"Ever since men became capable of free speculation, their actions, in innumerable important respect, have depended upon their theories as to the world and human life, as to what is good and what is evil. This is as true in the present day as at any former time. To understand an age or a nation, we must understand its philosophy, and to understand its philosophy we must ourselves be in some degree philosophers. There is here a reciprocal causation: the circumstances of men's lives do much to determine their philosophy, but, conversely, their philosophy does much to determine their circumstances" (Bertrand Russell, p.ll. History of Western Philosophy, 1946).

Summary: The conception of what a science ought to be like and ought to achieve determines (as Russell above points out) the problems towards which research in that science is oriented, and how research efforts and funds are apportioned. However, if there has been a misconception and the dreamed of achievements are unattainable, what can be the possible relevance of work said to be a "contribution" towards their attainment?

The aim in the first and major part of this proposal is to show that Psychology cannot be the kind of science that we often think of as being the fully developed ideal of a science; that the aim of producing a complete Theory of Behaviour by a cumulative, step-by-step process (in fact, by any process) is unattainable; and that "pure" research purporting to be relevant to that end is, if of any general use at all, only of use by accident and to the extent that any specialised experience may be valuable in certain circumstances. In fact, the aim is to show that all formalised theories must fail just in that sense in which it was hoped that they would be successful.

The aim of any research should be to provide knowledge relevant to human concerns; but not only that, it is essential that the knowledge be both reliable and communicable. The formula devised in the physical sciences for ensuring this is inappropriate for Psychology and, in certain cases, may even be socially danger
 

"The given is -so one could say -forms of life"⁽¹⁾

Introduction - Theories: It is very much easier to obtain public agreement on the structure or form of states of affairs than on their meaning or content. This fact is made use of in mathematical logic, where all crucial judgements are reduced to ones concerning merely like or unlike sentence forms. Probably much of the attractiveness of formal theories comes from the feeling of surety that accompanies the simple decisions of whether a form is like or unlike another. It probably also contributes to the belief that knowledge embodied in a formal theory is in the most highly developed and general form, i.e. it is "scientific" knowledge. "We reserve the term 'science' (say, Cohen & Nagel, 1934) for knowledge which is general and systematic; that is, in which specific propositions are deduced from a few general principles." Psychology is often said to be a science (or a "hope of a science") because it both seeks such general laws (with regard to human behaviour) and uses systematic methods to attain them.

In such surveys as that of Marx and Hillix (1963), expositions of many of the more recent types of theoretical formulations in Psychology may be found. All purport to be a suitable basis for extension into or relevant for inclusion in a complete Theory of Behaviour. Hypothetico-deductive theories vie with field theories, and causal with purposive ones. But, in general, I think that it is true to say that all theories in Psychology are inadequate in at least some of the ways in which their critics say they are. This is the difficulty in which Psychology finds itself: there are no adequate theories, nor any which even approach adequacy.

In moments of despair psychologists admit that there is a real psychological component in the work of the historian, biographer, novelist, dramatist, etc., and that psychological problems can only be defined and discussed in terms of the material of these fields. It is sometimes difficult to attain any clear conception of a pure, general, contentless form of psychology. However, in moments of resolution psychologists assert that another and more adequate formal theory may yet be found; or that, as Hebb (as we shall see below) counsels, just more. and yet more of the same is needed; consideration of alternatives (to "mechanical cause and effect processes") is "weak-kneed." But even seventeen years ago (Marx, 1951), there was disquiet: "Although many psychologists have becorme increasingly alert wi thin recent years to the basic problems of theory construction, unfortunately there does not seem to have been a comparable general improvement in the actual construction of psychological theories," Marx thought that it would help to increase psychologists' skill in theory construction: I do not think that it has.

Over recent years there has been a tremendous increase in our understanding of both theory and method in science, much of it has come from the behavioural sciences. However, scientific psychology - the sort of psychology carried out in laboratories with electrodes and counters, where the results are evaluated by powerful statistical techniques - this sort of psychology still does not yet appear to give us that type of knowledge which seems to be of any help with very ordinary human problems .I think that it is because the "laboratory knowledge" is being gained with a totally unattainable end in mind and is only directly relevant to that goal. But as Wittgenstein (1953) put it: "In Psychology there are scientific methods and conceptual confusion" (p.232).

I think that the difficulty in which Psychology finds itself is not due to its failure to find the 'right' theoretical approach: it is due to its failure to appreciate what is involved in having any formal or formalised theory whatever (be it 'causal' or 'purposive'). The difficulty is not due to the fact that psychologists have formal theories, the truth is that more often than not they do not: it is due to the fact that they are pursuing such goals. Any denial that this is so must occasion the comment: if not these goals, then what ones?
 

The established theoretical approach in academic Psychology is the same as that in all other branches of science, and indeed it is the same as that in most areas of philosophy since Plato: it is the belief that if one is called upon to validate an assertion, to show that what one thought is correct, then it is both sufficient and necessary to show that if followed by accepted processes of reasoning from a set of acceptable or self-evident premises. Thus the aim in Psychology has been to produce theories exhibiting that or an equivalent formal structure; as Hull (1943) says:

"In science an observed event is said to be explained when the proposition expressing it has been logically derived from a set of definitions and postulates coupled with certain observed conditions antecedent to the event."

Now often it is denied that a full formalisation is the aim in Psychology. But nowhere in the mainstream of Psychology are general alternatives stated. While many psychologists have relinquished with relief the hypothetico-deductive method, in one form or another⁽²⁾ it is always the formalist's aim that is given explicit statement; and it is the formalist's aim that is communicated to students as being the overall purpose of Psychology; "(The) ultimate aim (of Psychology) is to produce theories about human behaviour that attain the same factual status as Newtonian theory has in its applications to physical objects" (Isaacs, Hutt & Blum, 1965).

The spectre of Newton lurks in the back-alleys of psychologists' minds. We think of climbing upon the shoulders of giants, that science and thus Psychology is cumulative, and that we can make step-by-step progress towards a body of knowledge that we can put on library shelves.

