an inductive argument for other minds

11
PETER RAY AN INDUCTIVE ARGUMENT FOR OTHER MINDS (Received 17 December, 1974) I observe that certain states of my body and certain of my words and actions are commonly accompanied by certain sensations and mental activities of mine. I also observe that Smith is often in similar physical states, utters similar words and performs similar actions. But I do not observe, in the same sense, that Smith's physical states and activities are correlated with feelings and thoughts on his part. He could be a mindless robot, as far as my observation goes. Nevertheless, I have no reason to think that he is a mindless robot, and his observable activities do so closely resemble the ones I engage in when I have thoughts and sensa- tions. So Smith, so much like me in the one respect, is probably like me in the other. He probably 'has a mind'. Such is the form of the usual analogical argument for the existence of other minds. Although an acceptable version of the argument may some- day be found, w;rsions that have so far been offered seem to be vulner- able to at least two of the attacks philosophers have made on them. First, the sample is restricted. The number of cases in which I have ob- served a correlation between mental and physical states or events is limited, and the ratio of observed correlations to the total number of observed physical states of the relevant kind is usually small. I have ob- served pain and screams together on only a few occasions, for example, and ratio of this number to the total number of screams I have heard is not high. My sample would seem too fragile to bear the weight of the conclusion which is supposed to rest on it. One can increase both the number of observed correlations and the ratio of these to the total number of observed physical states of the relevant sort, however, by choosing one's version of the analogical argument carefully. Consider this version, in which the sample class consists of a single entity: Philosophical Studies 29 (1976) 129-139. All Rights Reserved Copyright 1976 by D. Reidel Publishing Company, Dordrecht-Holland

Upload: peter-ray

Post on 10-Jul-2016

213 views

Category:

Documents


1 download

TRANSCRIPT

PETER RAY

AN I N D U C T I V E A R G U M E N T F O R O T H E R M I N D S

(Received 17 December, 1974)

I observe that certain states of my body and certain of my words and actions are commonly accompanied by certain sensations and mental activities of mine. I also observe that Smith is often in similar physical states, utters similar words and performs similar actions. But I do not observe, in the same sense, that Smith's physical states and activities are correlated with feelings and thoughts on his part. He could be a mindless robot, as far as my observation goes. Nevertheless, I have no reason to think that he is a mindless robot, and his observable activities do so closely resemble the ones I engage in when I have thoughts and sensa- tions. So Smith, so much like me in the one respect, is probably like me in the other. He probably 'has a mind'.

Such is the form of the usual analogical argument for the existence of other minds. Although an acceptable version of the argument may some- day be found, w;rsions that have so far been offered seem to be vulner- able to at least two of the attacks philosophers have made on them. First, the sample is restricted. The number of cases in which I have ob- served a correlation between mental and physical states or events is limited, and the ratio of observed correlations to the total number of observed physical states of the relevant kind is usually small. I have ob- served pain and screams together on only a few occasions, for example, and ratio of this number to the total number of screams I have heard is not high. My sample would seem too fragile to bear the weight of the conclusion which is supposed to rest on it.

One can increase both the number of observed correlations and the ratio of these to the total number of observed physical states of the relevant sort, however, by choosing one's version of the analogical argument carefully. Consider this version, in which the sample class consists of a single entity:

Philosophical Studies 29 (1976) 129-139. All Rights Reserved Copyright �9 1976 by D. Reidel Publishing Company, Dordrecht-Holland

130 PETEK R A Y

(A) (1) Smith and I have properties P2... Pn in common. (2) I have property Px (being in pain when screaming). Therefore, probably (3) Smith has property Pl (and hence experiences sensations,

'has a mind').

Compare it with this version, in which the sample includes thousands of entities:

(B) (1) Smith's words and my words have properties P2 ... Pn in common, including the property of being uttered by people who have properties of Po.-. Pz in common.

(2) Most of the thousands of words I utter every day have property P1 (being accompanied by a thought).

Therefore, probably (3) Smith's words also have property P1.

Now compare these with a third version, in which both the number of observed correlations of the mental and physical entities is large, and the ratio of observed correlations to observed physical entities of the relevant kind is high. This version seemed particularly convincing when I met Smith early this morning after emerging from ten years' seclusion:

(C) (l) Almost one hundred per cent of the millions of words I have heard during the past ten years have property P1 (being accompanied by a thought).

