Radical Embodied Cognitive Science | Anthony Chemero, 2013

Keypoints:

  • Historical perspective on Radical Embodied Cognitive Science (RECS)
  • Interdisciplinary: phenomenology + ecological psychology + dynamical systems modeling
  • Direct descendant from Jamesian functionalist approach to psychology.
  • Current cognitive psychology makes occasional reference to the body, but departs mainly from mental representations (so this is cognitive-friendly and not radical embodied cognitive science)
  • RECS instead is "skeptical of the explanatory usefulness of mental representations"

Introduction

Wundtian Structuralist vs Jamesian Functionalist schools.

Wundt follows Descartes' approach: phenomenon should be understood by breaking it down in its constituent parts and then recombine those into more complex mechanism (cf. contemporary neomechanism).

Thus, Wundt’s psychology focused on simple sensations, that is, the atoms of conscious experience. Its goal was to see how these simple sensations could be combined into more complex experiences. In contrast, William James created a psychology in the broadly Darwinian tradition. Rather than focusing on the atoms of conscious experience, he said, psychology should tell us how the mind adapts us to our environment. That is, to understand, say, emotions, we needed to understand the role that emotions play in adapting an animal to its environment over behavioral, developmental, and evolutionary timescales.

The scope of Wundtian psychology is thus very narrow, whereas James includes very high-order phenomena such as feelings or desires.

Very cool quote of James, W. (1890). The principles of psychology. New York, NY: Henry Holt.:

“The boundary-line of the mental is certainly vague. It is better not to be pedantic, but to let the science be as vague as its subject, and include such phenomena as these if by so doing we can throw any light on the main business in hand” (p. 6).

Descendants of Jamesian functionalism are Skinnerian behaviorism, and Gibsonean ecological psychology.

A descendant of Wundtian Structuralism is the cognitive revolution. Jerry Fodor stated the modern contribution is basically the computer metaphor. Since the 70s the main approach has been cognitive psychology in the line of Wundtian structuralism.

This paper on RECS offers an approach based on the Jamesian tradition:

The psychology, to be described below, combines nonlinear dynamical modeling with ideas about the nature of the mind from James Gibson and/or phenomenological philosophers, and is currently practiced by two tight-knit and loosely interacting groups of scientists.

Two major RECS groups (the terms is coined by Andy Clark):

1) ecological psychologists * Focus on Gibson * E.g. Turvey, Shaw 2) enactivists * Focus on phenomenology * E.g. Varela, Thompson

Major goal of this paper and overall owrk of Chemero is to bring these groups together and make them realize they mostly agree.

Embodied Cognitive Science

"Embodied" usually used in a very weak sense. Historical overview, chronologically:

Gibson, versus computational theories of mind:

Gibson’s ecological approach to perception has three major tenets. First, perception is direct, which is to say that it does not involve computation or mental representations. That is, Gibson thought that perception was not a matter of internally adding stored information to sensations. Second, perception is primarily for the guidance of action, and not for action-neutral information-gathering. We perceive the environment in order to do things. The third tenet follows from the first two: Because perception does not involve mental addition of information to stimuli, yet is able to guide behavior adaptively, all the information necessary for guiding adaptive behavior must be available in the environment to be perceived. Thus, the third tenet of Gibson’s ecological approach is that perception is of affordances, that is, directly perceivable, environmental opportunities for behavior.

Affordances are thus not subjective or objective, and equally environmental as behavioral. This idea is still relevant in a toned down form today.

Situation semantics (Barwise and Perry) provide semantics for Gibsonian psychology, because they describe that for understanding meaning and cognition we must realize that "thinkers are spatially located (i.e. situated) and so have only incomplete, locally available information at their disposal." (146)

Important features:

1) Indexicals (you, here, now) are central and considered ubiquitous 2) Meaning is possible without reference to mental representations * Instead: "The meaning of thoughts and sentences is a matter of the relationship between thinkers/speakers and information in their environments."

How Rodney Brooks picked up this idea:

When Brooks said that his simple, mobile robots are situated, he meant that, because they are in the midst of a changing world, they do not need to use representations of the world to plan or guide their behavior. Instead they interact with the world itself. The idea is that there is no need to store information on-board and make predictions about how things will change during an action, when you can just act and check again. Brooks summed up this antirepresentationalism with the slogan, “The world is its own best model.” (146)

But Brook adds something. The thinker is not merely situated, but also needs to be physically embodied in an environment. Embodiment implies situatedness.

