Thursday, September 28, 2006

When the Bough Breaks...

Yesterday in Philosophical Foundations of Cognitive Science we talked some about Andy Clark's idea that the mind extends beyond the individual and into his surroundings. It goes something like this: we are accustomed to using tools to complete tasks - even mental tasks. When we have some complicated math to do, we tend to reach for pencil and paper. Clark argues that the pencil and paper are part and parcel to the "mind," part of the machinery that gets the calculation done. Thus, the distinction between the individual and his surroundings is blurred.

I don't necessarily disagree with this. I have a lot of sympathy for the Functionalist notion that "mind" isn't limited to instantiation in neurons. So in a meaningful sense I'm committed to the idea that minds can be made out of pencil and paper too. That said, I'm not an uncritical fan of Clark's argument here, and it's worth going through why not.

First, I dislike his use of the words "scaffolding" and "props" to steer readers away from other words with less convenient associations, like "tools." There can be little doubt that pencil and pad are tools. They are external objects used to aid in a specific task for which they are suited and then discarded. By any meaningful definition of "tool," they qualify. However, this word has other associations he doesn't like. For example, a "tool" implies a user distinct from the tool. Whereas we think of "props" as being part of the whole of the play, and whereas scaffolding blends nicely into the construction it aids, tools are generally under the control of an agent, definitely not "part" of the whole. Scaffolding and props are useful simply by being present; a tool must be actively used. And the fact of its being "used" implies a hierarchical relationship between the agent and the tool.

Now, Clark would no doubt respond that he's just being precise. Sure, pen and paper are tools, granted - but they're a tool of a different kind because of the relationship they bear to the nominal "user." His choice of words is meant to draw attention to this difference. And that's fair enough as far as it goes. It's just that I am not really convinced that they are "tools of a different kind." To illustrate, let's consider a screwdriver, a prototypical "tool" if ever there was one. A scredriver is meant to aid a person in turning a screw. And, in fact, in many (most?) cases it it indispensible to the task. A person can turn a screw without a screwdriver, of course, but usually not with a very high degree of success. And yet, it would strike us as unusual to think of a screwdriver as an extension of an arm. But why? Surely there is no bigger leap being made here than saying a pencil and paper are part of the mind? More to the point, the screwdriver is used in exactly the same sense that a pad and pencil are. It aids the user in a specific task to which it is suited and is then discarded. The relevant features seem to be the same. We seem to need to stretch the words "prop" and "scaffolding" a bit to get the desired effect, but "tool" simply applies. Now, of course analogies are useful for pumping people's intuitions, so there's nothing necessarily inappropriate about comandeering these words to get a point across. But one gets the feeling that the argument rests on the words rather than simply using them to challenge conventional thinking and make a point.

Which leads to the second objection - namely that I think there's a sense in which Clark isn't appreciating what his analogy is really useful for - much the same way that I believe Searle doesn't fully understand the implications of his own Chinese Room thought experiment.

Why can Clark "get away" with calling a pencil and paper "scaffolding" when we wouldn't let him do the same with a screwdriver? Well, he wouldn't be able to "get away with it" for Searle. Searle would simply reply that a pencil is not made of neurons, ergo it isn't any part of a mind - it's a tool, end of story. Which is interesting because it means that the only people who are susceptible to Clark's analogy are people already sympathetic to Functionalism. You have to already be committed to the idea of mind as substance-independent system to think this analogy is worth talking about at all. (To be fair, it's possible for people in Clark's camp to buy that minds have to be made primarily out of neurons - i.e. I'm stretching the rules a bit by saying that only those sympathetic with Functionalism will bite. In fact, someone could believe that mind extends beyond the brain, but that the brain part must necessarily be made of neurons, that there are appropriate building materials to appropriate modules, etc. A mind is a neuronal processor plus some extensions, or whatever. I personally think this opinion would be difficult to maintain (because writing instruments are clearly the sorts of things that can be instantiated in many materials - stone tablets, keyboards and screes, quill and parchment, etc.), but it wouldn't be impossible to do so.) And yet, there's something in us that rebels against it all the same.

