The Extended Room, Or, What Otto Didn’t Know

The Extended Room, Or, What Otto Didn’t Know

By Ryan Victor, The University of Pennsylvania

1. The Extended Mind

Clark and Chalmers’ thesis of “the Extended Mind”¹ involves the problem of where to draw the boundaries of the mind. This question is addressed by recourse to a couple of examples. The first involves a Tetris-like set of scenarios. Let us imagine a situation where, in order to ascertain whether a certain shape block will fit within a group of differently shaped blocks (I assume we’ve all played Tetris), we can either mentally imagine it, or use a button (which can do it faster) to rotate it. Clearly, we imagine the first case as a mental rotation, and the second the rotation as external. Now, imagine a case in which the ability to perform the fast rotation is implanted into our visual system, or, if that is too far a stretch, that the rotation can be activated on screen by thinking about it (this is technology currently under development to assist the disabled).² This presents a problem: is this rotation mental or external? To try and solve the problem, Clark & Chalmers propose the ‘parity principle’, which states that, “If, as we confront some task, 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.”³ So we are to view this prima facie external case as, in fact, just as mental a rotation as if we had imagined it.

The second example involves a man, Otto, who has memory loss, and thus must write everything in a notebook in order to be able to function normally. Thus, when asked what he believes about a certain topic (e.g. where is MoMa) he must consult his book before being able to answer the question (53^rd St). Clark et al. claim that this means the notebook is as much a store of his beliefs as the internal part of the brain in which that information might otherwise be located (e.g. hippocampus). Thus, we should treat Otto storing his belief about MoMa in his book and me storing my belief in my brain as exactly the same. The conclusion of these two examples, then, is that any physical system that is recruited as part of a cognitive process is, by virtue of its participation in that process, part of the mind itself. Furthermore, an agent’s beliefs, being as they are part of cognitive processes, can be stored in any physical location, inside or outside the brain. The criteria supplied for judging whether or not something like a notebook would be a candidate for a part of Otto’s mind are that (i) the resource is easily consulted, (ii) it is consulted often, and (iii) the contents are always or for the most part instantly affirmed. The case under which these criteria are satisfied is known as ‘strong coupling’.

In support of this argument, an analogy can be drawn between adding extra memory to a computer and the notebook example. An external hard drive should not be viewed as any less part of a computer than an internal one, so why should we favour our own internal grey matter over an external memory storage device such as a notebook? A bias towards seeing anything external as automatically not part of the mind is not an a priori reason to discount the view. Further, one might suggest that an actual part of the brain could be transplanted outside the body, kept in a vat, and connected via radio contact with the brain. In this case, the external item is performing cognitive operations, and so what, apart from the fact that it is outside the actual skull, separates this from being part of the mind?

2. Objections and Counterexamples to the Extended Mind

To try to drive a wedge between the two examples above, let us consider a further example – a watch. During the day, I may consult my watch multiple times per hour, I will instantly affirm that the time represented on it is correct, and it is incredibly easy to consult – a turn of the wrist is all that is required. Thus, a watch satisfies all three of Clark and Chalmers’ ‘coupling’ criteria. Were I to use my watch to aid in a cognitive process, would we consider the watch to be part of my mind? If an adherent to the externalist view would still claim that it is, then let us push the example further. Imagine a time before wristwatches. People in earlier times consulted the sun in order to figure out the time of day. They could do this multiple times per day, there would be no reason to doubt the position of the sun, and what is easier to do than to simply look up. Thus, consulting the sun for the time seems to satisfy the exact same criteria as using a wristwatch. By the same logic, then, it seems that the sun, too, would be part of our mind. Even a dogmatic believer in externalism might want to reconsider this assertion. It seems as though the criteria can easily lead us too far astray.

Clark et al’s assertion is that our intuitions are deceiving us into using an arbitrary boundary, the skull, in order to define where the mind ends. It is perfectly reasonable to insist that this distinction is not as clear-cut as it seems, and questions of exactly where we draw the line are valid. However, in place of this distinction, they substitute another, more arbitrary set of criteria. The same question of where we draw the line is equally, if not even more, relevant in this case. For example, what constitutes a high degree of reliance – consulting the notebook once an hour, once a day, once a week?

