Lecture 4

Instrumentalism

DENNETT DEFENDS THE IDEA THAT WHEN WE EXPLAIN SOMEONE'S BEHAVIOUR IN TERMS OF BELIEF AND DESIRES WE ARE NOT INVOKING INTERNAL STATES. WE ARE INVOKING THEORETICAL CONSTRUCTS WHICH HAVE BEEN ELABORATED BECAUSE OF THEIR EXPLANATORY POWER.

ORDINARILY WE USE 'BELIEF+DESIRE' EXPLANATIONS OF HUMAN HAVIOUR

So far we have mostly addressed 'the problem of consciousness': Place's thesis that science will probably show one day that consciousness is a brain process, Jackson's counterargument that this is something science will never be able to do, because of what we know about the nature of consciousness now.

I want now to turn to the question of what a belief is.

Buzz: in groups try and answer the question What is a belief? Try and write down a one sentence answer.

I now want now to develop a particular view on what beliefs are or are not. It comes from Daniel Dennett, my hero. And it is a view that is compatible with a scientific approach.

Dennett says beliefs and desires are constructs - theoretical constructs.

Note: Dennett, and other contributors to the debates covered in the Reader, change their minds from time to time. In discussing their views it is best to refer to partiicular articles or books.

See the paper by Dennett called 'True Believers' from page 150 in Lycan, 1st edition.

THE CONCEPT OF A 'THEORETICAL CONSTRUCT'

Dennett's starting point is the very important interest he thinks we have in predicting people's behaviour. We are constantly observing what the other people around us are doing and constantly we are interested in predicting what they will do next. This is the stuff of everyday human intercourse.

To help us predict the behaviour of others we have developed it is argued, a theory. We have developed a sort of informal, everyday psychology. It has been called folk psychology. The theory is a good one - we find that when we use it to try and predict people's behaviour it works.

________________________________

FOLK PSYCHOLOGY AND THE EVERYDAY PREDICTION OF BEHAVIOUR.

What is it, this theory?

Its basic terms are belief and desire Or belief and want, if you prefer.

Folk psychology explains almost all human behaviour in terms of belief and desire.

"He did this because he wanted such and such and thought doing this would get him it."

That is a sort of explanation schema into which a great deal of everyday explanation of human behaviour fits.

Try it.

I shall suggest behaviours and you suggest some possible explanations.

Why have I got this key?

Why did I come to campus along the A6 this morning?

Why did Monica Lewinsky keep that dress?

Why did they try and separate those twins last night?

The pattern is: we believe such and such; we want such and such; this leads us to do such and such.

___________________________________

BUZZ: Can anyone think of a piece of human behaviour which we would not ordinarily invoke a belief-desire explanation?

___________________________________

This approach to explanation goes along with belief-desire approach to prediction.

What will Charlie do next?

It will depend on what he wants and what he thinks will get him it.

Supposing we know he has missed his lunch. We may guess that he will be hungry. We may think he believes that eating a little something in the lecture will alleviate his hunger. And so we may predict that he will start eating before long.

Will Brown cut fuel tax?

What does he want? He wants to avoid sending the economy into a spin. He believes that cutting fuel tax significantly will send the economy into a spin. Therefore one predicts that he will not cut them.

So here we have folk psychology at work, explaining and then generating predictions of behaviour by ascribing beliefs and desires to people.

You can easily think of this as a sort of mechanical theory. It seems to be saying what brings a particular bit of behaviour about is the occurrence of two things, a particular belief and a particular desire. When these occur together the behaviour is generated.

That way of thinking treats the belief as something that is in place before the behaviour occurs, and it treats desires in the same way. It suggests these are two things which are in place or take place in the brain or in the mind and that they bring about the behaviour.

Buzz: who thinks this is the way to think of beliefs and desires?

But Dennett says it needn't be like that.

You don't have to think of a belief as something that is there in the brain or in the mind prior to the behaviour.

What is his alternative?

He says what we have posited is that predictions based on a belief-desire analysis are quite good. Let's just leave it at that! Talk of beliefs and desires gives good predictive power. We don't have to say any more about them. We don't have to think of them as exiting in the barin or mind of the person behaving. We find that by attributing beliefs (and desires) to a person we can predict their behaviour. But this doesn't mean these things are any kind of episode or state in the mind or brain. They could be just 'constructs'.

This success of the belief-desire model means that these concepts have a certain utility. They have proved themselves useful in predicting behaviour. They are instrumental in the task of prediction.

This is how Dennett thinks of the concept of belief: it is useful in a very successful theory we have developed for explaining human behaviour.

Does this means that Dennett thinks beliefs don't actually exist? This is a good question.

Dennett thinks the concept of belief can be useful - instrumental - in prediction without their having to refer to anything particularly.

They don't have to refer to events, or states, or episodes,

to physical events or states or

to mental events or states.

