Clark

Repeated rejection of the notion of representation is a kind of cornerstone for cognitive science, and there has been puzzled by the depiction of the world itself as perceiver dependent. So since that last bit, thinking of the world itself as perceiver dependent, was the one bit that I don’t seem to have quite come around to so far.

I thought that I’d explore a bit there and see how close I could get. So all these themes are prominent in that kind of multi-orthodox tour before the embedded mind, and also in the other solar work, so here’s a little passage from the re-enchantment of the concrete. In that passage, the bits of matter are the idea that perception doesn’t consist in the recovery of a predetermined world, but rather in the guidance of action in the world inseparable from our sensory motor capacities. And then at the end we can summarize by saying that cognition consists not in representation, but in embodied action.

That’s kind of alright by me nowadays. Corellatively, the rare we know is not pre-established. It’s rather inactive through our history of structural coupling. And it’s that last bit that I continue to find kind of puzzling, just what does that mean? And I think I’m coming closer and closer to at least finding a way of understanding it that works for me, so that’s what I’m going to go through. So first, the simplest explanation of this, it seemed to me, was that the cash value of the notion of a perceiver dependent rare had to be something pretty weak. It had to amount to, let’s say, the innocent idea, oh sorry, that was my question. It had to amount to the very innocent idea that a world may rather weakly be said to be perceiver dependent just because a perceiver selectively represents some things and met others. That’s a rather weak sense.

You might think here of, for example, that classic work of Von Uccskol’s where he talks about the invents of animals and men, the example of the tick, the tick whose world on wealth comprises only a few simple cues. It is supposed to comprise just a few stimuli. For example, butyric acid found in the skin of mammals. Detection of butyric acid caused the tick to drop off the foliage, hopefully onto the skin of a nice warm-blooded animal. Detection of heat then initiates burrowing in the tick. So it’s supposed to be sensitive too.

As Von Uccskol put it, nasty picture here of a tick doing something nasty to you, out of the vast rare which surrounds the tick, three stimuli shine forth from the dark like beacons and serve as guides to lead her unairingly to her goal. But clearly, however, there’s nothing there that the realist needs to worry about. The story there really is just one of selective representing. The tick picks out a few items if you like from the great list available to physics, but with particular interest to ticks. So if we want to go beyond the notion of selective representing, we might think for example about weak co-constitution.

Maybe the kinds of environmental property and feature that matter to different animals aren’t just sort of different with a set of ticks, no pun intended, put on a great list of physics, but instead there’s something kind of gerrymandered, something rather odd looking, something that really really is something specific about the kind of body in particular that a creature has. So for example, teacups. I’m an Englishman and you know if I live in a world populated largely by teacups, oh yeah, good point.

Yeah, if I live in a world populated largely by teacups, I don’t know how to do this. Fine, here. How about that? Is that all right?

Okay, well I’m going to sit back and relax now. If I live in a world populated largely by teacups, this is something about human motor structures. Reflects a kind of item that a creature like me can easily grab and lift. And it says something about my own historically and culturally determined habits and customs. Now that seems a little more promising. But most promising of all it now seems to me is to read these comments about perceiver dependence as part of a sustained attempt to take very seriously the widely recognized but I still think underestimated fact that organisms in general and we humans to an amazing degree are the active constructors of special environments that in turn mold and construct us.

So what I want to do is try and think about perceiver dependent worlds in terms of the kind of ratcheted iterated effects of niche construction. Okay. So this kind of consideration isn’t part of the traditional metaphysical debate. I think what happened to me is, you know, I’m a philosopher, I read these as claims about metaphysics basically. But it seems to me that it really isn’t part of a traditional metaphysical debate at all. Indeed, for Ella himself says that what he wants to do is to sidestep the whole realism non realism debate by bouncing a different model of the agents relation to her world. Okay.

I took a bit out to save time. So this going is explicitly stated in passages, for example, this one from the embodied mind. They say that multi authors there say we’d like to negotiate a middle part between the silver of cognition as a recovery of a pre given outer world and the suretice of cognition as a projection of a pre given inner world. Most importantly for my purposes, they say our intention is to bypass entirely this logical geography of inner versus outer by studying cognition, not as recovery or projection, but as embodied action. And the rest of this talk, I want to just pursue that middle way a little bit by exploring some specific ways in which a kind of combination of active niche construction and active boundary construction. So the organism in acts its own boundaries, some boundaries with the world that it’s inhabiting might indeed also the logical geography of the inner versus outer.