Current ECS does use representations, but in a limited and very specific action-oriented manner:

The representations they call on are indexical-functional (Agre & Chapman, 1987), pushmi-pullyu (Millikan, 1995), action-oriented (Clark, 1997), or emulator representations (Grush, 2004). In what follows, I will refer to these collectively with Clark’s term action-oriented representations. Action-oriented representations differ from representations in earlier computationalist theories of mind in that they represent things in a nonneutral way, as geared to an animal’s actions, as affordances. (146)

ECS is however currently in a strange position, because it is still being practiced as a computational theory of mind, while having its roots in Jamesian functionalism that opposed that approach.

E.g. the extended mind position (Clark, 1997) describes both mind, body and behavior in computational terms, as the mind quite literally expanding beyond the body (Also called "wide computationalism").

Radical Embodied Cognitive Science

In other words, ECS took inspiration on the previous thinkers, but did not take over their anti-representationalism (which would be against the computational theory of the mind)

Radical ECS instead also takes over the anti-representationalism. Chemero therefore argues that RECS is not really a radicalization of ECS, but simply stays through to its historical origins that trace back to Jamesian functionalism.

So:

1) RECS is part of a "venerable scientific tradition" and is no "radicalism for its own sake" 2) This also means that RECS is not just an arbitrary push for antirepresentationalism and a break away from the mainstream computational approach. Quite the oppositie: antirepresentationalism was always the heart of the matter from the origin of this tradition.

RECS is inspired by phenomenologists (Heidegger, Merleau-Ponty), and Gibson, whose account of cognition always starts with a critique of the "structuralist" or computationalist view of cognition, that goes back to Descartes. Against the view that the world causally affects our senses, after which an internal representation is built of that world. I don't have to repeat that criticism here.

However, this general criticism and the alternative view that follows from it is hard to turn into empirical science. So how do we do that? With dynamical systems modeling!

Dynamical Systems Modeling

The goal is to try and explain action without using a internal conversion from sensory representations to motor representations.

Idea: humans self-organize like other complex systems also do. We can try to model this process.

1) Bridge between theory and practice, allows data gathering 2) Can model intelligent behavior using dynamic parameters by modeling both environment, agent, and their coupling with differential equations. I.e. agent and environment form one dyanmical system and are not strictly separated. Dynamical modeling describes how this whole systems unfolds over time.

So in a formula: RECS = Phenomenology + Dynamical Systems Modeling

Example

See Dotov, Nie & Chemero, 2010 for the testing of a derived claim of Heidegger in SuZ about the transition between readyness-to-hand and unreadiness-to-hand. A ready-to-hand tool only becomes noticable as an object once its unreflective use is interrupted, e.g. when it breaks down.

To test Heidegger’s theory, we had participants play a simple video game, in which they used a mouse to control a cursor on a monitor. Their task was to move the cursor so to “herd” moving objects to a circle on the center of the monitor. At some point during the trial, the connection between the mouse and cursor was temporarily disrupted, so that movements of the cursor did not correspond to movements of the mouse. Later, control would return. We measured their hand movements to determine whether they showed 1/f scaling, a property of physiological systems that would indicate the mouse was being treated as part of a single system along with the rest of their body (Riley & Holden, 2012; West, 2006). As predicted, the hand-mouse movements of the participants exhibited 1/f scaling while the video game was working correctly. The 1/f scaling decreased, almost to the point of exhibiting pure white noise, during the mouse perturbation, then returned to 1/f scaling when they regained control. So, while participants were smoothly playing the video game, they were part of a human⫺computer system that had the same pattern of variability as a well-functioning physiological system; when we temporarily dis- rupted performance, that pattern of variability temporarily disap- peared. This is evidence that the mouse was experienced as ready-to- hand while it was working correctly, and became unready-to-hand during the perturbation (see Dotov et al., 2010, for more details). It is important to note the role of dynamical modeling. The

Pretty cool that in this way a claim of Heidegger could be tested empirically!

If it turns out dynamic systems modeling is a good fit for this theory, it can in turn be used to propose new experiments and generate hypotheses.

Chemero ends with a plea for explanatory pluralism.