Hebb (1949) puts it this way:

"It might be argued that the task of the psychologist, the task of understanding behaviour and reducing the vagaries of human conduct to a mechanical process of cause and effect is a more difficult one than that of any other scientist."

"The problem of understanding behaviour is the problem of understanding the total action of the nervous system, and vice versa."

"It is only too easy, no matter what formal theory of behaviour one espouses, to entertain a concealed mysticism in one's thinking about that large segment of behaviour which theory does not handle adequately. To deal with behaviour at present, one must over simplify. The risk, on the one hand, is of forgetting that one has over simplified the problem; one may forget or even deny those inconvenient facts that one's theory does not subsume. On the other hand is the risk of accepting the week-kneed discouragement of the vitalist, of being content to show that existing theories are imperfect without seeking to improve them. We can take for granted that any theory of behaviour at present must be inadequate and incomplete. But it is never enough to say, because "we" have not yet found out how to reduce behaviour to the control of the brain, that no one in the future will be able to do so."

"Against the perspective of two thousand years⁽³⁾ the speed of our advance in studying human nature seems cheerful. At a rough guess, two hundred more years may bring the study of behaviour up to the level which physics reached in Newton's time."

But once again this is the formalist's aim in disguise. As Goethe said: "All that is fact is already theory, " and one of the crucial tasks in Psychology, which Skinner fails to notice, is to elucidate how we come to appreciate and to agree on what constitutes a "fact". Facts appear in Skinner's system by fiat, either in terms of positivist formulations or in terms of the mechanical formalism of a Skinner-box (at least, in terms of those few aspects of it that are deemed relevant).

It is to ways that one might proceed in Psychology without the need for formalisms that this research would be directed.

Later I want to bring to bear on the thesis of this proposal arguments derived from difficulties that have arisen in the field of Artificial Intelligences. But first I want to outline some very general arguments that have to do with the essential social nature of knowledge - the fact that theories need to be understood to be used by people other than those who discover them.

Given that Hull had been successful in completing his Principles of Behaviour, anyone possessing it would still be faced with the task of understanding it as Hull understood it: That is, they would first face the task of recognising that an observed event was truly an event of the types dealt with in the theory, (i.e. they would have to appreciate the perceptual ca tegories that the theory imposed and then recognise exemplars of such categories)⁽⁴⁾; next, it would also be necessary to produce an appropriate codification (formalisation) of the relevant antecedant conditions to the event. This presents the same type of recognition problem as above plus the necessity for an explicit determination of which of the antecedant conditions are actually relevant to the event that is going to follow them. However, we might ask, should not the theory, if it does make claims of being complete, indicate explicitly how the recognition, codification and decisions as to what is relevant are to be achieved? These are difficult enough demands to make, and I shall suggest in the discussions of mechanical pattern-Recognition and mechanical Problem-Solving that follow later that there are grounds for suspecting that they cannJt be met -attempts to satisfy them lead to irresolvable paradoxes.

But here I want to drive other paradoxes home. If a Theory of Behaviour is truly complete should we not also expect it to give, not only an account of the way in which the premises come to be looked upon as self-evident but also of the we.y in which the processes of reasoning are both carried out and agreed upon as acceptable? In other words: Stated within the formalisms representing the theory should be an account 1) of the rules or conventions via which the formalisms can be said to represent empirical events, and 2) those via which they can be related to each other to form the 'predictions' of the theory. However this cannot be, for even if some of the f?rmalisms of the theory were used to construct such an account, the rules via which they themselves were understood remain unaccounted for. Clearly any attempt to account for them formally would lead to an infini te regress; in practice we break the regress by saying: "We understand the rule to imply this, and this (supplying examples of the rule's use until the listener provides evidence that he has grasped it - perhaps a set of exercises could accompany the theory)."

Thus any theory, even the supposedly complete Theory of Behaviour, must remain essentially incomplete, and in a way which is particularly unfortunate for a Scientific Psychlology oriented towards a general theory - the lacuna appears right at the heart of the attempt to understand the nature of "understanding."

There are other ways in which one might approach the task of showing that the formalist's aim in Psychology is an unattainable one. I shall just list the following assertions summarily:

1) A Theory of Behaviour cannot contain an account of how it can be decided whether it is complete or consistent;
2) A Theory of Behaviour cannot be used by A to predict B's behaviour if B also possesses the theory;

but I have expanded upon the argument concerned with "understanding" as I think it is the one from which the most important implications for alternative ways to proceed in academic Psychology will follow.

After sketching some of the difficulties with formal theories, I now want to approach the thesis of the proposal from another angle. At the moment some fully formalised or "rational" theories do exist in Psychology: Mathematical learning and choice theories, Games Theory, Information Theory, Statistical Detection Theory, to name the main ones. What is interesting about these is that after an initial period - said to be a "warm-up," "practice", or just a "transitional" period - the theories do come to describe quite well what the subjects will do on average when in a specified situation. However, workers using such theories are unable to counter the suggestion that as long as the subject in some way comes to understand the experimental situation that he is in, in the same way as the experimenter, e.g. that in a choice experiment he accepts that the only options open to him really are those that the experimenter believes that he has presented, then the subject can come to play the Information Theory-game or the Statistical Detection Theory-game or whatever⁽⁵⁾.* The subject in coming to act as if he were following the tacit rules of the experimenter's game comes to exhibit behaviour that can be formally characterised: mechanical activity. That this demonstrates that humans, within a closed system of options, can come to mimic mechanical actions is not in dispute. It is the inability of any mechanical, formal theory to encompass the so-called "transitional" behaviour - which is ignored as irrelevant for the purposes of the experiment - that is the important point. It is precisely this ability: to come to act like a machine, as if to appreciate a tacit set of rules and act accordingly, that constitutes a central, if not the central, problem of Psychology.

