Weak vs Strong Embodiment

Weak Embodiment

The idea that we have non-propositional representations. This approach doesn't take representations to be symbolic and these representations don't represent states of affairs in the world.

Instead, they call them B-Representations (body representations); they represent states of affairs of the body in a non-propositional way. B-representations are still features of the brain.

Representations are constituted by neural properties.

Pushes away from Fodor's 'mentalese' and LOT.

Mirror Neurons are important here. They are active, say, when I pick up a cup -- and they are active in exactly the same way when I see someone else pick up a cup. Potentially important for Social Cognition -- where we understand what other people are doing. This is important because it places the body front and center in the question of cognition because in order for the mirror neurons to become active in the first place, we need to have the kind of body that allows us to pick up cups. So, you need a body which requires a stronger link between the brain and the body than is required in Fodor's "Methodological Solipsism."

So -- mirror neurons are there for re-use. In order to understand what you are doing, I reuse the information from what I have done.

This is one example of the Weak Embodiment claim.

But, to emphasize, it is still weak because it is still representational and it's still putting the brain at the seat of cognition. So, it's not spreading cognition into the rest of the body like stronger forms of EC.

Strong Embodiment

Strong EC suggests that anatomy and movement are nontrivial contributors to cognition both prior to brain processing and subsequent to brain processing. Embodiment in this case means that extra-neural structural features of the body shape our cognitive process.

(this would be very much contrary to classic computationalism)

Strong EC views cognition and the body as a Dynamic System -- the states dynamically affects and depend on one another similar to the way a Watt Governor works.

The state one one part of the Watt Governor necessarily affects the states of other connected parts. Because the state of one part dynamically affects the states of other parts, you can't describe the effect sequentially. They operations happen simultaneously, not sequentially.

My Thought -- I like this idea. Something to think about.

So, let's consider how this affects embodiment in detail.

Part 2: Differences between dynamic and computational cognition

Timing

1. Linear Causality: cause --> effect
2. Circular Causality: cause --> effect --> cause --> effect, etc.

Peep this: Our body (and thus, perceptions) are constrained by our physical capabilities.

So how does this tie in with timing?

Here it is: perception depends on movement, and movement makes affordances apparent (circular)

"Cognition depends upon the kinds of experience that come from having a body with various sensorimotor capacities. [...] Perception and action are fundamentally inseparable in lived cognition.

My thought Why does this seem sort-of wrong to me? This is a bit like saying that a CPU depends on user input to process information. The CPU is a processor. Period. What does it process? Information; and some of that information comes from user input. But the input has to be in a certain format for the CPU to use it, which has to have been programmed by someone (user input) which might be seen as being "prior to CPU processing" in the same way that strong EC says that anatomy and movement are "nontrivial contributors to cognition". Right? What am I missing here? Ah. Maybe it's the issue of timing described above? Well, the Watt Governor example is not a correct example of a system that works instantaneously in a circular way. At the micro level, there is a chain of events that are chronologically linear. There is a linear causality. We just don't see it.

PhiloAssist clarifies:

Strong EC treats cognition as a single, integrated system rather than a brain that “receives input” from the body. But the car metaphor still leans too “modular” for the radical view:

• Car: engine, steering, tyres are separable parts bolted together; you can swap a tyre without changing the steering algorithm.
• Strong EC: brain, body, world are non-decomposable; change the arm’s length or the slope of the ground and the very control law you instantiate changes.

Think instead of a kayaker in a rapid: paddle, current, kayak, and body adjust to each other in the same millisecond; there’s no “input stage” followed by “output command”—the river’s force literally is part of the computational loop.

The explanatory role of mental representations

The Role of Computation in Cognition

• Advocates of LOT say there are rules that determine how the syntactic structures of symbols change (similar to linguistic grammar rules) e.g. add 's' to a noun to make it plural
• Can we get computation without positing internal rules?

Gas Canisters Let's say we have 2 gas canisters:

• Gas canister doesn't 'represent' Boyle's law
• But we can use one canister to predict the behavior of another.

OR

The Spring Scale

• It's an 'adding machine' that doesn't use computation.
• It has no 'internally stored' rules

So, the question is: Can this view of the Spring Scale be applied to human cognition?

Challenges:

1. Spring scale does one task
2. Isn't human cognition just a bit more complicated than that?

Martin Davies & Tony Stone

• We are justified in positing internal rules when we are describing transitions between representational states.
• There are no representations in gas canisters (no internally represented rules) because the canisters are not representing anything to themselves.
• So the gas canisters don't need to store internal rules
• We only need to posit rules when we need to explain transitions between representational states

For example:

Learning that the letter B has a certain sound requires an internally stored rule for the sake of future replication.

Connectionism

Representation and computation WITHOUT stored rules. Computation occurs within (or as a feature?) of the system.