So long as we know how to use the concepts consistently, we don't need to make any assumptions of this kind.

They do their work,

play their role in helping us predict behaviour accurately, quite irrespective of what events

or what states

may or may not be inside the person.

POSSIBLE CANDIDATES FOR 'THEORETICAL CONSTRUCT' STATUS.

LINES OF FORCE.

There is perhaps a parallel with lines of force.

If you take a bar magnet and let iron filings arrange themselves freely they will fall into a certain pattern.

You can predict the pattern by saying:

there are lines of force

set up by the magnet's poles,

and the filings arrange themselves

so as to align themselves with these lines.

What are the lines of force?

The line of force is a concept which is instrumental in making the behaviour of the filings predictable.

That is all we need say.

There is not necessarily any event or state or process corresponding to them in the region of the magnet.

The concept doesn't get its sense

by referring to any such things.

It is useful.

It proves its usefulness by playing apart in successful prediction.

So with beliefs and desires.

These concepts are instrumental in making successful predictions of human behaviour. But that doesn't mean they refer to real states or real events within the person.

This is the instrumentalist construal of mental states, of beliefs and desires in particular.

Is there a better parallel than lines of force?

ATOMS; ELECTRONS; CHROMOSOMES; GENES; FORCES; VICE-CHANCELLORS, BLACKBURN.

Some philosophers have been taking an instrumentalist view of all concepts used in theories where straightforward observation seems to be excluded.

The concept of the atom would then be an example.

Atoms, at some point at any rate, were thought to be unobservable.

But they were and are invoked by successful physical theory.

The success of that theory did not show, instrumentalists argued, that atoms must exist.

The concept of the atom could and should be regarded as purely instrumental:

not referring to entities at all, but just constructed by the theory to make prediction possible.

The atom is sometimes said to be on this account a 'theoretical construct'.

The Electron

The concept of the electron is another example. It is a mistake to think of electrons being very small Ping-Pong balls, or very small clouds of charge; wrong to think that with sufficient magnification we might be able to see one.

The electron is a construct to help us predict.

Others

Chromosomes, genes, forces, vice-chancellors, have all received this treatment at various times.

These terms, it is said, do not refer to real things

that are too small

or too remote

to be seen.

They are purely conceptual devices to help us explain what happens and predict what will happen next.

Inflation, confidence, public opinion, the student grant are also vulnerable.

This is Dennett's approach to the question of what a belief is. But it bears back on the problem of consciousness.

It is Dennett's programme to show how beliefs and desires could be regarded as theoretical constructs.

[Later:

The advantage of this strategy is that it bypasses the traditional impasse that concentrating on consciousness gets us into.

If we understand beliefs etc as conceptual devices paying possible part in explanation we don't need to bother whether system is conscious or not. The key question becomes: does it display behaviour of such kind and of such complexity that intentional explanations - explanations in terms of belief and desires, - are required to explain/predict it.]

DENNETT DEVELOPS HIS POSITION THAT BELIEFS AND DESIRES AND OTHER SO-CALLED MENTAL WHATSITS ARE REALLY LOGICAL CONSTRUCTS BY DEVELOPING THE CONCEPT OF THE INTENTIONAL SYSTEM.

IF WE WISH TO PREDICT AND/OR EXPLAIN BEHAVIOUR, SAYS DENNETT, WE MAY ADOPT ANY OF THREE DIFFERENT STANCES.

1. The physical stance.

If science is on the right lines, it should in principle be possible to predict the behaviour of all physical systems by knowing the position of all the elementary particles they are made up of and the forces impinging on them.

This is the vision of Lapalce in the 19th Century, that if you knew the position of all the atoms of the universe , and all the forces acting on them you could predict what the state of the universe would be at any moment in time: the future as well as the past would be open to your eyes.

More mundanely, I can predict the behaviour of this table by treating it as a physical object subject to the laws of physics. On that basis I think it will probably stay put for the forseeable future. We can and do predict the bahviour of lots of things by thinking of them as physical objects subject to familiar physical laws.

'From this stance our predictions are based on the actual state of the particular system, and are worked out by applying whatever knowledge we have of the laws of nature.' (Dennett, in Lycan, p.168)

For a system as complex as a computer running a chess programme, the task of using the physical state of the machine to predict its future states would be prodigious - but possible in principle.

BUZZ

Do you think computer programs are predictable in that way - in principle?

2. The design stance.

If you take a clock - a clockwork clock - there is an new way in which you might predict its behaviour. You could apply physics to it, and that might enable you to work out that in 15 minutes' time it will show 5 o'clock.

But you could make this prediction on the basis of knowing that that it had been designed to tell the time.

You would then be adopting the 'design' stance.

'The essential feature of the design stance is that we make predictions solely from knowledge or assumptions about the system's design, often without making any examination of the innards of the particular object.' Dennett, quoted by Stich, in Lycan, p.168.

3. The intentional stance.

But sometimes there is a third possibility.