What I'm getting at is this: Clark's analogy is useful more for pumping the intuitions of Functionalists (and similar) about their beliefs than it is for actually introducing us to a novel way of thinking about the world.

As to that last bit, I really don't think that Clark's "observation" that minds extend beyond the individual and into the environment is all that helpful. Sure, it gives perspective, but there's something artificial about it. Namely, we can connect anything in the universe to anything else in some finite number of analogy applications. So right, granted, the boundaries that we draw are ultimately arbitrary in some sense. A screwdriver can be thought of as an extension of an arm if we push it, a pad can be thought of as a temporary memory extension, etc. But take this too far and you lose the concept you started with.

As an example, consider the distinctions that Environmentalists like to draw between the "natural" and the "artificial." Although beavers build dams and this is "natural," when a human builds a glass house it's "artificial." Why? Well, there is no clear answer, and these boundaries will be different for different people. Pretty much everyone agrees that tools made of sticks and stones are "natural." Once we add metal, though, intuitions start to diverge. For some there's no problem, as long as there's no large-scale use of it. Swords are fine, highrises not so much. For others, highrises are even fine: that's just the product of an animal instinct for shelter, etc. For still others, even cars and space shuttles might be "natural" in the sense that it is in human nature to make and use tools, etc. But once we've reached this point we get the feeling that the distinction has ceased to exist. If even space shuttles are "natural," then it seems people are willing to accept anything made of atoms according to the laws of physics as "natural." Which covers everything, and the concept ceases to make any distinctions.

There's an element of that going on in Clark's analogy, I think. If we accept anything that makes an impression on a mind as a part of that mind, we start to run out of use for the concept "outside world." To illustrate this intuition a bit - let's say that Clark had chosen a calculator instead of a pad and pencil. Well, one gets the feeling this wouldn't work as well. With a calculator, the feeling is more that we've farmed out the thinking to something else. Of course, a calculator is still a tool, but its functioning is largely hidden. We hit the buttons in the prescribed way and an answer comes back, but we don't really know how. There isn't an obvious link between the buttons and the answer - save that it's doing the math we want. Hitting the buttons is more like communicating with the device, which then communicates back through its output screen. And yet, I could easily describe a calculator as either a "tool" or a "prop." (If I stretch it I might even get "scaffolding" in the sense that it helps me get access to the parts of my mind I need fast access to, or something.) Well, this gets even worse if you're like me and your computer is always on. When I need some quick calculating done, I usually just flick on the screen and open GNUPlot or bc. It's even harder to stretch the analogy here: computers are getting ever closer to being "other minds." And of course we reach the fixed point when we start to think of other people as extensions of our minds. If I need to know a fact about geology, I could look it up, but if there's a geologist in the room with me I'll probably just as him instead. Is he part of my mind? We are all together coo-coo-ca-choo?

What I'm illustrating, of course, is that like the "natural/artificial" distinction, the boundary between "my mind" and "outside world" can be played with to an arbitrary degree - the only reason we don't is because if we take it too far we obliterate the distinction. The world has sort of settled on a nice intuitive definition of the boundary: what I perceive as "internal" is - and mind is all internal. Pen and paper are tools because they are outside me. They aid my mind rather than being part of it. There is, as I said, no principled reason (for a Functionalist, anyway) not to adopt Clark's concept of extended mind. The boundary is arbitrary. But because it's arbitrary we have to be careful not to take things too far - in order to preserve the concept (provided we still think it's useful - which we do at present). Things which require close interaction between mind and object (like an abacus or pencil and paper) do better, of course, than things like calculators which simply return an answer without much interaction. But coming up with a principled distinction seems hard, and if Clark can stretch "mind" to pencil and paper, it's only a matter of time before someone else wants to add in the calculator, the computer, or even (gasp!) the other person.