(2) Smith's words have properties P2... P, in common with all these millions of words.

Therefore, probably (3) Smith's words have property P1.

It appears that if restricted samples are a problem for the analogical argument, it is not just a problem of too small a number or percentage of observed correlations between mental and physical events, but rather one of too restricted a kind. It is not that I have failed to experience enough cases to justify the conclusions of (B) and (C), but rather that my experience has somehow not been broad enough.

Unfortunately, my sample is not to be broadened by any further ex- perience. Perhaps I could 'observe' another person's pain when he is

AN INDUCTIVE ARGUMENT FOR OTHER MINDS 131

screaming by feeling the same pain myself - at least no one has yet shown that two people cannot feel the same pain. But I may not add this 'obser- vation' to my sample in (A) unless I know that it is an observation of that otherperson's pain and not just of my own. And it is difficult to see how I could determine this without first determining that the other person is feeling a pain at all. 1 Thus, if I am to base my argument for other minds on my observation that screaming is accompanied by pain on the part of the screamer, my argument will have to be based on the correlations I know to hold in just one case - my own. As Ayer says, " I f it is required of an inductive argument that the generalization to which it leads should be based on a wide variety of experienced instances", the analogical argument fails the test. 2

Some philosophers have suggested that an inductive argument may yield knowledge~, or at least 'reasonable belief', even if it is not based on 'a wide variety of experienced instances'. All that is required, according to this view, is that we have no evidence that our sample is biased. Michael Slote writes:

If all the x ' s one has sampled have been found to be f ( a n d one has sampled numerous x's), and if one has after careful, rational, thorough examination of one's evidence dis- covered no reason to think that one's sample is unfair or biased, then it is reasonable to believe that all x 's a r e f a

I doubt that this is an acceptable principle, however. If my friend Klock gives me ten bags of marbles and I look in just one, finding it to contain only red marbles, I shall not have much evidence that the other bags do or do not have red marbles in them. I might carefully examine each marble in that fLrst bag, think about all the bags of marbles I have been given in the past (none), consider everything I know about the donor (nothing relevant to this issue), and conclude that I have no reason to think that the sample I have examined is unfair or biased. But surely it would not be reasonable for me to conclude that all (or any) of the mar- bles in the other bags are red. My sample may be large, but it is restricted to that one bag. After taking fifty red marbles from the bag I may infer that the next one I take from it will also be red, but not that the marbles in any other bag are (or are not) red.

Some philosophers have thought that an inductive argument is suff• ent for knowledge, provided that its sample is not known to be biased and that it is 'in principle' impossible to broaden the sample. In the case of

132 PETER RAY

the marbles, I can look into the other bags, but I cannot in principle look into any other minds to strengthen the case for Smith's mindlessness. Since my sample is not only limited, according to this view, but is (in some sense) necessarily limited, I should be satisfied with it.

This view also seems to me to be mistaken. Showing that an analogical argument with a limited sample has a necessarily limited sample not only fails to strengthen the argument; it even appears in some cases to weaken its plausibility. Consider, for example, this argument, whose sample is limited to objects of certain size:

(D) (1) All objects I have observed have a determinate position and momentum at any one time.

(2) This electron is an object. Therefore, probably (3) This electron has a determinate position and momentum

at any one time.

In this case, I cannot expand my sample to include any electrons, for it is in principle impossible to determine an electron's simultaneous position and momentum. This fact has led Bohr, Heisenberg and others to suggest that electrons do not have a determinate position and momen- tum at any one time. 4 That is, some people seem to believe that the con- clusion of(D) is false precisely because the sample is necessarily restricted. Now, I do not mean to imply that the impossibility of expanding a sample always undermines an inductive argument. But I do think that this example shows that a necessarily restricted sample does not guarantee the validity of an inductive inference.

II

Let us turn for a moment to another problem which is faced by the analogical argument for other minds: the ditficulty of excluding parallel arguments which yield unwanted conclusions. My argument (A), for example, can be altered very slightly to give a new conclusion:

(E) (1) Smith and I have properties P2.-. P, in common. (2) I have property P1 (being in pain when I am screaming). Therefore, probably (3) Smith has propertyP1 (being in pain when I am screaming).