But could there be a tacit set of rules via which one comes to appreciate all the tacit rule-systems that make up the various game-like situations in life? Does one come to appreciate the tacit rules of the grammar of a language in the same way that one learns and appreciates the rules in chess? While there is no clear goal in the language 'game', there is in the chess game - the 'rules' are obviously of a different character in these two cases. The cards do look stacked against the idea of constructing an artificial General Games Player (GGP).

The proposed research will aim at elucidating the proposition that: "Life is not the Game of games" or "GGP's cannot be made by formalist methods."

Introduction: Until now I do not think that the full force of the inadequacies in formal theories has been felt as, on the one hand, a clear understanding of their nature has only recently been attained in mathematics, and on the other, there has been no rigorous method of manipulating them in order to put them to test. Theorists find it very difficult to relax and just let the logic of the theory guidetheir actions; they are continually giving a theory the benefit of the doubt by bringing their own intuitions to bear in its interpretation. Psychologists have hitherto been able to disguise the failings of their formal theories from each other. As I have suggested, when the theories are fully formal and small scale, subjects come to act in accordance with them. When they are large scale theories tend to lapse into unfalsifiability, and their limitations cannot be seen.

Now, due to the fact that it is thought that any truly formal theory of behaviour can now be couched in terms of a progamme for a computer⁽⁶⁾, the possibilities inherent in the theories may now be mechanically and impartially derived. Thus the theory's power to encompass important aspects of human behaviour may be put to realistic test.

There is now no doubt that some very fundamental difficulties lie in the way of any major theoretical advances in Psychology. If a computer cannot be made to exhibit activities that we can appreciate as intelligent then the consequences of this will reflect back in an important way on Psychology. For the consequences will not only suggest, as other arguments do, that the aim of the general, formal, complete Theory of Behaviour is unattainable, but also they will help to make clear many of the more detailed implications of this main conclusion in the various special problem areas: for instance, pattern recognition, problem solving and language.

Pattern Recognition: Traditionally , approaches to the problem of pattern recognition have assumed that exemplars of a "universal" all have certain invariant features in common. These features may be quite concrete or very abstract. "Recognition" first requires identifying and listing these features and then programming them into machine. The machine must then be provided wi th procedures whereby, when confronted with an unknown pattern, it can extract the significant from the irrelevant features and then decide which of the lists of features that it contains most nearly matches those that it has extracted. The unknown pattern is then named with the name of the feature list. The most advanced form of such an approach is Uhr & Vossler's (1962), where the machine has the ability to extract that set of features which at any one moment allows it to reduce its tendency to confuse members of a limited family of patterns to a minimum.

This machine, although currently the most sophisticated of the well- known general (there are many highly developed successful recognisers for specific families of patterns) pattern recognisers, demonstrates a very paltry ability.

There is a difficulty with this sort of approach which has yet to be faced - there are no indications that it has yet been discussed, never mind proposals for its solution made. An example is the following, consider the pattern: A. Within the context of alphabetic characters it is an "upper-case 'a'," within the context of school geometry it is a "triangle with two sides extended," within the context of a drawing of a kitchen (say) it is a "representation of a pair of step-ladders, side-view," within the context of ...and so on.

One might say that the solution to this problem of context is simply to introduce into the pattern recognition scheme an explicit codification of the context in which the pattern appears, but this would also need recognising by the computer as a pattern. Thus this proposal is just to say in other words that the problem of pattern recognition can be solved by recognising patterns ad infinitum...

None of the well-known artificial pattern recognisers are able to exhibit that very common ability to recognise, as Wittgenstein puts it, family resemblances. (The situation where a group of objects belong together but there is no one feature common to all of them). We can come, after experience with a suitable range of examples, to appreciate that two physically quite dissimilar sequences of events can be accounted the same significance. For example, consider the expression "Keep cool cat." Are the initial onsets the same or different, they would all be marked by the phoneme /k/? Obviously they are all different, conditioned by the vowel that is going to follow them. They do have features in common, but these alone would not be sufficient to prevent confusions between /k/, /t/, /b/, and /p/ .How we are actually able to recognise /k/ sounds is still a mystery.

I have (Shotter, 1968) outlined an approach to pattern recognition that does to some extent cope with family resemblances, but it would be no help in building a general speech recogniser. Our abili ty to characterise the sound system in our language in terms of about 40 or so symbols seems to follow from our ability to use and understand language rather than to be a step in the process of understanding. Thus, while my recogniser can appreciate tbe family resemblances between patterns made up of well-defined elements, it does not contribute at all towards helping us to understand how we decide what are the significant elements in a newly encountered language - an ability that we can easily demonstrate. We may note, however, that this cannot be achieved by a completely non-participant observer, i.e. one who only hears tape recordings. In other words, use of language seems to precede any explici t apprecia tion of its formal structure and is a necessary pre-requisite. With regard to pattern recognition in general, an appreciation of the significance of the pattern would seem to be an essential part of the recognition process.

This shifts the fccus of the difficulties in pattern-recognition towards making an attempt to embody "intentions" or "purposes" into a computer. For it seems that the problem of context, raised above, is avoided by humans because all recognition takes place relative to certain concerns that a person has - in some way we are able to accept that certain aspects of our environment are relevant to these concerns while others are not. At the moment it is difficul t to indicate how these concerns may be formalised, but an alternative is to attempt to make a computer an autonomous entity such that it must be 'concerned' to maintain itself. But these are dreams of the future.

Addendum: Before leaving this section it may be worthwhile to point out that the preoccupation with "universals" (Plato's influence again?) has bemused almost all the workers in this field. However, it is quite clear that in practice not only do we use the same word in very different situations but quite often use different words in what are objectively the same situations. It is this last fact that has persistently been ignored.

Problem solving: Bold claims were made for the potential abilities of computers. It used to be asserted that computers could be made to perform any activity that could be described. (an assertion that was attributed to McCulloch & Pitts, 1943). Simon (1965) has now said that: "machines will be capable, within twenty years, of doing work that men can do." (Perhaps he would be on safer ground if he had said "... had done."). Earlier Simon (1957) had predicted that within ten years a computer would be chess champion and would prove an important mathematical theorem. These predictions have not been fulfilled.