So the plan is this, I start off with that little sort of puzzle. So what can perceive a dependence means so that a kind of more or less unreconstructed realist like me can take it on board, something like that. And I’m going to make a few suggestions about that one is that we enact our own boundaries. Secondly, we construct our own niche. And lastly, that when you put those two things together, you get a very powerful sort of ratchet effect. So I’m going to start with this enacting the boundaries. I think constructing the niche is more is a more common thing to think about nowadays, at least thanks in part to the work of people like Susan O’Yama who’s with us here. So I’m going to start with a colorful example, but probably ought to strike you as trivial.

I think it will lead to maybe one or two things that aren’t so trivial. It’s somewhat by a performance artist, an Australian performance artist called Stellark. Stellark is a performance artist devoted to trying to explore the difference between inner and outer, between biological and non-biological, and between self and other. And here’s Stellark performing using a faired hand. So for a lot of Stellark’s performances, he uses this faired hand, not so much nowadays, about five years ago you were seeing a lot of this. It’s an electronic device mounted on the biological right arm, but independently controlled by means of canally placed electrodes.

What happens with these is that EMG signals detected by electrodes placed basically on the legs and on the abdomen, what happens is the brain sends commands to those that are then relayed to the prosthesis. Since those sites are not normally used for hand control, Stellark’s actually become very, very fluent at using the faired hand. You can see them in the picture here, right in the word evolution with three hands simultaneously.

Wow, I can’t even do it with two, and I have been trying, I promise. The point is, it’s Stellark’s brain, if you like, that has to give those commands to the abdomen and thigh muscles and so on. But as far as Stellark, the conscious agent is concerned, that’s not what he’s doing. What he’s doing is controlling the third hand.

Phenomenologically, Stellark’s experience is absolutely that of simply telling the third hand what to do. And in this respect, I think it’s just like you doing the washing up. When you do the washing up, you don’t feel, for example, as if you’re telling your hand to do the washing up.

You just feel as if you’re washing up. Your hand functions as what some philosophers have called transparent equipment. Equipment through which you, the agent, can act on the world without first willing and act on anything else. So, Heidegger, I guess, would often be cited in this context and a bunch of people since then. So, here’s the first familiar, but I think important sense, in which intelligent agents can enact their own bounds and bodies. The boundaries of the phenomenologically experienced body are negotiable.

They can be reshaped in the kind of heated furnace of fluent action. Experientially, at least, the body-world boundary seems to be a science and technology developer. For example, as we’re now in the direct neural control of robot arms, of wheelchairs, of on-screen cursors, and so on, I think this negotiability of our own bodily boundaries will become increasingly important. And even without high technology, skilled use of sports and musical equipment already provides a kind of intermediate case.

At a quite basic level, however, this shouldn’t really come as a surprise. For biological beings, whose natural bodies grow and alter. Basically, you come into the world as a baby with a kind of body that you or your brain have to learn to control in just the sort of way, a spell-up, that’s often more of a developmental psychologist talking about body-babbling. See the little infant lying in the car that sees and wants to make move the mobile, doesn’t really know how to do it. Has to try out various neural commands in effect to see what’s going to get this piece of equipment, my hand, to do the job.

When eventually, or what’s going to do the job, the men probably find out which bit of equipment’s doing it, actually. So there’s a sense which a human baby encounters her own body in much the same way Stelarc first encounters the third hand. More surprising then when bodily boundary in action, I think, maybe the extent to which our sensory contact with the world is equally open to activity-based extension and transformation. So this was a topic that caught the Erez air in attention.

In passages where he described the work of someone who at one point I thought was going to be the session chair but something altered, Paul Barquerita. This is where a lot of you or may have worked from a long time ago, look at that picture, very 1970s picture. This is a blind subject fitted with a head-mounted camera and one of these grids that you can wear on the back, a tactile visual substitution grid. Those of you that haven’t seen this thing before, it’s like those things that you can get in sort of toy shots to make imprints of your face like that.