A N I N D U C T I V E A R G U M E N T F O R O T H E R M I N D S 133

An even more disturbing conclusion results from this version of (B):

(F) (1) Smith's screams and my screams have properties P2... Pn in common.

(2) All of my screams have property P1 (being accompanied by pain on my part).

Therefore, probably (3) Smith's screams also have property P1 (being accom-

panied by pain on my part).

Plantinga, Slote and others have searched for principles which would separate valid from invalid analogical arguments, or more precisely, which would exclude arguments whose premises seem to be true but whose conclusions we believe to be false, while leaving untouched those whose premises and conclusions both seem to us to be true. 5 But no principle has yet been found which adequately does this job, a fact which may indicate that analogical arguments like (A)-(F) do not yield know- ledge.

Parallel argurnLents might be less troublesome if we could resolve the problem of restricted sample. If I could observe (or otherwise know) not only that I feel pain when I am screaming, but also that Jones, Barnes and Randolph feel pain when they are screaming, my inductive infer- ences about Smith's mental life might be more reliable. We could then argue, for example, that:

(c) (1) Smith, Jones, Barnes, Randolph and I have properties P2..- Pn in common.

(2) I am in pain when I am screaming. (3) Randolph is in pain when he is screaming. (4) Barnes is in pain when he is screaming. (5) Jones is in pain when he is screaming. Therefore, probably (6) Smith is in pain when he is screaming.

This argument has a broad sample, and equally important, a plausibility which is not wea&ened by the existence of parallel arguments having un- wanted conclusions. Unfortunately, I cannot use (G) to show that Smith is in pain when he is screaming without first showing that the premises,

134 VETER RAY

particularly (3)-(5), are true. Since I cannot observe that anyone else is in pain when he is screaming, I seem to be left with my original problem: Is it fair to generalize from my own case?

I I I

I believe that there is an acceptable inductive argument for the existence of other minds which does not generalize from my own case - an argu- ment which is not based on my observation that I have a mind or that others resemble me in certain respects. It is, like (G), an inductive argu- ment based on a broad sample, but unlike (G), it is not an analogical argument and it contains only premises which I know to be true inde- pendently of the other minds issue. The 'argument from Smith's reliabili- ty' might roughly be sketched as follows:

(H) (1) Smith usually tells the truth. (2) Smith tells me that he is in pain. Therefore, probably (3) Smith is in pain.

I f (H) is to be acceptable as a proof of other minds, its premises will have to be interpreted in such a way as not to presuppose intentions, feelings, or mental processes on Smith's part. The statement 'Smith is usually honest', for example, could not be used in (H) because it ascribes a certain mental state to Smith, and so begs the question. I f Smith is to be honest, it is not enough that the propositions he utters happen to conform to the facts; he must also intend that they do so. Is this also the case with 'telling the truth?' Can one only tell the truth if one intends that one's statements be true?

Perhaps we can sidestep this question by eliminating 'tells the truth' from premise (1). We are not concerned with Smith's intentions in this premise, but only with the fact that what he says generally turns out to be true.

(I) (1) Most of the propositions uttered by Smith are true. (2) 'I, Smith, am in pain', is a proposition uttered by Smith. Therefore, probably (3) 'I, Smith, am in pain', is true.

AN I N D U C T I V E A R G U M E N T FOR O T H E R M I N D S 135

It may be objected that 'uttering propositions' is also an intentional phrase. Perhaps the noises 'I, Smith, am in pain' are not meant by Smith to convey a proposition at all. Perhaps they are like the rustling of the leaves which may sometimes seem to be conveying a proposition, but really are not.

Yet, the noises 'I, Smith, am in pain' do express a proposition in my language. Whether or not the words mean anything to Smith, and whether or not tkey mean the same thing to him that they mean to me, I know perfectly well what they mean to me. And being a close companion of Smith, I have heard him utter a large number of noises which express propositions in my language. I have carefully checked to see whether the propositions expressed in my language by all these noises are true, and have discovered in most cases that they are. Thus, when he utters the noise 'I, Smith, am in pain', I infer inductively that this noise also ex- presses a true proposition in my language. A further revision of our argument may help to make this dear:

(J) (1) Most of the noises that Smith has made which are such that they express propositions in my language are such that these propositions are true.