One aspect of the difficulty seems to be in giving the appropriate explicit descriptions of the relevant activities - like patterns, activities seem only to heve family resemblances, which makes an explicit general codification of possibilities, as in pattern recognition, difficult. In fact, it seems to me, that it is precisely an equivalent difficulty.

Significant early successes in the field of problem solving came from the treatment of very well defined problem areas such as logic, chess, algebra, where there are families of very similar problems and it was possible to take advantage of the special structural features of "interesting" sets of problems in these areas. Thus it was possible to use heuristics - short cut, rules of thumb - which, while not guaranteeing success, worked more often than not. This was the line followed by Newell, Simon & Shaw (1964) in the construction of their General Problem Solver (GPS).

However, what GPS was unable to do was to learn from its experience. It was unable to appreciate that the problem it was currently engaged upon was very similar to ones that it had already solved. It embodied only heuristics provided by the programmers, ones drawn from their own intuitions, (i.e. N, S & S's intuitions).and was unable to capitalise upon the family resemblances between groups of problems to construct higher-order heuristics. The difficulty is to make it appreciate what is significant about the structure of a problem and what is irrelevant variation.

In mathematics, problems do seem to fall into families of "interesting" problems; interesting because they seem to require similar metriods for their solution, but also because of the implications that certain special members of the family hold out for further mathematical developments. If one were to assert that mathematician's intuitions, of what might constitute a further development and the point at which it was most likely to be made, could be formalised, it would be equivalent, again, to saying that their concerns, intentions and interests could be formalised.

These seem to be intractable problems even when the problem is one cf a well-defined family; but most of the problems we face do not occur as part of a logic-game, they involve us, they are of an ill-structured, once-off type: like choosing a new car. Here, just as in the information theory experiment, an unlimited set of conditions are potentially relevant. In the experiment, the subject is supposed to choose only between pressing one of eight buttons, he may initially consider: leaving the experiment, screaming, dreaming, attacking the experimenter, pressing a button, etc., as among the set of options open to him. Eventually, however, he comes to play the game and his activities can then be formally described. But in choosing a car there is no game to play. The person may take into account: his wife's aversion to pink, his son's sickness with soft springing, his own desire for excitement tempered by considerations of his bank balance, etc., or he may simply decide not to buy a car after all but a horse and cart, or even consider having his cwn high-speed automatic walking trousers manufactured. It would seem that there might be a potential infinity of facts of demonstrable relevance to the decision. In other words, there just seems to be no general rules about car buying. This may seem contrary tc fact, but what I mean is, that only within a closed system of options could there be such general rules, and that potentially at least the choice of how to limit the relevant options is still up to the individual concerned. Even when cultural pressures may operate to suggest uniformity, there is no way of finally robbing the individual of that choice: to create and limit his options as he sees fit. Thus (the Theory of Games, von Neumann & Morgenstern, 1944, notwithstanding) it is quite impossible for us to make an artificial General Car Buyer (G.C.B.) or any other General Something or other Buyer for that matter.

Even the preceding problems seem re1atively well-defined though when compared with the problems of understanding activities carried out for religious, ritualistic or (literally) mystical reasons, or simply by nature of play: Just as activities in themselves, they present the problem exemplified in: Browning's poem, The Grammarian: " I will not live, until I know how to live." In a social context, before life is routine, before it is clear what to do next, how does one know, in what sense does one know, how to proceed in the world? At the individual level the problem is even more mysterious. It now seems to be a possibility (Pronko et al, 1966) that the child 'creates' the "objects" of his world. Yet he does not passively do this, unable to act until he knows what to act towards; it seems to be only through the consequences of his activity that he comes to assign to certain portions of his environment the status of objects. But here also the child can view "this" as an object, now "that;" from this point of view, now that; with this interest in mind, now another. Where a "hand" stops and a "wrist" begins, where the "wrist" stops and the "forearm" begins, and so on, is undecidable in any objective fashion - it is relative to the current intents and purposes of the child (or adult) when using such names - a hard and fast decision is never made.

The child comes to 'create' the 'objects' of his world during play : but not while playing a game. Playing and playing games are two quite different activities. This is a distinction that once clearly drawn will, I feel sure, be a rich source of important implications.

One result of playful activity is that occasionally we recognise that its products fulfil an "up-till-then-unfelt" need. Necessity is more often than not, not the mother of invention; often the results of play are only much later put to use. For example, the child's development of speech would seem to conform to this pattern (Vygotsky, 1962). Where, onlyafter the child seems to have gained a degree of voluntary control over his articulators is this ability put to use as language. Also, there is evidence that many technological advances were derived from toys long in existence. The first steam turbine existed as a toy in Alexander's court long before it was developed for utilitarian purposes (Mumford: The Myth of the Machine, 1967).

What is the nature of play such that so much time, energy, concentration and feeling of emotional necessity goes into it?

Language: Chomsky's brilliant analyses of the structure of language have brought us up against the following problem: Is there any sense in which we might say that there exists "The English Language" that can be fully characterised in terms of sets of primitive components at various levels and rule-systems (phrase structure and tranformational rules in his scheme) for forming permissible sequences?

Well, Wittgenstein has outlined the notion of a language-game and its dependence upon a "form of life." More empirically oriented researchers (especially Bernstein, 1965) have noted the existence of self-contained language systems used within different sub-cultures in this country. Some speakers may have ready access to very general or academic codes (elaborated codes, as he calls them) others may only have ready access to very restricted codes. Thus it is perfectly feasible that two people speaking English can find each other quite unintelligible - it often happens, and is to be expected.