A further and deeper point is that our intuition seems to strongly disallow the notebook, but allow the external brain segment. Why? I propose a different concept of “strong coupling”. I would use the concept to describe the degree to which the actual cognitive process is ingrained and connected to the external device. The reason why the notebook seems so implausible is twofold: Firstly, it requires the beliefs to be translated into natural language, and then into symbols by the person before being recorded, and then the inverse when the information is accessed. This need for translation has a decoupling effect – the integration into the cognitive process seems less and less direct as we add further stages of translation and manipulation. Secondly, a notebook is something that anyone can access and thus change or erase – this is something we would not claim for the information stored in our hippocampus. Something connected in a more direct way and less liable to be altered by external forces seems more plausible as a candidate for a part of the mind.

The watch and especially the sun examples seem implausible because, under the extended mind view, they are being considered as part of the cognitive process in the same way as the cerebral cortex may be considered part of the cognitive process, despite the fact these are two very different things. Checking one’s watch, or the angle subtended by the sun involves consulting aspects of the external world in order to aid in belief and desire formation. For example, knowing that the sun is directly overhead can cause the belief that it is around noon, and thus the desire to begin making lunch. However, the actual cognitive process itself consists in the interpretation of the sun’s angle as a functional isomorphism for a particular hour. The sun itself is not actually performing that interpretive task, and thus it is not part of the mind. The distinction here is between (a) the attributes of the physical world that we, as humans, are capable of interpreting as representing important information needed to form beliefs and desires, versus (b) the actual substantive matter that provides us with the ability to create and manipulate these beliefs and desires in order to produce goal directed action.

To try to provide a counterexample that more closely parallels the general logic of the extended mind argument, let us imagine a new scenario. I wish to try and define exactly what is means to be ‘part of a computer’. Normally, we would say that the computer ends where the boundaries of the machine are, yet this may simply be an arbitrary bias. Surely, if I were experiencing a difficulty with a program in that it took up too much space on the hard drive, I could conceivably print out the information, and then, when necessary, input the code back into the machine when the program required it. This would mean that the computer’s memory is not only located inside the machine, but also on the page. Why should we say, then, that the paper is not part of the computer? After all, it requires the information on the page every single time it runs a specific program. We should not bias this kind of storage space simply because it is external, should we? Now, what if I simply memorised the code? Then, wouldn’t I be part of the computer too? Again, our intuitions scream no, and for good reason. Neither the actual piece of paper nor my mind are coupled, or connected, in the same way as the actual circuits in the computer are. There is an extra decoupling stage, the printout, or the memorisation of the code, which should automatically disqualify those things from being part of the computer. In the same way, this kind of decoupling should automatically disqualify a notebook as a candidate for part of a mind.

Now let’s take a brief detour to a well-known example, to see if it might shed some light on the extended mind hypothesis.

3. The Chinese Room and the Systems Reply

In Mind, Brains and Programs, Searle⁴ creates a thought experiment in an attempt to show that formal symbol manipulation is not a sufficient condition for thought, and thus passing the Turing test is not a guarantee of a computer’s understanding of language.

The thought experiment is very simple. Searle is in a room and is passed cards with Chinese characters on them. He has a look up table in a giant book, which tells him what to write as a response to any given set of characters. He passes his answer back, and the person on the other side, who only sees the replies written on cards, believes he is having a conversation with whomever is inside the room. Since Searle does not speak Chinese, it is absurd to claim that, armed with his rulebook, he suddenly understands the content of the discussion in which he is involved. This is evidenced by the fact that, were the interlocutor to pass through a message telling Searle that the room is on fire, although he might pass a card back which says “thanks for telling me, I’ll leave,” Searle would not actually do so.

This experiment, then, can be thought of as analogous to a computer, programmed with a look up table of English sentences and rules for applying them. In both cases, there is indeed formal symbol manipulation, and in both cases it is certainly possible that a human may well be fooled into believing that they are part of an interesting conversation, but the fact remains that, in both cases, there is no understanding. Searle himself is acting in exactly the same way as the CPU – simply following instructions blindly.

The systems reply is a response to Searle’s thought experiment in which the term ‘understanding’ is applied not to just Searle, but to the system of the room, the book, and Searle as a whole. The counterargument admits that Searle alone does not understand Chinese, but insists that the system as a whole does. As Block puts the analogy, “If the whole system understands Chinese, that should not lead us to expect the CPU to understand Chinese.” Searle’s response is a good one. Imagine that Searle simply internalises the rule- book by memorising the entire thing. Remove the room. Let the entire process be in Searle’s head. Searle still does not understand Chinese.

The major error in the systems reply is that the word ‘understanding’ is being used in an equivocal fashion. There are many systems composed of multiple parts that function as a whole in modern society. For example, the postal system is made up of numerous complex pieces. Whereas we might say that the postman understood where to deliver a letter, would we say, “the Postal System understood that my letter needed to get to Kentucky in two days”? Absolutely not – we do no speak of systems in this way. Understanding is limited, in everyday usage, to, at the largest, the consciousness of an individual.