Since there's no particular reason in principle to object to Clark's monkeying with the common concept of mind, it seems fair to ask whether there's also a reason he's doing it? Monkeying with the concept, after all, sort of disturbs other dependent mental concepts. It creates side effects. It's like running the installer to upgrade a piece of software only to be told that the new version is no longer compatible with some other dependencies. So you can't just grab Qt4, you've got to grab a bunch of other seemingly unrelated stuff too - like a new version of X11 or whatever. What you thought would be a 2min. installation ends up taking the better part of 20min.

And that's definitely going on here. Changing our concept of mind to extend beyond the individual body makes us nervous because it means we might have to "update" some of our other concepts as well: like the concept of individuality, or the internal/external distinction, or perhaps consciousness, etc. The default position would seem to be to leave well alone!

So the relevant question is whether Clark's idea buys us anything?

Not that I can think of. In class there was some talk about some guy Hutchins who consults for airplane manufacturers who believes that it's more useful to think of the cockpit - pilot and all! - as a single system. One assumption made in class was that if he can get manufacturers to pay him for this, it must indeed be a useful concept. But of course that doesn't follow at all. It might just be a catchy sales point; he may, in fact, have nothing particularly special to offer that another consultant couldn't provide. In fact, it's completely counterintuitive to think of a cockpit as a single functioning unit insofar as there are obvious and apparently relevant differences between the machinery and the pilots (pilots don't "break down" or have backup systems, they don't behave in a deterministic fashion, they have intuitions and stored memories of flying history which the instruments don't, etc.). Thinking of pilots as cogs in a system seems like just a way of talking. It's difficult to imagine what this actually buys you in terms of insights.

And so it is with Andy Clark's analogy too, I think. It's not that we can't stretch our concept of mind to acconomdate it, it's just that there doesn't seem to be any particular motivation to do so. If we grant (which any Functionalist surely must) that there is no easy physical determinate of the boundary of mind, that this inner/outer distinction is in some sense arbitrary, then shouldn't we just go with the general intuition - leave the language untouched? People across many cultures have managed to agree on this boundary. Further, they all have words to accomodate the concept Clark is pushing: namely "tool." They are all capable metaphorically of speaking of tools as extensions. So there doesn't seem to be any real point or insight to what amounts to advocating changing the definition of "mind." The world's languages are already rich enough to deal with any situation that would seem to require us to bend the default concept. Why, therefore, do it in general?

I can't think of a reason. But I can think that it makes me nervous to override a general and apparently stable intuition (everyone seems to "just know" what is internal to them and what's not) just to make a minor cute point.

So to get back to the main idea: the real benefit to reading Clark seems to be for its use as a nice thought experiment. It's sort of a point in Ned Block's column. He's immune to it, of course, since minds are neurons for him. But the rest of us are compelled to answer it. We have to admit that there's a kind of "leak" here - that people can push us to accept fairly counterintuitive boundaries for minds. And in that sense, Clark has done something very useful. But is it an insight? I'm not convinced.


At 6:10 PM, Blogger noahpoah said...

Perhaps perversely, Clark's leaky-mind metaphor can function as an insight for precisely those folks that you think won't buy it at all.

If you're committed to the reification of neurons-as-mind, the fact that people regularly depend on tools like pencils and paper, calculators, and fancy math programs to do long division and multiplication may well tell you something about what kinds of computations neuron-minds are (not so) good at.

It's interesting to me that you're insistent on keeping the mind/not-mind boundary as is, but that you're also sympathetic to functionalism. Aside from tradition or neuron-love, neither of which are particularly convincing, what's the reason for insisting that the mind is body-internal?

Let me be clear that I'm sympathetic to regarding all tools as extensions of our minds and/or bodies. I don't think it's all that problematic to keep tools distinct from other non-mind-extending, non-body physical objects, and it seems to me that, within the category of tools, there are surely degrees of mind-extension.