But while we have ready access to a number of different codes or ready made language-games (which seems the preferable term to "codes") we also very often demonstrate the ability to construct when necessary an appropriate novel language-game to meet special circumstances. Consider a novice coming to understand the terms used in sailing or in golf; no amount of talk will teach anyone how to play golf, you must try to play to learn to play. Once a modicum of experience has been gained experiences can be compared and contrasted, via language, with other players of the game - talk is only meaningful and useful if you are already a member of the 'club', having gained admission via your purposeful and actual attempts to play the game.

If this ability, to construct new language-games when necessary, were to be mechanised it would entail there being a formal description of an essential prerequieite: the nature of the shared non-linguistic knowledge. Is it to be described in a so-called "neutral" language? Once again we must bear in mind that, to those present in the environment attempting to communicate, they may take one aspect, one viewpoint as relevant and ignore the rest, and then to make what was relevant irrelevant and shift their attention elsewhere. But any formal language can only represent states of affairs via an unchanging set of pre-formed perceptual categories, whether it is a so-called "neutral" set or not. Once again the problem is: how can one decide what aspects, among the infinity possible, of the environment that people share, do they see as significant and what, but for accidental reasons, need not be there?

It was just this fundamental inability to formalise the knowledge over and above the linguistic knowledge that a successful translator must have to produce anything other than literal translations, that lead to the failure of mechanical translation projects (Bar-Hillel; 1959)⁽⁷⁾. Similarly, it was my initial failure to appreciate the full nature of the relation between the activities that accompany a child's speech and the speech itself that allowed me to propose some time ago (Shotter, 196a) that language acquisition could be understood in terms of a rule-learning machine. The recognition of the existence and central importance of that problem and my inability to deal with it, led to the abandonment of my project, to simulate natural language acquisi tion on a computer, and the writing of the current proposal.

As Chomsky points out, a child cannot learn his language by any known associationist or other principles. Important dependencies between the words in a sentence are not objectively marked. Consider Chomsky's example:

1) John is eager to please.
2) John is easy to please.
3) John is trying to please.

If we write:

4) John is eager to please us.
5) John is easy to please us.
6) John is trying to please us.

Or:

7) It is eager to please John.
8) It is easy to please John.
9) It is trying to please John.

we succeed in showing that "eager" and "easy" function in quite different ways to each other in 1) and 2) , but that "trying" can apparently function in both these ways (its ambiguity is semantic, not syntactical). Chomsky calls the actual sentence as uttered, the "surface structure" proposing that this is derived in an orderly manner (by the rules of transformational grammar) from a supposedly "deep structure" containing in an abstract (but formalisable) manner all the relevant information necessary to indicate the special types of dependencies involved. It is our ability to arrive at the "deep structure" of a sentence in some way that allows us to understand the differences between sentences 1), 2) and 3) even though they are not objectively indicated in the "surface structure." It is due to the remarks at the beginning of this section that I feel that Chomsky's endeavour: "to outline the processes that relate linguistic structures to physical events," is bound to fail.

"Unless we are able to specify and define, we are hardly in a position to argue. "(Leonard, 1968).

With regard to the problem of constructing blind aids, the procedure outlined in the appendix is perhaps a possibility, with regard to what Psychology at large might do I think the following discussion might be instructive: I have had teachers say to me that they cannot see the point of experimental psychology at all, as none of the theories of learning seem to have any relevance in the classroom. They then go on to talk about the problems in a vocabulary significantly enriched with terms derived from the failed psychological experiments. It seems that it is not the results of the experiments (what results there are) that are valuable, but the conceptual categories that researchers have invented in their attempts to understand the situation. The teacher seems to be able to take these over and, on occasions when he is in trouble and needs to look at his situation with another perspective, use them.

There has been a gradual change in Psychology from the use of "theories" to that of "models" without, as yet, change of accompanying investigatory methods nor a thorough analysis of the epistemology of.models of metaphors or analogies, etc.

If Philosophy is as Wittgenstein conceived it to be: a medicine to cure diseases, to cure aberrations of the system. Then perhaps it is possible to view Psycho1ogy as one such disease or aberration. As similar to that flood of talk that a child delivers to himself when he is trying to understand what actually is the problem that confronts him (cf., Vygotsky), how he should view it, how he should plan his course of action, what he is trying to achieve. It is here that the perceptual categories marked in his language and handed on to him from his culture help, as they tend in general to precondition him to categorise the situation in ways that have led to success in the past⁽⁹⁾. Psychology could be just that activity where not "what is" is investigated but "what might be." Such activity cannot be carried out without risk and in a disinterested manner; it is essentially a moral endeavour - I can see no way of avoiding that conclusion.

To the extent that social scientists are not just involved in arguing, but with acting; to the extent that they are involved in not only advising people about day-to-day affairs, but in advising governments; and to the extent that their proposals are based on formal theories that cannot contain any representation of intangibles, their activities are dangerous. It amounts to recommending that people should be either fitted to theories or denied existence; in essence, that freedom should be restricted for arbitrary reasons; that societies, now all but closed, should be finally closed; that a totalitarianism of rationalism should prevail; that the very processes via which the categories develop through which life becomes intelligible should be controlled or denied.

When, to quote Macfarlane Burnett quoted in the SSRC Newsletter No.2, social scientists can say: "In the eyes of the modern anthropologist the problem of today is how to use the intelligence of a relatively small number of men and women to devise ways in which patterns of behaviour laid down in a million years can be modified, kicked and twisted if necessary to allow a tolerable human existence in a crowded world." When social scientists view it as their destiny to rationally devise for us a pattern of life (where work is viewed as "occupational therapy", to quote from the same article) then I think that the time to be seriously worried has come.

I feel that the danger cannot be emphasised too strongly - there is ample evidence, from the writings of Herman Kahn⁽¹⁰⁾, that the formalists' mode of thought, that he argues is morally neutral or "Byzantine", is now the current mode of social planning in the U.S.A. I feel an alternative mode must be found, which allows discussion and decision making involving essentially ineffable values left unreduced, rather than unsuccessfully made tangible in terms of dollars, a mode that does not allow those concerned with human values being browbeaten into silence by accusations of "wishy- washy humanism" or "weak-kneed vitalism" because they are unable to make appeals to so-called objective facts. We face a shift from content to form, from the spirit to the letter, from what is desirable to what is feasible, from what is just to what is economic. Alternatives to our current procedures must be found.