(2) Smith has made the noise 'I, Smith, am in pain'. (3) The noise 'I, Smith, am in pain' expresses a proposition

in my language. Therefore, probably (4) The proposition expressed in my language by the noise

'I, Smith, am in pain' is true.

Perhaps Smith means by 'I, Smith, am in pain' what I mean by 'I, Smith, am not in pain', or by 'Remember the Maine!' Or perhaps uttering 'I, Smith, am in pain' is just Smith's way of sighing or yawning. It makes not the slightest difference what meaning, if any, he assigns to the words. (J) is not an argument from the existence of public language. It does not state that if Smith's noises express true propositions then he must assign the same meanings to them that I do, or that he must assign any meanings to them at all. Rather, it states simply that the relationship which has been found to obtain in the past between certain of Smith's noises and the truth value of certain propositions in my language probably continues to hold when Smith's noise is 'I, Smith, am in pain'.

136 PETER RAY

The distinction between our reliability argument and Price's 'language argument' 6 may be put in another way: As soon as I discover that most of Smith's noises which convery propositions in my language convey true propositions in that language, I may be tempted to speculate that Smith has a mind, that he speaks the same language I speak, and so on. But it should be remembered that this is speculation; it is one possible explana- tion of why Smith's noises have a special meaning for me that somehow corresponds to the truth. Perhaps it is a useful explanation for some purposes. Perhaps it is the best explanation I have been able to come up with so far. Yet it is not until I hear Smith say 'I am in pain' or 'I think' or 'I hope' that I know that he engages in mental activities as I do. And this knowledge is based not on my knowledge that Smith speaks English, but rather on my knowledge that his noises are generally reliable indica- tors of what is happening in the world, just as the height of the mercury in a thermometer is a reliable indicator of what the temperature of the air is. 7

IV

There are several considerations which may appear to limit the value or undermine the effectiveness of arguments like (J). I f (J) is successful in giving me knowledge that Smith experiences sensations, for example, it is only because Smith is an unusually 'reliable' fellow. The noises of most human beings either express false propositions in my language much more often than Smith's do, or else they express no propositions in my language at all. Thus, the application of arguments like (J) would seem to be ex- tremely narrow. I can use them to establish the mindedness of only a few very truthful and knowledgeable English speaking people.

This limited application, however, is sufficient to give the reliability argument considerable value. For having used it to show that a few 'reliable' people probably experience sensations, we can go on to use analogical arguments like (G) to extend our knowledge of other minds to anyone we wish, so long as each new subject looks and acts enough like our original 'reliable' ones. The importance of arguments like (J) is not derived from their wide applicability, but from the fact that they allow us to break through the barrier of 'generalizing from a single case'.

A second and more serious problem for the reliability argument lies in our capability of building machines to imitate Smith both in his general

AN I N D U C T I V E A R G U M E N T FOR OTHER MINDS 137

reliability and inL his claim that he sometimes feels pain. I f we were to program a present-day computer, for example, to print 'I, Computer, am in pain' at the end of each print-out, we could construct an argument like (J) with a presumably false conclusion: that the computer experiences sensations. Since the new argument would not yield knowledge of another mind, this objectlion goes, there is no reason to think that (J) does so either.

There is something illegitimate about our use of this new argument, however. It is as if we took a bag of marbles all of which we knew to be red, placed one unseen marble in the very bottom of the bag, drew the others out one by one, finding each to be red, and concluded that the unseen marble must also be red. Surely, putting the unseen marble into the bag of already examined marbles discredits any inductive inference we may make about its color, while leaving untouched similar inferences with regard to marbles in bags that have not been tampered with. Similar- ly, the possibility of our programming a reliable computer specifically to say that it is in pain does not discredit (J).

I f an extremely reliable and sophisticated machine not specifically programmed to do so suddenly made the noise 'I, Machine, am in pain', on the other hand, I suggest that we would have some reason - and in the absence of any counter-evidence, adequate reason - to conclude that it really is in pain. As Micahel Scriven has pointed out, 8 one way to find out whether robots can be made which have sensations is to build a few 'truthful' (reliable?) ones and ask them whether they do. If the answer is 'No' , we construct further robots on different principles and ask again. When the answer finally is 'Yes', we may reasonably infer that there can be robots which have sensations - provided, of course, that we do not have stronger evidence to the contrary.