Sometimes you can get good predictions by assuming that the thing is rational.

You might predict the behaviour of a plant by attributing to it the aim of getting maximum illumination, and the belief that keeping its leaves flat towards the source of light is the way to do this.

Dennett is now in a position to define 'intentional system'.

WHENEVER, HE SAYS, ONE CAN SUCCESSFULLY ADOPT THE INTENTIONAL STANCE TOWARDS AN OBJECT, THE OBJECT IS CALLED AN INTENTIONAL SYSTEM.

He then explains what he means by being able to adopt the intentional stance 'successfully'. This judgement is to be understood as a pragmatic one. The test is whether the system behaviour can be successfully predicted, 'and most efficiently predicted' ... 'by adopting the intentional stance towards them.' Whether the system 'really' has intentions, beliefs, thoughts, plans, desires, purposes or not is not the issue. The issue is whether by treating it as though it had these things, or things of this sort, you can predict what it is going to do. If you can, it is in Dennett's terms an intentional system. If you can't, it isn't.

By this test, some computers, good chess-playing computers included, are intentional systems.

STICH GLOSSES THE DEFINITION THUS: 'ANY OBJECT WILL COUNT AS AN INTENTIONAL SYSTEM IF WE CAN USEFULLY PREDICT ITS BEHAVIOUR BY ASSUMING THAT IT WILL BEHAVE RATIONALLY. ' STICH, IN LYCAN, 1st edition, p.168.

One and the same intentional system can be instantiated in many different ways.

Notice that an intentional system may be programmed or designed in a plurality of ways: just as a single programme may be run on a plurality of different hardware systems.

Different programmes can result in the same gross behaviour.

In order to treat a system as intentional, we have to attribute to it a large number of beliefs and desires - the beliefs and desires it would be rational for such a system to have, given its nature and history. But we need not assume that the beliefs and desires correspond in any systematic way to internal states, defined physically or functionally.

Dennett makes the point in terms of imagining three different instantaitions of one and the same intentional system: Mary, Ruth and Sally.

Mary: a person

Mary is a person. Ruth and Sally are two robots, each designed as copies of Mary.

Ruth: copies Mary's design.

Ruth is designed with the idea that if you are to get a robot to simulate Mary you had better reproduce in the robot what you know of how Mary works. You won't be able to use organic materials, but you will be able to attempt a thorough simulation with silicon chips of the functioning of Mary's organism, and in particular her brain.

If you get it right, Ruth will behave exactly like Mary because you have copied the design of Mary.

Sally: different design, but achieves the simulation nonetheless.

Sally, on the other hand, although she too simulates Mary, is designed completely differently. There is no attempt to copy the way Mary's behaviour is produced in Mary.

Sally's input-output pattern is the same as Mary's (and Ruth's), but the internal workings are different.

This means, says Dennett, that Sally is likely to be 'psychologically different' from the other two.

This may come out when something goes wrong. Sally's errors will be different from Mary's.

A parallel: different sound reproduction technologies

[Think of a cassette version of Motzart's Coronation Concerto compared with a cd. They sound much the same when they are working properly, but they go wrong differently, in ways you can tell.]

The concept of programming , as in programming a computer, is very helpful in seeing distinctions here.

If you think of Mary as following the behaviour she does because of the programme built into her, you can think of Ruth as equipped with the same programme. That is why she simulates Mary. They are not physically identical, but they are following the same programme. Mary may be a Mackintosh to Ruth's IBM PC: but both are running WORD 6.

On this analogy - it is of course not just an analogy - Sally is not running the same programme. She is running a different programme, but with the same overall observable effect.

If we think of each system transforming inputs into outputs, stimuli into behaviour, all three perform the same transformations, all three produce the same outputs from the same inputs.

Ruth achieves this by running the same programme as Mary. Sally runs a different programme, but one that achieves the same transformations.

The concept of the black box

If we talk of black boxes, all three are the same black-box:

input > > output

stimuli > > behaviour

This is like treating sound reproduction technology as a black box:

Birmingham Symphony with Murray Pariah

Wonderful noise in the living room

In the black box could be

an vinyl LP player or

a tape deck or

a cd player

[How many beliefs do you have? Roughly?]

Dennett's programme: to develop a concept of an 'intentional system', with the following role in mind:

to 'legitimise' mentalistic predicates

to explain the theoretical strategy of cognitive psychology and artificial intelligence

to achieve a reconciliation between 'our vision of ourselves as responsible, free, rational agents, and our vision of ourselves as complex parts of the physical world of science.' Dennett, quoted by Stitch, in Lycan, p.167.

Dennett makes repeated use of the illustration of the chess-playing computer (Stich, p.167).

'If one knows exactly how the computer's programme has been designed ... one can predict the computer's designed response to any move one makes. One's prediction will come true provided only that the computer performs as designed, that is, without breakdown. ...

END