The conclusions here will all be negative ones, the proposals to follow will be a little more positive.

There are many more implications of the attempt to produce formal, general, or univocal theories of human behaviour that one might draw out in a more detailed, technical and orderly way than I have done here. But the force of my argument (it has only really been sketched above) so far is this: that in those areas in Psychology where formal theories have been truly put to the test, by having them embodied in electronic computers, apparently insurmountable difficul ties have arisen. These amount not only to the difficulty of formalising the context or environment within which the behaviour takes place but the difficulty of finding some way to decide upon what aspects of the environment are at any one moment relevant and what unimportant.

I think that it should be possible, by reference to the fundamental theories of mathematical automata⁽¹¹⁾, to produce a convincing demonstration⁽¹²⁾ that very essential aspects of human activity are forever unformalisable. I think that not only will it amount to showing that, not everything that is essential to human intelligence can be formalised, but also the reciprocal assertion, that nothing essential to human intelligence can be formalised. Thus if these essentially ineffable attributes of human intelligence are not to be denied existence (a danger which currently seems to be very real), then some publically acceptable way of dealing with them, communicating about them, making decisions involving them, must be developed. In Psychology, this amounts to outlining alternative, non-formalist modes of investigatory procedures and ways for communicating the resultant information.

It may be useful to point out that while language translation has proved to be a non-formalisable process (an irrational process? Not of course not but certainly intuitive), nonetheless we do have ways of convincing ourselves that for all intents and purposes the translation is an adequate one. When a social scientist goes to live in an alien culture and then later attempts to "translate" his experiences and describe them in his native tongue how can we be sure that his description is an adequate one? Perhaps the parallels between these two types of activity are worth investigating.

That's a maybe, but there seem to be few doubts that the social sciences contain some radical misconceptions that are due to their attempts to ape physics. I think that these misconceptions can, if they are not already doing so, lead to some extremely socially pernicious consequences.

Addendum: I do not deny that anything that man has already done cannot be formalised, in some way or other, from some point of view or another. But I certainly do deny that it is possible to formalise everything that he is going to do or should do. Man seems to be not only creative in the systematic sense, i.e. that once he has a formal system he may derive infinitely many 'theorems' from it, but he is also creative in a transcendental sense, i.e. he can also create formal systems. I want here to coin the terms: "systematic creativity" and "transcendental creativity."

References:

Bar-Hil1e1, Y. (1959) Current Research &. Deve1opment in Scientific Documentation (Washington, Nat. Sci. Foundation) No.5, p.64.
Berstein, B. (1965) "A Socio-Linguistic Approach to Social Learning." In, Penguin Social Science Survey, 1965. Ed. J. Gould.
Broadbent, D.E. (1961) Behaviour, London: Methuen.
Chomsky, N. (1965) Aspects of the Theory of Syntax. Cambridge: MIT.
Cohen & Nagel (1934) An Introduction to Logic and Scientific Method. London: Rout1edge & Kegan Paul.
Festinger & Lawrence, (1962) Deterrents and Reinforcements. London: Tavistock.
Hebb, D.O. (1949) Organisation of Behavior. New York: Wiley.
Hull, C.L. (1943) Principles of Behavior. New York: Appleton-Century-Crofts.
Isaacson, Hutt & Blum, (1965) Psychology: The Science of Behavior. New York: Harper.
Kahn, H. & Weiner (1967) Daede1us, Summer 1967.
Leonard, J.A. (1968) "To formalise or not to formalise: that is NOT the question." Department of Psychology, Nottingham. (private circulation).
Marx (1951) Psycho1ogica1 Theories.
Marx & Hi11ix (1963) Psychological Theories. New York: McGraw-Hi11.
McCulloch & Pitts (1943) "A logical calculus of the ideas immanent in nervous activity." Bu11. Math. Biophys. 2,115-133.
Minsky, M. (1967) Computation. New York: J. Wiley.
Mumford, L. (1967) The Myth of the Machine.
Neumann & Morgenstern (1944) Theory of Games & Economic Behavior. Princeton Univ. Press.
Newell, Shaw & Simon (1959) "A report and general problem solving program." Proc. of International Conference on Inf. Processing, Paris.
Pronko, Ebert and Greenberg (1966) "A Critical Review of Theories of Perception." in Kidd & Rivoire, Perceptual Dev. in Children. London: Univ. of Lond. Press.
Rosenthal & Fode (1963) Behavioral Science, 8, Part 3, July 1963.
Shotter (1966) "The existence of the crossroads policeman." Nature, No.5047, Lond. July 23. pp.343-345.
Shotter, J. (1968) "A Note on a machine that 'learns' rules." Br. J. Psychol. 59, pp.173-177.
Simon, H. (1965) The shape of automation for Men and Management. New York.
Skinner, B.F. (1953) Science and Human Behavior. Appleton-Century Crofts.
Turing, A..M. (1936) "On Computable Numbers with an Application to the Entscheidungs problem," Proc. Lond. Math. Soc. ser.2. 42, pp.230-265.
Uhr & Vossler ( 1962) " A pattern-recognition program that generates, evaluates, and adjusts its own operators." In Feigenbaum & Feldman, Computers & Thought 1963. New York: McGraw Hill.
Vygotsky (1962) Language and Thought. MIT Press.
 

(i) Appendix: The Philosophy of Informationally Indeterminate Blind Aids

Introduction

Elsewhere I have offered objections to the view that everything that is essential to human intelligence can be formalised. One implication of these objections is that everything that is essential to visual perception cannot be codified: thus when a blind-aid, such as the Kay sonic aid, is asserted to provide a signal rationally calculated to provide sufficient ir!formation to enable blind people to make decisions regarding the structure of their environment, such an assertion must be denied.

