26 Feb 2013

Andy Clark. 8.6 of Being There, “Continuous Reciprocal Causation”, summary


summary by
Corry Shores
[
Search Blog Here. Index-tags are found on the bottom of the left column.]
[Central Entry Directory]

[Posthumanism Entry Directory]

[Andy Clark, Being There, entry directory]

 

[My own commentary is in brackets. All boldface and underlining is my own.]



Andy Clark

Being There:
Putting Brain, Body, and World Together Again

Ch.8
Being, Computing, Representing


Part 8.6
Continuous Reciprocal Causation



Brief Summary:

Separate parts of a system can be in a state of continuous reciprocal causation, meaning that the behavior of each part simultaneously affects the behavior of the other parts. In such cases, it is not best to explain the whole system’s by analyzing the system into insulated parts. And also, representational accounts might not best explain how one part can be found internally affecting another part.



Summary

[Recall that a position is “representationalist if it depicts whole systems of identifiable inner states (local or distributed) or processes (temporal sequences of such states) as having the function of bearing specific types of information about external or bodily states of affairs”. (147a)] Clark will offer one last way to make a strong anti-representationalist argument. He will appeal to “the presence of continuous, mutually modulatory influences linking brain, body, and world.” (163b) Clark previously described the neuronal processes involved in vision, which had “hints of such mutually modulatory complexity in the interior workings of the brain itself.” (163b) Clark now wonders what if “something like this level of interactive complexity characterized some of the link among neural circuitry, physical bodies, and aspects of the local environment?” (163b)

 

[Consider if a radio and a transmitter were near one another, and the transmitter is broadcasting music from a turntable, also nearby. This means that low frequencies playing on the radio will disrupt the needle on the record, but the disruption of the needle on the record will change what the radio is playing.] Clark gives this example.

Consider a radio receiver, the input signal to which is best treated as a continuous modulator of the radio’s “behavior” (its sound output). Now imagine (here is where I adapt the analogy to press the point) that the radio’s output is also a continuous modulator of the external device (the transmitter) delivering the input signal. In such a case, we observe a truly complex and temporally dense interplay between the two system components – one which could lead to different overall dynamics (e.g. of positive feedback or stable equalibria) depending on the precise details of the interplay. The key fact is that, given the continuous nature of the mutual modulations, a common analytic strategy yields scant rewards. The common strategy is, of course, componential analysis, as described in chapter 6. To be sure, we can and should identify different components here. But the strategy breaks down if we then try to understand the behavior unfolding of one favored component (say, the receiver) by treating it as a unity insulated from its local environment by the traditional boundaries of transduction and action, for such boundaries, in view of the facts of continuous mutual modulation, look arbitrary with respect to this specific behavioral unfolding. They would not be arbitrary if, for example, the receiver unit displayed discrete time-stepped behaviors of signal | receiving and subsequent broadcast. Were that the case, we could reconceptualize the surrounding events as the world’s giving inputs to a device which then gives outputs (“actions”) which affect the world and hence help mold the next input down the line – for example, we could develop an interactive “catch and toss” version of the componential analysis, as predicted in chapter 6. (163-164)

[So if we were to analyze for example the component of the radio as if insulated from its environment, we would not know where to begin, assuming that the process had already begun. But if each causal event happened in temporal steps with gaps between, then we could analyze the components of the causal relation.]


Clark offers a second example (from Randy Beer). [First consider this description of oscillating or reverberating circuits in Marieb and Hoehn’s Human Anatomy & Physiology (quoting):

In reverberating, or oscillating, circuits, the incoming signal travels through a chain of neurons, each of which makes collateral synapses with neurons in a previous part of the pathway.

As a result of the positive feedback, the impulses reverberate (are sent through the circuit again and again), giving a continuous output signal until one neuron in the circuit fails to fire. Reverberating circuits are involved in control of rhythmic activities, such as the sleep-wake cycle, breathing, and certain motor activities (such as arm swinging when walking). Some researchers believe that such circuits underlie short-term memory. Depending on the specific circuit, reverberating circuits may continue to oscillate for seconds, hours, or (in the case of the circuit controlling the rhythm of breathing) a lifetime. (Marieb and Hoehn, 422d)


Andy Clark’s second example involves such oscillating neurons,] he writes:

Consider a simple two-neuron system. Suppose that neither neuron, in isolation, exhibits any tendency toward rhythmic oscillation. Nonetheless, it is sometimes the case that two such neurons, when linked by some process of continuous signaling, will modulate each other's behavior so as to yield oscillatory dynamics. Call neuron 1 "the brain" and neuron 2 "the environment." What concrete value would such a division have for understanding the oscillatory behavior? (164a.b)

[So the neurons mutually modify one another, because they have both inputs from and outputs to one another.]


When we are interested in the behavior of the two insofar as they are mutually affecting one another, it would not make sense to analyze the workings into insulated components, even though indeed the system is made of discrete parts.

in the case of biological brains and local environments it would indeed be perverse—as Butler (to appear) rightly insists—to pretend that we do not confront distinct components. The question, however, must be whether certain target phenomena are best explained by granting a kind of special status to one component (the brain) and treating the other as merely a source of inputs and a space for outputs. In cases where the target behavior involves continuous reciprocal causation between the components, such a strategy seems ill motivated. In such cases, we do not, I concede, confront a single undifferentiated system. But the target phenomenon is an emergent property of the coupling of the two (perfectly real) components, and should not be "assigned" to either alone. (164c.d)


Such continuous reciprocal causation is common in our everyday lives.