Imagine a grid rather like that with the regions of the grid controlled by the input from the head-mounted camera. So wearing a device like that while moving and acting in the world, what happens is at first you feel a tingle on your back. But after a while, this is very well known, patients report that they stop feeling the tickling on the back and start to report rough quasi-visual experiences of looming objects and so on. So a bull thrown at the head causes instinctive and appropriate ducking.

The causal change is a bit deviant here. It runs via systematic input to that back-mounted grid. But in some sense, the nature of the information carried or perhaps following Albers talk, the way that input can participate in the control of action is suggestive of the visual modality and something we could explore maybe is what makes something fall into the visual modality rather than another one given a very action-oriented perspective. Now, there is in place of that clunky old back-throttle thing, you have things like this, little tongue-mounted grids that work very well because the tongue survivors are really electrolyte, but doing the same kind of job. Performance using these devices is quite impressive. Here’s Barkey Rita in 2001 says, these systems have been sufficient to perform complex perception and eye-hand coordination, camera hand coordination tasks.

These have included face recognition, accurate judgment of the speed and direction of a rolling ball with over 95% accuracy in back in the ball as it rolls over a table age, as well as complex inspection assembly skills. So there’s review of this stuff in recent trends in cognitive sciences. Now, just as an inactive model would predict, the absolute key to effective sensory substitution is goal-driven motor engagement. It’s absolutely crucial for these effects that the head-mounted camera or whatever other sensory substitution mode you’re using be under the subject’s intentional motor control. This allows for a particular kind of active learning via the motor system as action commands then systematically affect the perceptual input, the stop that’s coming back from the camera. For example, as you hear someone approaching from the left, you turn your head that way. At that moment, your brain is in a position to begin to learn something about certain sensory motor contingencies.

Importantly, it’s the agent’s ability to control the actual sensor, in this case the camera, that determines where the input becomes experienced. So in recent work with neprosy patients using data gloves, this is particularly striking. So you have patients here who have lost the feeling in their hands a long, long time ago. They’re fitted with a sensor made in glove and they have a sort of disc put on the forehead that does the sort of tactile stimulation.

What happens rapidly is that they stop reporting feeling anything on the forehead and start reporting feeling sensations in their hands because after all, the sensation is here being projected to, as it were, the site at which motor control can systematically vary the perceptual input. I think that’s an interesting case. Okay. So what all that shows, I think, is that as far as our sensory contact with the world is concerned, brand new agent world circuits become calibrated in the context of whole agents engaged in goal driven activity. And that for this to occur, all we really need is closed loop signaling, so that motor commands can systematically affect sensory input.

If these conditions are met, the experiential impact is quite striking. The details of the circuitry fall silent in use, and you get that sense of, as it were, transparent equipment. The conscious agent is aware of the oncoming ball, a lot of C in the ball, more of using a tactile visual sensory substitution channel to see the ball. What they’re aware of is ball coming towards me. So here’s the second sense there, and again, a fairly trivial and straightforward one, but an important one, I think, in which the embodied agent actively constructs her own bounds. Sensing rather than being seen as a kind of fixed transduction point at which the embodied agent encounters a wider world, can itself be seen as in part and within minutes that I think we still need to determine. A motor generated achievement, open to radical reshaping by the very activity, that’s to say intentional engaged action, that it supports. So the picture that’s now emerging here is the image of what I’m later going to be calling a profoundly embodied agent. These agents are integrated, but constantly negotiable platforms of sensing, acting, and I don’t want to argue reasoning. And why you should call that profoundly embodied rather than disembodied is something that I might get to if I get a chance.

Before pursuing this any further though, I think it helps to directly address a rather reasonable worry, a kind of conservative reinterpretation of what I’ve been saying so far that might go like this. Why think about all this as the activity-based inaction of body and sensing, rather than as something like a stovable cause of body and sensing that they just get influent, they just come into be an influent command of a few new tools and gadgets. You can think of this as a question, what makes it incorporation into the embodied active agent rather than just use? That’s an issue that’s become particularly salient for me in the case of reasoning that we’ll get to in a minute.