Due to the structure of the Kay sonic aid - its transducers and electronic circuitry - explicit decisions about what aspects of the environment to code and what not to code must at some point have been made. It was these that must have been the basis of its design - it hardly matters whether this information was derived from the intuitions of its inventors or from the results of rigorously conducted experiments, the fact that it was incorporated in a formalised way into the aid is sufficient for us to be able to say that it must be essentially inappropriate. That is, inappropriate in the sense that it does not allow the user to select those aspects of the situation relevant to his purposes but imposes upon him a pre-established set of formal perceptual categories.

The fundamental theoretical arguments for this can be found elsewhere but, informally, they can be put like this: A person viewing a scene may first view it like this, then like that; from one point of view, then another; with one interest in mind and then with another. His interests determine both his viewpoint and what he deems relevant. Thus it is quite impossible to pontificate beforehand upon what alternative codifications of the information received from a visual scene would allow this sort of activity. Thus it is a totally inappropriate philosophy to assume that the provision of spatial distance receptors for the blind to aid their description of their spatial environment.

The following example might serve to clarify the situation further: The user of the long cane develops ways of manipulating it in order to make it reveal the sort of information that he can interpret and that he finds relevant to his need to progress confidently. Its limitations, are, at least to us sighted persons, obvious; although I would contend, we will run into difficulties if we try to state their precise nature. Now consider the following attempt to extend the compass of the long cane: we take the tip and the handle and cut out the intervening aluminum tube that transmits the perturbations of the tip to the handle, and substitute in the tip a radio transmitter and in the handle a receiver. (It would now be possible, if some way of making the tip's movement dependent upon the movement of the handle, for the range of the cane to be altered at will - say by variation of pressure on the handle.) However, what we have to decide now - a decision that we had bypassed before - is what movements of the tip are relevant and irrelevant, as it is only the relevant motions that will be encoded into the radio signal. Furthermore, we also have to decide upon the nature of the signal presentation at the handle: another decision as to what might be relevant or irrelevant. These decisions would, I maintain, present insuperable difficulties in the way of such a development of the long cane. It is impossible to decide beforehand what a long cane user deems relevant and the way in which he deems it so, i.e. categorises or encodes it. Thus it is necessary in the development of distance receptors to avoid the necessity for the imposition of pre-formed perceptual categories and to allow for the development by the user of his own idiosyncratic ones.

I have pointed out that the activity of the tip of a long cane is transmitted to the handle where it is interpreted by a user relative to his needs and relative to his manipulations of the cane, and that at no point in the process do pre-formed human perceptual categories intervene. It is now necessary to outline ways in which blind aids might be developed.

I can conceive of no ways other than the aid being developed by a blind person or by a sighted person prepared to go functionally blind for a period of time. It is only the blind person who can tell whether the aid that he is trying is of any use to him - for what means and in what situations. This is, in fact, what the philosophy of leaving the nature (the categorical form) of the information indeterminate directly entails. As we face an essential inability to make explicit the nature of the blind persons' needs, intentions and purposes, and to make explicit the way in which information may or may not be relevant to them, there is no other way than to take the actual body that incorporates them, and the ability to appreciate what is appropriate to them, and make use of that.

In one sense this amounts to saying that any activity that a blind person can carry out, or potentially carry out, at a distance may be of some use to him - not all of it will necessarily be of any help with regard to his mobility, it may be relevant though to deciding other things. To put it into slogan form it is activity at a distance that is important.

Let me give some wild examples, not worrying about social acceptability for the moment: It might be quite a useful thing for a blind person simply to have a stone on the end of a string. He could throw it every so often and, relative to his appreciation of the way in which he threw it, make some use of the time taken for it to strike something and the nature of the sound it then made. This activity would, of course, be no use at all if it was not used in conjunction with the blind person's attempt to achieve some purpose in that environment - this is essential. It was the failure to develop the Kay sonic aid with any clear feelings by its inventors of the purposes that it must satisfy so that they could gain immediate appreciation of its worth, that makes it virtually unusable by the blind. They provided information which the blind person should have found sufficient. However, it is not a question of sufficiency, it is a question of it being relevant to a purpose.

Now, to make the stone on the end of a string more socially acceptable we might conceive the idea of an electronic gun that would fire an electronic bullet into the environment at a certain velocity. It might then be feasible to make the nature of the gun's recoil dependent upon its time to hit something (using echo time) and the angle at which it was hit. The rate at which the trigger was pressed would, of course, be under the control of the user. Here again I am thinking of the user interpreting his environment in terms of the action that he was attempting to have upon it , and the action that it had on him as a consequence.

Deaf children, who do not learn their language as an accompaniment of activity are said to have no "depth of language." Blind people, who have to try to interpret the output of blind-aid that gives an output in terms of perceptual categories developed for entirely different purposes than their needs are in a similar position. Their problem is to try to decide what to do as a consequence of what they hear in the blind-aid. This reverses the natural requirement, to know something relevant to one's intentions and purposes.
 

Appendix II

Richard S. Rudner's (1966) arguments which seem to run exactly counter to my own arguments have been brought to my attention.

He asserts: "It is the function of science to describe the world, not to reproduce it."

I would like to introduce my criticism of his arguments in the following way. Consider an early carpenter with his tools making various items. I rather doubt if he begun his tasks with the appropriate blueprints before him; it is much more likely that he just set to and fashioned things according to his needs. It would only be much later, when he began to make things for other people, to satisfy their needs, that drawings became necessary for the customer to convey the nature of his need to the carpenter. Now the carpenter needs a picture that he, as a carpenter, understands. He might say: "But legs that thin can never be strongly fastened into the base - we'll have to change that." In other words, the drawing must depict something that the carpenter has it within his power to achieve relative to the skills, tools and materials that he knows that he has available.

Now some time ago, Eddington befuddled everyone with his introduction of the "physicist's table." A cloud of atoms in constant motion, restrained by various mutual attractions, repulsions and boundary conditions. He pointed out that a description of the table could be given by listing the position and velocity of all the particles involved. He then went on to ask which of the tables, the carpenter's or the physicist's, was the more 'real'. But notice he asked, "which of the tables" not "which of the descriptions."

