An essay by Andy Clark and David Chalmers published in the 1998 issue of the Oxford philosophy journal “Analysis.”¹ It sparked various discussions relating to embodied cognition, active externalism, and more.

Introduction

Two common answers to “Where does the mind stop and the world begin?” are 1) the mind is inside the skin and skull, and everything outside is the world, and 2) the externalist view that the mind need not be confined to the head. On the externalist view, it is reasonable to see part of the mind as outside the body. What “part” means varies with the specific philosophical position: semantic externalism, for example, argues that the meanings of words do not exist solely in people’s heads.

The authors argue for another position, active externalism. Explaining and advocating it is the purpose of this essay. Active externalism emphasizes that parts of cognitive processes exist outside the body (in the environment), and that the environment does not merely respond passively to active cognitive processes inside the body: it can itself play a highly active role. It is called “active” externalism because cognition outside the body is not simply there, but plays an active role.

Extended cognition

Imagine a shape and a socket that could hold it on a computer screen, and a task of deciding whether the shape fits the socket. An ordinary mental rotation task: participants rotate the shape in their heads to determine whether it fits.

The authors offer two variations.

First variation: A button lets participants actually rotate the shape on screen. Rather than imagining it, they simply press the button to rotate it.
Second variation: Instead of a button, a neural implant is placed in the brain. Participants can now command the computer to rotate the shape on screen simply by thinking, without pressing a button.

In each of these three situations (the ordinary setup, variation 1, variation 2), what should count as part of the cognitive process? To put it another way, how much cognition exists in each situation? The authors argue that we should consider all three equivalent.

If we insist that only what happens inside the head counts as cognition, the neural implant is clearly inside the head, so it must count. If we accept this, there is no fundamental reason to exclude the implant from cognition merely because it takes the form of a button outside the head.

An actual experiment (On Distinguishing Epistemic from Pragmatic Action) observed people playing Tetris and reached an interesting conclusion compatible with this argument. Rotating blocks in Tetris serves two purposes: pragmatic action, rotating a shape to actually change its orientation, and epistemic action, rotating it to see whether it fits. The authors argue that epistemic action should be viewed as part of the cognitive process, rather than merely an action. Epistemic actions create new information or transform existing information; this is called information self-structuring.

In this context, the world contains not only problems for us to solve but also the process of solving them. Epistemic action should be understood as modifying the world so that it can perform the work of problem-solving effectively.

Active externalism

The discussion above yields the following active externalist position:

Humans and the environment can be viewed as a single system interacting in both directions.
There is no particular reason to regard the environment as more passive than the human; both should be seen as playing equally active roles.

The authors argue that active externalism can explain many everyday activities more naturally. Explanatory economy is a characteristic of good theories. Consider Scrabble, a board game in which players arrange letter tiles horizontally and vertically to make words. While waiting their turn, players rearrange their tiles, then form a particular word when their turn arrives. To explain the cognitive processes involved, it is more reasonable to describe “rearranging the tiles” itself as thinking (an epistemic action) than as “internal cognitive processes plus a lengthy series of visual inputs and hand movements.”

(Scrabble is not widely known in Korea, so this may not be a very good analogy. Think of beginners moving the cards around in their hands while playing a card game.)

From cognition to mind

The examples above show that the environment can actively participate in cognitive processes in a narrow sense. The authors now introduce another thought experiment to argue for active externalism about the mind in a broader sense.

There are two people, Otto and Inga. Otto has Alzheimer’s disease, so his long-term memory does not work. He therefore writes everything he wants to remember in a notebook and consults it all the time.

Both want to visit a gallery to see a new exhibition: they share the same desire. Inga recalls that the gallery is on 53rd Street and heads there. In other words, she believes the gallery is on 53rd Street. What about Otto? He takes out his notebook, checks the gallery’s location, and heads to 53rd Street. Both Inga and Otto are now on their way to the gallery.

Schematically, their actions can be explained as follows:

Inga: Wants to go to the gallery. Goes to 53rd Street based on her belief about its location.
Otto: Wants to go to the gallery. Checks its location in his notebook and goes to 53rd Street.

If someone asked, “Why is Inga going that way?”, anyone would naturally answer, “Because she wants to go to the gallery and believes it’s that way.” But if someone asked, “Why is Otto going that way?”

Two answers are possible.

First: Because Otto wants to go to the gallery and believes it is that way.
Second: Because Otto wants to go to the gallery, believes its location is written in his notebook, checks the notebook, and finds it says the gallery is that way.

The authors propose this thought experiment, of course, to argue that the first answer is more appropriate than the second.

In their view, there is no need to spell out Otto’s act of checking the notebook. Inga, too, does not constantly keep in mind that the gallery is on 53rd Street; she has to retrieve it from long-term memory. Yet no one spells out this process. In Otto’s case we feel tempted to do so because the process is externalized (reading a notebook), but the argument is that this is not an essential difference.

The point is that the notebook is not only Otto’s long-term memory: what it contains are Otto’s beliefs. His beliefs are externalized.

If what makes information a belief rather than mere information is the role it plays, there is no reason to assume that role must be performed only inside the head:

The moral is that when it comes to belief, there is nothing sacred about skull and skin. What makes some information count as a belief is the role it plays, and there is no reason why the relevant role can be played only from inside the body.

In this passage, the authors extend the parity principle beyond cognitive processes to conscious experiences such as belief. The parity principle says that if something occurring in the world performs a function like something occurring in the head, it should count as part of the cognitive process.

If … a part of the world functions as a process which, were it to go on in the head, we would have no hesitation in accepting as part of the cognitive process, then that part of the world is (for that time) part of the cognitive process).

This passage is not in the original essay; it is excerpted from Supersizing the Mind, a later book by coauthor Andy Clark.

Beyond the outer limits

Extended beliefs

For information to count as a belief, it must play the following roles.

It is a constant presence in the person’s life, and when a situation calls for consulting it, the person almost always does so and acts on it.
It must be accessible whenever the person wants to consult it, without difficulty.
Once consulted, it is accepted without question.
It is there because at some point in the past the person trusted it and recorded it there.

The last criterion is debatable (beliefs can form unconsciously, memories can be manipulated, etc.), but the first three are necessary constituents of belief. Information possessing all these characteristics is a belief, rather than mere information. Otto’s notebook satisfies them all, so it constitutes extended belief.

Socially extended cognition

Could socially extended cognition exist? Could part of my mental state extend into another person? In principle, yes. In a highly interdependent couple, one partner’s beliefs could play a role for the other similar to Otto’s notebook. The keys are high levels of trust, stability, and accessibility. A waiter at a regular restaurant, a secretary, an accountant, or a colleague could partially serve this role. Language is important for socially extended cognition to work.

The extended self

Could there also be an extended self? Probably. Most of us already accept that the self extends beyond the boundaries of consciousness. Otto’s notebook, for example, is a key constituent of Otto’s identity as a cognitive agent. It makes more sense to view Otto as a single system comprising a biological body and external resources (the notebook).

This perspective has significant ethical implications.

Footnotes

consc.net/papers/extended.html