So let’s just stay with the easier cases for now. To resolve this kind of worry, I think, you can’t just appeal to what it feels like. Instead, I think it’s useful to add a little bit of neuroscience. So there’s some suggestive data I think in recent work on the effects of tool use on bimodal neurons. Here’s what bimodal neurons are, long quote here from Maravita and Iriki, but basically what matters about the bimodal neurons is that they’re neurons that have a sort of joint response. They respond both to somatosensory stimuli, like a gentle touch on the hand, and to objects falling within a certain visual space around the hand.

So they’re neurons that have, that are attuned to both of those parameters, if you like. So recordings were taken from bimodal neurons in the intra-parietal cortex of Japanese macaques when they learnt to reach for food, an intentional act, reaching for food using a rake. What the experimenters found was that after five minutes of rake use, the responses of some of those bimodal neurons alter. Basically the bimodal neurons that used to respond to touch here and this bit of visual space now responded to touch here and this bit of visual space. And another bunch of bimodal neurons had the same sort of effect, but there’s a picture there you can sort of get the idea. Basically the rake is now being treated by those neurons in the same way as the space around the hand was being treated before.

After surveying a number of related studies, Neoravita and Iriki concluded that these visual receptive field expansions may constitute the neural substrat of use dependent assimilation of the tool into the body schema, as suggested by classical neurology. The body schema, with an odd tishongalica who’s here somewhere, shouldn’t be confused with the body image. The body image is a conscious construct able to inform thought and reasoning about the body.

The body schema is something more like a suite of neural settings that implicitly and non-consciously define a body in terms of its capabilities for action, for example by defining the extent of directly reachable space for action programs. I noticed too that in line with our earlier comments, any of these neural changes only follows active intentional use of the tool, not just putting it in the monkey’s hand or whatever, like you have to be using it to do some raking. Okay, rather decalibration of the body schema is also evident in recent experiments involving tool use by patients with unilateral visual neglect. So this is kind of a sort of oddly pathological version of the same effect. These are patients who selectively fail to see visual stimuli occur in some region of egocentrically coded visual space, rather surprisingly I think, in a patient who’s neglects selectively affected space close to one side of the body, use of the stick extends the neglect to the whole area, now within stick reachable space.

So it’s a kind of, it’s the same effect. 30 and the fascinating conclusion perfectly in line with Maravita and Iriki was that the brain makes a distinction between far space, the space beyond reach and distance, and near space, the space within reach and distance, and that simply holding a stick causes a remap in a far space to near space. In effect the brain, at least for some purposes, now treats the stick as out with half of the body. Changes in the response characteristics of certain populations of bimodal neurons may go some way then towards explaining some of the mechanisms by which some equipment actually falls experientially transparent in use.

To clarify that result may be a little thought experiment. We can imagine I think tool users whose brains weren’t like that. Their brains aren’t engineered so as to be plastic like that, so as to adapt the body schema in these ways. These are creatures who must always use tools the way we typically use them at first, by representing the tool in its features, in its powers, its length and so on, and then kind of calculating what you might do with it. I think that they’re on a philosopher and we’re good at imagining, but I think I can even imagine creatures who are so good at this that they became as fluent in using their tools as we are in using ours. But they’re actually using a very different route to do in it. Their route is a route of representing infer.

Our route seems to be the route of also a suite of neural seconds that implicitly define the space around the body in terms of what it can do. So the end issue is concerning the efficiency of use then. I think there’s this deeper issue, this contrast between an agent that first represents the shape, the dimensions and the powers of the tool, or body part, and then impairs what she can do with it. And agents were constituted so that some of those inferences don’t need to be made.

Some of those results now fall out automatically. Okay, what I think that stuff does is it helps us give some sense of the biological agent as being at home in her world. The body presence like this isn’t represented as a resource. So much is inhabited as a nexus of possibilities for action. And there I say the body, of course, I mean equally the body as augmented by tools and so on. Now I’m tempted to see this as part and parcel of the broader motion developed in the embedded mind of structural coupling.