Clearly, neither of the descriptions is more 'real': the carpenter's blueprint nor the physicist's listing of particles' positions and velocities, etc. The carpenter's blueprint would be of no more use to the physicist, with his atom smashers, electron accelerators and electron microscopes, than the physicist's listing of particle characteristics would be to the carpenter with his planes and chisels and so on. Any old description will not do: a description is only of use if it bears a relation to the methods that people have available for taking some sort of action in the world. It is Rudner's total failure to appreciate this that I wish to criticise.

I would now like to reproduce at least some of Rudner's argument to demonstrate its flavour:

". ..there have always been philosophers (Bergson and Whitehead are notable examples) who, while agreeing that causal or scientific explanations of physical phenomena are as much as can be understood by physical science, have nevertheless taken this very fact to be symptomatic of the deficiency or limitations of the scientific method - even as employed in physics. A typical claim maintains that science distorts (through abstraction from) physical reality. It is held, for example, that a scientific description of a tornado conveys in only a feeble, truncated manner what is, on the other hand, conveyed with overpowering richness and fullness by the direct experience of a tornado.
    ....the shortcomings of this view should be clear. We need but to remember Einstein's remark that it is not the function of science to give the taste of soup. It is the function of science to describe the world, not to reproduce it. Of course a description of a tornado is not the same thing as a tornado! And incidentally, the description does not "fail" to be a tornado on account of being incomplete, truncated, generalized, or abstract. Even if it were a "complete" description of a tornado - whatever that might be - it would still be a description of a tornado and not a tornado. Moreover, a description of a tornado no more fails to be a tornado than does a tornado fail to be a description.
    How, in the end, does all of this bear on Winch's argument? The answer is that Winch's argument commits a rather subtle form of the "reproductive fallacy"- reminding ourselves of the character of the fallaoy in the more neutral context of physical science (e.g., illustratively using tornados) allows us to discuss it more easily here. The claim that the only understanding appropriate to social science is one that consists of a reproduction of the conditions or states of affairs being studied, is logically the same as the claim that the only understanding appropriate to the investigation of tornados is that gained in the direct experience of tornados.
    We can scarcely entertain the idea that the only kind of understanding at which we can aim in the investigation of tornados must come from the experiencing of tornados. Notice that in rejecting Winch's thesis, it is not necessary to deny that some sort of knowledge or understanding of, say, religion is gained in "playing" the "religious game," any more than it is necessary to deny that some sort of knowledge or understanding is gained in experiencing tornados. The point is that nothing whatever in such a concession implies that these direct understandings are either the only ones possible for the social scientist or that they are a substitute for a scientific understanding of social phenomena.
    Neither Weber's arguments nor the more contemporary but still rather Weberian arguments of Winch are decisive, then, in compelling the conclusion either that social science must fail of achieving the methodological objectivity of the rest of science or that social science must employ a radically distinct methodology."

It is clear that Rudner does go as far as to talk about "some sort of knowledge or understanding," thus he acknowledges that there are at least different sorts. I quite agree with him that "direct understandings" are not "the only ones possible." The fact that social phenomena have been studied in the ways that he supports constitutes an existence proof of the possibility and makes his arguments as to their possibility quite redundant. What is at issue here is not whether any sort of understanding can be gained, but what type of understanding, and to what purpose.

I would like to say that the type of knowledge that can be gained by the use of methods of the same forms as those in the physical sciences yield knowledge of legislative use. It is knowledge which does not take into account Man's essential quality to create new circumstances.

References:

Rudner, R.S. (1966), Philosophy of Social Science, Prentice-Hall.
Winch, P. (1958) The Idea of a Social Science. Lond: Routledge & Kegan Paul.

Notes:

1. I think that Wittgenstein's notion, of a "form of life", will prove to be central to the research proposed here. At the moment I feel that we shall not be able to make intelligent computers until we can, in some way, give them a "form of life." They will need the equivalent of arms and legs, eyes and ears for acting in the world - they must have the capacity to lead their own autonomous lives. But it is just this - a "form of life" - that I think we cannot formalise in advance, preliminary to building them. In other words, should we ever, by some means, succeed in making autonomous artificial beings, we would face the same difficulties in understanding them as we now face in understanding ourselves. It is the elucidation of such assertions as the above to which the research would be directed.

2. Sometimes there is the assertion of a quasi-formal set of postulates with an appended demonstration experiment to make them clear. For instance, Festinger & Lawerence (1962) in "Deterrents and Reinforcements."

3. Is he here referring to the inception of this mode of thought with Plato?

4. It was to meet this need, of course, that led to the use of "operational" definitions.

5. It looks as if this might be the point at which to attempt to clarify the findings of Rosenthal & Fode (1963), on "experimenter bias." The issue of: "Who is conditioning who" in the classical "Boy have I got this guy conditioned"-joke has never been completely clarified either.

6. This follows from what is the fundamental hypothesis - Turing's hypothesis (1936) - in the theory of computing.

7. Either one can say that the difficulty was that a potential infinity of facte must be taken into account if certain sentences are to be disambiguated, or context could not be incorporated, or relevancy indicated. All are equivalent problems.

8. Due to their slaves, Greek intellectuals, perhaps, were only concerned to talk.

9. They may, of course, if the problem is really new, prevent him from ever viewing it in a suitable way: "There are children playing in the street who could solve some of my top problems in physics, because they have modes of sensory perception that I lost long ago." J. Robert Oppenheimer.

10. In Daedalus: Summer 1967, "Toward the year 2000."

11. Minsky (1967) Computation, J. Wiley.

12. Not to prove, that would entail the paradox of imposing a formal system on a human being again in order to show that they could not be one. The aim would be to show that by whatever logic the formalists endeavoured to prove their assertions concerning human behaviour, by that very same token their assertions could be shown to be inadequate.