Introduction
This essay explores the key similarities and differences between two prominent approaches to understanding cognition: the dynamical systems approach, often metaphorically described as ‘mind as a Watt governor,’ and the Computational-Representational Understanding of Mind (CRUM), referred to as ‘mind as computation.’ Drawing primarily from foundational materials in cognitive science, including Bermúdez’s textbook, this discussion situates these perspectives within the broader context of cognitive mechanisms and models. The essay first outlines the core assumptions of each approach, then compares their similarities and differences, and finally addresses the weaknesses of CRUM while highlighting the advantages of the dynamical systems approach. By doing so, it aims to provide a clear and balanced analysis of how these theories explain the complex nature of cognition, particularly in terms of temporal dependencies between mind, body, and environment.
Core Assumptions of Each Approach
The CRUM approach, rooted in the computational theory of mind, posits that cognition operates like a computer. It assumes that mental processes involve the manipulation of symbols or representations according to formal rules, akin to software running on hardware. This perspective, explored in depth during earlier modules on ‘mind as information-processor,’ views the mind as a system that processes inputs (sensory data) to produce outputs (behaviours or decisions) through discrete, sequential computations (Bermúdez, 2022, p. 112). The brain, in this metaphor, is the hardware, while mental states are the software, governed by representational structures and algorithms.
In contrast, the dynamical systems approach, often illustrated by the metaphor of the ‘mind as a Watt governor,’ suggests that cognition emerges from continuous, interdependent interactions between mind, body, and environment. Drawing from content in the module on ‘mind, body, and environment,’ this perspective likens cognition to the self-regulating mechanism of a Watt governor, a device that maintains equilibrium in a steam engine through feedback loops (Bermúdez, 2022, p. 245). Rather than discrete computations, cognition is seen as a dynamic process unfolding over time, shaped by real-time interactions and feedback within a coupled system. This approach rejects isolated representational models, instead emphasising embedded and embodied processes.
Similarities Between Dynamical Systems and CRUM
Despite their fundamental differences, both approaches aim to explain how cognition enables organisms to interact with and adapt to their environments. Indeed, each framework seeks to account for complex mental phenomena—such as perception, learning, and decision-making—through systematic processes. For instance, both acknowledge the importance of inputs and outputs in cognitive activity, albeit interpreted differently: CRUM sees inputs as symbolic data processed into behaviours, while dynamical systems view them as continuous perturbations within a coupled system (Bermúdez, 2022, p. 118, p. 250). Additionally, both approaches, to some extent, draw inspiration from mechanical or engineered systems—computers for CRUM and self-regulating devices for dynamical systems—using these as metaphors to conceptualise abstract cognitive processes.
Furthermore, both theories have been applied across various cognitive domains, including problem-solving and action coordination, demonstrating their versatility as explanatory tools. They share an underlying commitment to scientific rigour, seeking to ground cognitive science in observable and testable mechanisms, whether through computational simulations or dynamical models of behaviour (Bermúdez, 2022, p. 255). However, their methods of achieving this understanding diverge significantly, as discussed below.
Differences Between Dynamical Systems and CRUM
The most striking difference lies in their conceptualisation of cognitive processes. CRUM relies on a representational and static view, where mental states are discrete, internal symbols manipulated by predetermined rules. This approach often isolates the mind from the body and environment, treating cognition as a largely internal affair (Bermúdez, 2022, p. 115). In contrast, the dynamical systems approach prioritises continuous, temporal dynamics, arguing that cognition cannot be separated from the body or the world in which it operates. It views the mind as inherently embedded, emerging from ongoing interactions rather than isolated computations (Bermúdez, 2022, p. 248).
Another key difference is their treatment of time. CRUM typically models cognition as a series of sequential steps, akin to a computer program executing line by line, which can overlook the real-time, fluid nature of mental activity (Bermúdez, 2022, p. 120). Dynamical systems, however, explicitly incorporate time as a central factor, using mathematical models to describe how cognitive states evolve continuously through feedback loops and environmental coupling (Bermúdez, 2022, p. 252). This makes the latter arguably more suited to explaining processes requiring adaptation to unpredictable changes, such as coordinating actions in a rapidly shifting context.
Weaknesses of the CRUM Approach
While CRUM has been instrumental in advancing cognitive science, particularly through its influence on artificial intelligence and early computational models, it has notable limitations. One primary weakness is its struggle to account for embodied and situated aspects of cognition. By focusing on internal representations, CRUM often neglects the role of the body and environment in shaping mental processes, which can lead to an incomplete picture of phenomena like perception or motor control (Bermúdez, 2022, p. 122). For instance, it struggles to explain how a person navigating a crowded street integrates sensory and physical feedback in real time, as its sequential processing model does not easily capture such immediacy.
Additionally, CRUM’s reliance on discrete, rule-based computation can be overly rigid, failing to address the flexibility and adaptability required for many cognitive tasks. Learning, for example, often involves non-linear and context-dependent adjustments that are difficult to model as purely symbolic operations (Bermúdez, 2022, p. 125). This limitation becomes particularly evident when considering chaotic or unpredictable environments, where static representations may not suffice.
Advantages of the Dynamical Systems Approach
The dynamical systems approach offers several advantages, particularly in addressing the shortcomings of CRUM. Firstly, it provides a more holistic framework by integrating mind, body, and environment into a unified system. This perspective is especially effective in explaining cognitive processes like perception and action coordination, where real-time interaction with the world is crucial. For example, Bermúdez (2022, p. 250) highlights how dynamical models can capture the fluid adjustments a cyclist makes to maintain balance, something CRUM struggles to represent without complex, ad-hoc rules.
Moreover, dynamical systems excel in modelling adaptability to complex, unpredictable settings—a critical feature for cognitive processes in natural environments. If one were designing an artificial intelligence system, such as a robot navigating uneven terrain, a dynamical systems ‘mind’ would likely be superior. Its emphasis on continuous feedback and temporal dependencies allows for ongoing adjustments to unexpected changes, unlike the more rigid, pre-programmed responses typical of a CRUM-based system (Bermúdez, 2022, p. 253). This adaptability underscores a significant practical advantage, especially for AI applications requiring robustness in real-world scenarios.
Conclusion
In summary, while both the dynamical systems approach and CRUM offer valuable insights into cognition, they differ fundamentally in their assumptions and explanatory power. CRUM’s computational, representational focus provides a structured model but falters in capturing embodied and dynamic aspects of cognition, particularly in unpredictable contexts. Conversely, the dynamical systems approach, with its emphasis on temporal interactions and system coupling, offers a more flexible and integrated perspective, proving advantageous for explaining real-time, environment-dependent processes. These differences highlight broader implications for cognitive science, suggesting that a complete understanding of the mind may require integrating elements of both frameworks. Future research could explore hybrid models to combine the precision of CRUM with the adaptability of dynamical systems, potentially offering a more comprehensive account of cognition.
References
- Bermúdez, J. L. (2022). Cognitive Science: An Introduction to the Science of the Mind. 4th Edition. Cambridge University Press.
