Dynamics and Soft Computation

A summary of section 1.9 of Andy Clark’s Supersizing the mind.

1.9. Dynamics and “Soft” Computation

DST’s anticipated response to the problem discussed in Time and mind:

There are ways of understanding this variability.
Any point in state space should be expressible as a vector of multiple components, allowing an approach that observes how those components change.

True, but this approach feels quite different from the “total state” vision emphasized by DST:

(a) Information-based readings of various substates and processes.
(b) Attending to internal information flow as much as to the unfolding of the total state over time.

Can this really be called a “radical alternative” to traditional computationalism?

It seems instead to be a hybrid: dynamical computationalism. On this view…

Internal information flow is as important as large-scale dynamics.
Some dynamical properties simultaneously play information-processing roles.
The block-copying scenario can also be explained from this perspective:
- Information processing: comparing properties of two blocks, for example.
- Dynamics: coupling (world and body) in which eye movements shift pointers into the world, emphasizing perceptual–motor timing, and so on.

Dynamical computationalism is already widely accepted in some respects:

A hypothetical example: interaction between two components, some CRC loops being internal. Here the two components are understood as exchanging (the dynamical perspective) different information or encodings (the computational perspective).
An actual example: interaction between gestural and verbal systems (to be discussed in chapter 6).

There will still be situations in which DST conflicts with computational/representational perspectives. In particular…

this will be so where the nature of the contributions being made by the “parts” is itself changing radically over time as a result of the multiple influences from elsewhere in the system.

Outside such extreme cases, however, DST offers an important additional dimension to the traditional perspective.