You mention a couple of examples - calculators and fancy programs as opposed to pencils and paper - but you fail to make the same distinction with regard to screwdrivers. Yes, screwdrivers and pencils/paper are tools, but the uses to which they are put bear very different relationships with our cognitive systems.

If you use pencil and paper to carry out a math problem, it seems pretty clear to me that you're offloading some of the memory functions your mind would have to keep track of otherwise. If you use a screwdriver to attach two pieces of wood to one another, you're adding functionality to your arm for sure, but not so much to your mind.

This all seems related to the sometimes tiresome debates about 'level of analysis' in cog-sci. I'm not so skeptical of the pilot-and-cockpit-as-system idea, as I can easily imagine a 'level of analysis' at which this is precisely the right way to look at it. Treating the pilot and cockpit as part of a plane flying system can help focus attention on the interface between the pilot and the controls, which can, perhaps, in turn enable changes to be made to how the cockpit is designed in order to make this interface as seamless as possible.

I would argue that, to the degree that cockpits are designed to conform to the way humans are built, cockpit designers are already thinking of pilots and cockpits as unitary systems.

Clearly, the basic idea of a leaky mind is not going to (directly) give us any testable hypotheses, but that doesn't mean it's useless. Maybe being 'a nice thought experiment' is all it needs to be.

At 8:08 AM, Blogger Joshua said...

You mention a couple of examples - calculators and fancy programs as opposed to pencils and paper - but you fail to make the same distinction with regard to screwdrivers. Yes, screwdrivers and pencils/paper are tools, but the uses to which they are put bear very different relationships with our cognitive systems.

The point of bringing up the screwdriver was not to deny that there are levels in the relationships tools bear to our cognitive systems. The screwdriver was just meant to illustrate that few people can be convinced it is an "extension" to their arm in the way that people can be convinced that a pad and pencil are an extension to the mind. What you're countering with here is implicit in that point. The overall point is simply that the only reason people can be convinced that mind can extend to paper and pencil is that they've already bought in to a concept of mind that is so extensible. This is revisited later in the point that the fact that this ability to metaphoricaly extend the mind to paper and pencil is, in fact, all that is required. We don't need to update our concept of mind in the way that Clark suggests becuase the concept is already recognized as flexible by the appropriate people (in a way that the concept of "arm" is not).

The short answer, in other words, is that the screwdriver is not meant to be a counterexample to the idea of extending the cognitive system through tools.

Treating the pilot and cockpit as part of a plane flying system can help focus attention on the interface between the pilot and the controls, which can, perhaps, in turn enable changes to be made to how the cockpit is designed in order to make this interface as seamless as possible.

I see it in just the opposite way. Designing an interface crucially depends on recognizing the two things being interfaced as different things. Part of the same system, sure, but an interface design will rather tend to concentrate on what makes them different - because its purpose, after all, is to smoothe over those differences.

Clearly, the basic idea of a leaky mind is not going to (directly) give us any testable hypotheses, but that doesn't mean it's useless. Maybe being 'a nice thought experiment' is all it needs to be.

Nothing in the post implies that the thought experiment is useless or that it needs to give us testable predictions. All that was said is that it is not useful in the way it was intended to be used. It seems to want to argue for a new concept of mind, and all I'm saying is that the old one is just fine, thanks, already sufficiently flexible for our purposes. But as an intuition pump it's very useful - and that, as you point out, is all that is required for a successful thought experiment. I'm arguing against the presentation and the specific conclusions, not the validity of the model.

At 11:02 AM, Blogger noahpoah said...

I think (once again) that we agree on the substance here, but that we're 'arguing' about some semantic nit-pickery that, really, your post is at pains to point out the silliness of. I agree, if I'm remembering Clark's position accurately (it's been a while since I read his stuff), that the metaphor won't do the work he intends it to do.


Post a Comment

<< Home