It might be countered that we do speak in this way occasionally, for example, “the FBI knew the man was headed to Venezuela.” Although we do accept these kinds of sentences, they are a matter of common parlance and not a matter of conceptualising the system as a whole. We could easily pinpoint the individual agents within the FBI who knew the fugitive’s whereabouts, for example. These individuals would still have the same understanding absent the agencies for which they work, and furthermore, the organisation as a whole can only be attributed this kind of ‘understanding’ in virtue of those individuals. In the same way, the room, book and Searle system requires only the rulebook and any person capable of following the instructions. The understanding can only be attributed to the person who wrote the rulebook, and not to the book by itself or the person following its instructions.

What lessons can we take from the systems reply that apply to the extended mind? The main reason the systems reply fails is due to equivocation between understanding by an individual and understanding by a ‘system’. Applying the term understanding to a system seems to lead us astray. I will now argue that the extended mind hypothesis also equivocates between definitions of ‘mind’ and ‘belief’, and its extension of these terms to systems greater than an individual leads to significant problems.

4. Analysis

The reason Clark & Chalmers seem to be able to argue with some conviction and success is because they equivocate between two distinct definitions of ‘mind’. The first kind of “mind” is the kind we all understand when we say the word – my mind is the seat of my consciousness, contained within my brain, the parts of which work together to perform all of the necessary cognitive processing, of language, perception, memory, etc. that I require to function. The second term, “mind” is used to denote the former, plus the objects of sense experience being processed by my mind. To equate these two things – the actual hardware necessary to process external stimuli in the first place, and the stimuli themselves, is to equate apples and oranges. Furthermore, the epistemological status of the link between the external object causing the experience and the perception itself is a matter of heated debate.

A further point is that the notion of ‘belief’ is also being confounded. In one sense, I would call my mind the store of my beliefs, but in a deep way, since it is the hardware that enables me to have beliefs in the first place and manipulate them in a holistic manner. An essay that expresses my opinion on some matter is also a store of my beliefs, but in a very different sense. In the case of the beliefs in my head, these may be very inarticulate and vague – many people have beliefs that are logically inconsistent. But when we actually have to pronounce on a topic, we formulate our beliefs into natural language, and these become codified into assertions. Dennett remarks that, in many cases, the pressure of having to pronounce a particular belief may cause us to say something we later regret.⁵ Thus, the beliefs on the page and the beliefs in the head are not one and the same. Secondly, we often speak of degrees of belief, a concept for which we seem to have an intuitive grasp when talking about beliefs in our head, but for which there is no easy expression in natural language. The example of the location of a museum blurs these distinctions because it is too simplistic – many beliefs are far more complex. Finally, expressing beliefs on a notebook precludes them from interacting in the holistic way that beliefs in the brain do. How can they when they are fixed representations on a page. Thus, it is not so easy to claim that the notebook is as legitimate a store of beliefs as the mind.

5. Conclusion and Final Thoughts

In general, the extended mind may first appear as a radical and exciting new argument for externalism. Its re-evaluation of boundaries opens up questions of the social nature not only of meaning but, more fundamentally, of the mind itself. Despite its promise, however, the thesis leaves even the most uncritically accepting with a somewhat sour taste in their mouths. The novel criteria proposed for drawing the boundaries of the mind end up being more arbitrary than those they were supposed to replace. More generally the question remains: where does extending our minds to a notebook or the Internet really get us? What new insights does accepting this paradigm enable us to discover? Putting our Popperian hats on for a minute, it seems like this kind of theory, below the surface, is devoid of real content. Almost anything, if interpreted in the right way, can be considered part of the mind, and thus the final nail in the coffin is that the theory admits of no refutation.

Postscript – The Meaning of Meaning and an Argument against Externalism

This discussion of the extended mind highlights an important confusion that plagues the externalist program, stemming from the definition of “meaning”. Meaning is something that cannot exist, like an abstract Platonic form, outside the mind. Every human mind is an isolated universe, in a sense. The meanings that I attach to words are my own and only my own, and they constitute what linguists call my "idiolect."⁶ Everyone has their own idiolect that we build by trial and error from interactions with other humans using the powerful linguistic inference tools with which we already come equipped.⁷ What makes communication possible is not that meaning is outside the head, but that the social nature of linguistic acquisition, together with shared common experiences and the assumption of an intersubjective world, guarantee much overlap in the idiolects of different people who belong to what we call the same "speech community."□