So there’s passages like this. Cognition in its most encompassing sense consists in the enactment or bringing forth of the world by a viable history of structural coupling. One way to connect this to the idea of the perceiver dependent bringing forth the world is then simply to reflect that now the body world division is itself being created in part by history of activity and engagement.

By processes of ongoing self-calibration in which bodily boundaries and sensory systems are inactive rather than simply given. And as I say, I think we’re going to in the very near technological future that will actually have more points than it does right now. What most goes to the division between the body and the world goes doubly, I want to argue. The division between mind and world.

Or in fact it probably goes to just the same reasons because there’s no good division between mind and body anyway. But let’s go through a kind of reasoning like case. Here we come close to territory that others, myself included, have lately been exploring.

So since I’ve been doing this next bit at great length elsewhere I’m going to be really unhelpfully brief today. But the idea, the question there would be could anything like the notion of incorporation rather than mere use actually get a grip on the domain of mind and cognition? Could human minds be extended and augmented by cultural and technological tweaks? Or is it still going to be just the same old mind with a shining new tool? And this lampra notion I think is very much the one you get out of strong evolutionary psychology. Same old mind’s shining new tools.

The story that I want to develop is one where that is just not true. The minds themselves are changing. They’re not just using new tools. So rather abstractly the story I’ve been telling is one in which external and non-biological information processing resources are also at the temporary or long-term recruitment and incorporation rather than simply for knowledge-based use. And to the extent that that’s true then we’re not just bodily and sensorally but also cognitively permeable agents. One specific way in which this manifests, a way I think that’s beautifully explored by Alvin Noghian and others is the idea of developing problem-solving routines that make feet with implicit commitments to the robust availability of certain operations and or bodies of information while you’re doing some problem-solving. So while Alvin was talking about there, my visual routines make robust and make implicit commitments to the robust availability of more detailed information by Saccardin around the scene. There’s a partial parallel here with at least one feature of the physical tool use case. Also I have this slide here.

Okay, let me backtrack. The story that I want to tell you is all about delicate temporal tuning of multiple participating elements. So you have this notion of trading easy motor-based access to information against maybe expensive internal local biological storage. And there’s a sense in which I think it really doesn’t matter which of those solutions you use and which strategy you use is going to be very context dependent.

There’s some nice work coming out. I’ll just come out in the journal Cognitive Science where Gray and Foo try to quantify this over a course of several experiments. Okay, so the thought is that the parallel with the tool use case is that perhaps then the brain doesn’t need to actively and repeatedly represent the worldly availability of specific information.

It just factors it in by growing its problem-solving routines around that availability. That’s that same kind of contrast again, I think. Now that probably isn’t quite right of... I mean, the conjecture I’m trying to make is this. When those sorts of episodes of representation of the resource, whether it’s a physical one or an informational one, don’t intervene when you visit where growing a way of solving the problem around their robust availability instead of representing that availability.

Then the resources, whether they’re informational or physical, should actually be counted as part of the system, either the cognitive or the physical system. I don’t think that conjecture is quite right the way it is, but it’s probably good enough to be going on with them now. So with that kind of half in mind, I want to add that second ingredient, and I’ve got what, 10 minutes? How long have I got?

A little bit longer? How many? Five. Five, didn’t... I thought you were a bit... I thought I’d start around six? Ten.

Okay, well, let me try and add the second ingredient, and then I’ll be really, really brutal with any third or fourth or fifth or sixth ingredients. So, in fact, what I’m going to do, I think, is go and ask this question. Is there any interest in sense there in which the world itself, and not just the agent world boundary, might be said to be somehow brought forth or inactive, rather than simply biologically given? Certainly, Valella and company often speak in this way. They say the overall concern of an inactive approach is to explain how action can be perceptually guided in a perceiver-dependent world. And as I say, I used to treat this as a rather dubious move in the debate between the realist and the anti-realist.

And I do kind of think that there is a metaphysically extravagant and dubious move available in the neversimity of claims like that one, but I don’t really think that we have to hear those claims in that way. Instead, I think, we can think that the world is perceiver-dependent, simply because insofar as it becomes structured in ways that both reflect and enable the characteristic activity patterns and problem-solving routines of the perceiver. That, I think, is probably the cognitive equivalent of that notion of laying down a path in walking. So, you know, there are lots of really simple cases here. The basic idea is that we build our own cognitive niches. Lots of people have done work on this kind of stuff.

