural theories can be integrated using the CMC framework, providing insights into potential bridges between neuroscience and cognitive architecture.
Overview of Neural Theories of Consciousness
The authors explore the Global Neuronal Workspace Theory (GNWT), Integrated Information Theory (IIT), Recurrent Processing Theory (RPT), and Predictive Processing and Neurorepresentationalism (PP/NREP). These theories are distinguished by their emphasis on different aspects of consciousness, such as access consciousness (AC) and phenomenal consciousness (PC). GNWT links consciousness to global broadcasting mechanisms in the brain, primarily focusing on the prefrontal cortex for AC. IIT offers a structural perspective, defining consciousness through unified cause-effect structures, albeit with controversial postulates like panpsychism. RPT posits that consciousness arises from local and global feedback loops, distinguishing between PC and AC. Meanwhile, PP/NREP emphasizes hierarchical predictive modeling, generating high-level representations that may correlate with PC.
Mapping onto the Common Model of Cognition
The CMC is utilized as a consensus cognitive architecture outlining fundamental components such as working memory (WM), procedural and declarative long-term memories, perception, and motor modules. The paper delineates how each neural theory maps onto these components:
- Global Neuronal Workspace Theory (GNWT): GNWT's global workspace is interpreted as analogous to the CMC's working memory, which acts as a hub for inter-module communication. The recurrent ignition process in GNWT is seen as similar to the cognitive cycle of the CMC.
- Integrated Information Theory (IIT): IIT’s structural claims are mapped onto the concept of problem spaces within the CMC. Conscious states correlate with maximally integrated representations in these spaces, facilitated by operators driving transitions among them.
- Recurrent Processing Theory (RPT): RPT’s focus on feedback loops aligns with local recurrent processes within CMC modules, and the global feedback facilitated by WM mirrors AC according to RPT.
- Predictive Processing and Neurorepresentationalism (PP/NREP): PP/NREP maps onto the CMC by illustrating how multimodal representations within WM can support both AC and PC. Predictive elements are embedded within CMC’s procedural and declarative modules despite being more implicit.
Implications and Speculations
The analysis provides coherence between diverse consciousness theories and the CMC, suggesting pathways towards developing a neurocognitive theory of consciousness. This integrative approach may assist in applying consciousness models to artificial agents, providing a structured cognition framework analogous to human-like processes.
As AI advances, mapping neural consciousness theories onto cognitive architectures like the CMC could enhance cognitive systems' ability to emulate aspects of consciousness, potentially influencing areas such as decision-making and perception in autonomous agents.
Future developments in AI could leverage the CMC's abstraction to explore consciousness-like processes, potentially extending applications beyond human analogs to more sophisticated AI systems, which require nuanced representations and recurrent processing capabilities. The alignment of elements across theories into the CMC framework fosters cross-disciplinary insight that is crucial for progressing our understanding of cognitive and neural dynamics in both organic and synthetic beings.