Nor, it seems to me, is continuous reciprocal causation a rare or exceptional case in human problem solving. The players in a jazz trio, when improvising, are immersed in just such a web of causal complexity. Each member's playing is continually responsive to the others' and at the same time exerts its own modulatory force. Dancing, playing interactive sports, and even having a group conversation all sometimes exhibit the kind of mutually modulatory dynamics which look to reward a wider perspective than one that focuses on one component and treats all the rest as mere inputs and outputs. Of course, these are all cases in which what counts is something like the social environment. But dense reciprocal interactions can equally well characterize our dealings with complex machinery (such as cars and airplanes) or even the ongoing interplay between musician and instrument. What matters is not whether the other component is itself a cognitive system but the nature of the causal coupling between components. Where that coupling provides for continuous and mutually modularity exchange, it will often be fruitful to consider the emergent dynamics of the overarching system. (165a.b)

[This is like Deleuze’s notion of rhythm in Spinoza’s affection, see the end of section 6 of my paper “Body and World in Merleau-Ponty and Deleuze”:

Our active self-affection and adaptive interaction with the world around us is what Deleuze here calls "rhythm." He also offers the example of swimming through a powerful wave. When we collide with the wave, its affection begins to decompose our body. Yet, by self-affectively altering the arrangements of our own body's parts, we may swim in conjunction with the wave and together form a larger composite body. Deleuze suggests another illustration to explain more clearly how affective rhythm involves couplings of continuous affective variations. He has us consider a dual improvisation of a violin and a piano. On the one hand, each one needs to improvisationally choose its own development. Yet, the musicians' decisions will influence how the other plays in concord with it. So, in order for both instruments to maintain their differential co-composition, they must make self-modifications that are differentially compatible with those of the other player. (Shores 203)

]

 

So when there is continuous reciprocal causation, there is little use for an analysis that looks at the parts of such systems as if they were separate.

Thus, to the extent that brain, body, and world can at times be joint participants in episodes of dense reciprocal causal influence, we will confront behavioral unfoldings that resist explanation in terms of inputs to and outputs from a supposedly insulated individual cognitive engine. (165c)

Clark thinks that there are then only two possibilities for the use of internal representation for cognitive scientific explanations. (165c)


To understand the first possibility, we consider a complex neural network, called ‘A’. It is coupled with its environment, and part of its dynamics is an ability to sense whether it the environmental processes it is coupled to are present. “Imagine a complex neural network, A, whose environmentally coupled dynamics include a specific spiking (firing) frequency which is used by other onboard networks as a source of information concerning the presence or absence of certain external environmental processes—the ones with which A is so closely coupled.” (165d) So internally we might say the system has patterns for when it is coupled to external processes. Now we are to consider those signals normally coming from outside to be produced from the inside, causing the system to ‘imagine’ being engaged with the environment rather than physically being so. This would be like internal representation.

The downstream networks thus use the response profiles of A as a stand-in for these environmental states of affairs. Imagine also that the coupled response profiles of A can sometimes be induced, in the absence of the environmental inputs, by top-down neural influences, and that when this happens the agent finds herself imagining engaging in the complex interaction in question (e.g., playing | in a jazz trio). In such circumstances, it seems natural and informative to treat A as a locus of internal representations, despite its involvement, at times, in episodes of dense reciprocal interaction with external events and processes.” (163-164)


The other possibility is that even such inner processes cannot operate unless they are coupled, and thus there are nonrepresentational dynamics at play.

A second possibility, however, is that the system simply never exhibits the kind of potentially decoupled inner evolution just described. This will be the case if, for example, certain inner resources participate only in densely coupled, continuous reciprocal environmental exchanges, and there seem to be no identifiable inner states or processes whose role in those interactions is to carry specific items of information about the outer events. Instead, the inner and the outer interact in adaptively valuable ways which simply fail to succumb to our attempts to fix determinate information processing roles to specific purely internal, components, states, or processes. In such a case the system displays what might be called nonrepresentational adaptive equilibrium. (A homely example is a tug of war: neither team is usefully thought of as a representation of the force being exerted by the other side, yet until the final collapse the two sets of forces influence and maintain each other in a very finely balanced way.) (166b.c)


Thus,

Where the inner and the outer exhibit this kind of continuous, mutually modulatory, non-decouplable coevolution, the tools of information processing decomposition are, I believe, at their weakest. What matters in such cases are the real, temporally rich properties of the ongoing exchange between organism and environment. (166c)

Such instances do not challenge the representational model, because they do not fall under the class of cases best suited for representational explanations. Clark will explain this in the next section. (166d)

 

 

Clark, Andy. Being There: Putting Brain, Body, and World Together Again. Cambridge, Massachusetts/London: MIT, 1997.

 

Marieb, Elaine N., & Katja Hoehn. Human Anatomy & Physiology. London: Pearson, 2007.

 

Shores, Corry. “Body and World in Merleau-Ponty and Deleuze” in Sudia Phaenomenologica, vol.12, 2012, pp.181-209.

https://cdn.anonfiles.com/1360747598945.pdf



No comments:

Post a Comment