Here’s a quick example. Some of you will know this one from either bartender, the expert bartender. You might think that the expert bartender has developed exquisite memory routines, biological memory routines for drinks orders. Beach in 1988 compared expert and novice bartenders under the condition of normal glassware and a special set of uniform glassware where there is no variation available. What they found was that in the regular condition, the experts were great, but in the uniform glassware condition where they were all the same, all of the experts of bar and pitch vanished.

The explanation is obvious enough. The expert bartenders haven’t really honed their biological memory skills in quite the way you might have thought. They don’t really remember biologically all that stuff. Instead, they associate each order with a special kind of glass cocktail furniture combination and lay the glasses out in order.

Notice that courtesy of this strategy, there’s no need for the bartender to explicitly biologically store the contents of what to drink or the temporal sequence of what to drink next. Instead of arrangement in the world, there’s those parts of the load. This is a world as external memory thing. Language itself I wanted to go on to claim actually behaves for us in the same sort of way. But where do language themselves maybe form a layer of humanly imposed structure whose impact on human cognition can be thought about in some of these sorts of ways? I’m going to do one example I hope in three minutes and then try and sum up.

I’m suffering a bit here from the five things. Okay, here’s a quick example. Here’s my favorite one. Two groups of chimps are taught to discriminate stainless and difference using presented pairs.

Two cups or a cup and a shoe. That sort of thing. And then one of the groups of chimps is while they’re doing that task also trained to associate the judgments with a plastic token. So when they make a stainless judgment, they associate that with say a red plastic square. When they make a difference judgment, they associate it say with a blue plastic triangle. In the second experiment then, with all of these little plastic tokens now taken away from both sets of chimps, the token users and the non-token users.

They try to get the chimps to do something much harder to respond to higher order sameness. Sameness or difference of exhibited relations. So the top set there exhibits the relation higher order sameness. The relation of sameness is present in each case. The bottom pair, the bottom pair of pairs exhibits a higher order relation lack of higher order sameness because the relation is different in each case.

Sameness in one difference in the other. You might ask then, why is it that only the token trained chimps can learn about the higher order regularity? Well plausibly, because experience with the labels is now reduced in the higher level pattern to the lower order one. So just to think about it almost naively, imagine that cup-cup recalls the image of a red square.

Cup-shoe recalls the image of a blue triangle. But once you’ve done that, the higher order task is reduced to the lower order one. So maybe there’s a sense here in which real symbols help us dilate and compress patterns in sensory space. In that sense, language itself is a kind of animal-built structure that reflects, of course reflects perhaps, but then structures and transforms our own thinking and problem solving. So here maybe is another place to breach those traditional boundaries. Our cognitive relation to our own words and language defies the simple logic of universe without a word and labels or first encounters as objects, as more structure in our world, but then they effectively reconfigure the problem space for biological cognition, changing the kinds of thinkers that we are. Okay then, how can I get to somewhere where I can stop?

Let me think. Okay, this is a kind of fun slide because sometimes I think we don’t really appreciate on what a cosmic scale we do this kind of stuff, but just think about this one. We quite deliberately as human beings build better tools to think with and we use those tools to discover still better tools to think with. And we tune the way we use those tools by using some of the tools to educate ourselves to use our best tools better.

And we tune the way we tune the way we use them by using some tools and practices to help build better tools and practices to educating ourselves in the use of our best tools and practices. I think when you get into that kind of cycle of constructing design and issues that go on to construct you and when you add to that the notion that the boundary between the agent and the world, both in physical bodily terms, in sensory terms and in cognitive terms, is constantly enacted by the way that we engage with our world. You get just this incredible potent set of ratchet effects in which really the world that we think about and perceive is very, very much an aspect of ourselves and our own activity. So my final slide was in the whole bit on profound embodiment but you got the idea, it didn’t matter. What I think this part gives us some kind of sense of the genuine sense in which ultimately we’re ever world and not just in it and I hope and think that’s something that Brella himself was interested in. So thank you.