A final objection to the reliability argument is that its sample, like those of (A)-(C), is ~lot broad enough to warrant its conclusionP For (J)'s sample is restricted to those of Smith's noises whose corresponding propositions in my language I can verify by observation, while its con- clusion is about a noise whose corresponding proposition in my language I cannot verify by observation. No matter how many of the former sort of noise I examine, according to this objection, I can infer nothing about the truth value of propositions expressed by the latter sort of noise.

The present objection to (J) is similar to W. T. Stace's claim that the existence of unperceived objects is in doubt because no inductive argu-

138 PETER RAY

ment for the existence of unperceived objects can have such an object in its sample, 1~ and like the claim of some philosophers that the existence of anything in the past is questionable because all arguments for past entities are based entirely on our knowledge of the present and future. There is a gap between sample and conclusion in each case which is thought to undermine these arguments' validity.

The gap between verifiable premises and unverifiable conclusion, how- ever, exists not only in arguments for the existence of unperceived objects and entities in the past, but also in a great many other inductive argu- ments which we normally accept, including all those with conclusions like 'Every substance has a solvent'. If the gap does not undermine the validity of all these arguments, I see no reason to think that it disturbs the inference in (J).

v

Having found that the usual analogical argument for other minds is grounded on too narrow a field of experience to yield knowledge, we have discovered an ordinary enumerative inductive argument which is based on a broader sample. The 'argument from Smith's reliability' states that since Smith's sentences have generally expressed true proposi- tions in my language in the past, they probably continue to do so when his sentence is 'I, Smith, am in pain'. I believe that the argument gives me knowledge of other minds. If it does not, its failure is matched by the failure of most enumerative inductive arguments which have heretofore been accepted.

University o f Pennsylvania

NOTES

z Cf. Alvin Plantinga, 'Induction and Other Minds', Review of Metaphysics 19 (1965-6), 441-461.

A. J. Ayer, The Problem of Knowledge, Penguin, London, 1971, p. 222. 8 Michael Slote, 'Induction and Other Minds', Review of Metaphysics 20 (1966-7), 341-360; quotation from p. 353. 4 See P. W. Bridgeman, 'Determinism in Modern Science', in Determinism and Free- dom in the Age of Modern Science (ed. by Sidney Hook), Collier-MacMillan, London, 1970, pp. 57-76. 5 See Plantinga, op. cit. ; Slote, op. cir.; Plantinga's 'Induction and Other Minds lI', Review of Metaphysics 21 (1967-8), 524-533; Slote's Reason and Skepticism, Humani-

AN I N D U C T I V E ARGUMENT FOR OTHER MINDS 139

ties Press, New York, 1970, pp. 111-135; and Jerome Gellman, 'Inductive Evidence For Other Minds', Philosophical Studies 25 (t974), 323-336. e H. H. Price, 'Our Evidence For the Existence of Other Minds', Philosophy 3 (1938), 425-456. 7 Another argument for other minds which bears superficial resemblance to the one proposed here has been offered by Stuart Hampshire, 'The Analogy of Feeling', Mind 61 (1952), 1-12. Hampshire claims that each person can establish the reliability of our ordinary inferences from someone's exhibiting pain-behavior to someone's being in pain, because he cart see that others correctly make such inferences about himself. "We are each generally in a position to reassure ourselves about our methods of inference to the feelings of others," he writes, "by confrontation with the successes and failures of others in talking about us." This 'reassurance', however, is not much help to the skeptic who doubts whether anyone else can be known to have a mind with which to make inferences in the first place. s Michael Striven, 'The Compleat Robot: A Prolegomena to Androidology', in Dimensions of Mind (ed. by Sidney Hook), Collier, New York, 1961, pp. 113-133. a Cf. G. Schlesinge~r, 'Induction and Other Minds', Australasian Journal of Philosophy 52 (1974), 3-21. ScMesinger's paper contains an elementary version of the reliability argument, which he, apparently arrived at independently of the present effort. 10 W. T. Staee, 'The Refutation of Realism', Mind 43 (1934), 145-155.