Chapter 12

Belefs, arguments, facts, intuitions and hunches

"Yes", you say, "but I'm still not convinced".
That is a fine notion, that what matters is for me to convince you - as if thirty seconds of exposure to my argument should suddenly transform your convictions - and you not needing to make any effort.

If you are interested in the nature of the world, then surely what you believe is not important?
Will the actual structure of the world be any different as a result of you believing this as opposed to that?
I do not mean as a result of your actions - these clearly matter - no, I mean, at the moment of you holding the belief - does the world change in conformity - not as it appears to you - but as it appears to me?
So what does what you or I think matter?
There are arguments. For example, there are the ones just given. You are interested in the nature of the world, or say you are, for if you are not then there is no point in us talking like this, we might as well go and watch a game of football together.
The arguments do not depend on the identity of he who made them nor on whether he happens to believe them or not.
I give a correct argument but believe the contrary. Does that make the argument invalid?

As it happens, I don't believe a word of it either.

In this, my thoughts had taken a wrong turning. I was of the opinion that the reason why people are so wedded to the notions of induction and the telementation view of language, and the deep internal conviction that words are carriers of meaning in the telementation sense, was that these notions had an instinctive appeal which outweighed any arguments that one might put to the contrary. I now think that although there may well be a genetic predisposition to hold such views, their retention in the face of arguments against induction and telementation, is, if not justifiable then at the very least, understandable.

This brings to mind arguments I had had with a friend, late into the night, regarding belief and the need to become convinced of the "truth" or "validity" of an argument. The sort of situation where a believer and an atheist try to convince the other that he is wrong.

The point which I was interminably puting to him was that only the arguments mattered and that one's personal belief in their veracity was immaterial. My feeling was that beliefs stand in the way of determining the facts because these beliefs colour one's perceptions and prevent one from pushing an argument forward to its logical conclusion. Given the conclusion, one can re-examine the arguments and try to re-interpret the facts accordingly. As a result of that re-interpretation, one's beliefs may change. That is, one may obtain a new perspective and hence a new view. So whatever beliefs one had prior to examining the argument would be immaterial with respect to the beliefs that the argument eventually led one to have. Since these would change in their turn, beliefs are immaterial and argument is everything.

I still agree with the basic line of this argument even though it is now clear that one cannot be wholly divorced from one's beliefs. The point which I had missed entirely, was that there is any number of arguments that can be pursued to their logical conclusion and that it is sometimes necessary to stand out against the facts; that sometimes it is necessary to follow one's intuitions. This is a dangerous line of argument. It is not to be taken as advocating bloody mindedness as a substitute for reason.

The discussion centered on the ideas of Popper regarding "refutation" as a solution to the problem of induction, ideas we've touched on in previous chapters. I had recently taken the refutation idea and applied it to the problem of automatic speech recognition. At the time of our meeting, I had just, as I then thought, discovered the precise difference between my "refutation approach" and pattern-matching. As it happens, I had to wait a further six months before the idea crystalised sufficiently for it to be coherent on paper.

Now, I had had a terrible time of it, trying to show the precise differences between these two. I knew, that is, believed, that pattern-matching was useless as a recognition technique despite the fact of it having been the approach single-mindedly pursued by every research laboratory in the world these past thirty five or more years. I felt that pattern-matching should not work and even now, whenever I see a speech recogniser working, I still experience incredulity. Of course these machines work, after a fashion, or should I say, within limits - as will be the case with any recognition device. As far as I was concerned, pattern-matching was not even eligible for consideration as a recognition method practicable in the real world, and refutation gave something like a solution but was nothing like the sort of solution one really wanted. It involved generating too many hypotheses for evaluation and at the time, I could not see how this number could be substantially reduced. As a computer scientist, every bit of me rebelled against this phletora of hypotheses. It still gives me cold sweats when I think about it. Furthermore, I did not really understand the approach I had developed. It looked very different to patern-matching but every time that I tried to show this difference I would fall flat on my face. On each attempt I was forced to conclude that there was no difference between them whatsoever!

This situation lasted some five or so years. The problem was that my thinking was still steeped in the logic of induction and the notions of similarity and still largely is. The comparissons I was performing were steeped in the logic of the opposing view. No wonder I kept falling flat on my face. The reason why I could continue despite the seemingly irrefutable conclusion that there was no real difference between the approaches, was the existence of a deep belief in the correctness of the line I was taking.

If I had been true to the line argued above, namely that of disregarding belief and dogmatically rejecting the refuted idea 11.57, then, I would have been forced to dispence with the "refutation approach". On the other hand, my actions were not unjustified given the, to me, overwhelming nature of my arguments against pattern-matching. Furthermore, there was no question of me ever building systems on the basis of pattern-matching. It was so clearly the wrong way of proceeding that every cell of my being cried out against it. I had no idea which could serve as a viable alternative to refutation. Imperfect as it was, there was always the chance that there did exist a formal difference between the techniques if only I dug deep enough. Additionally, by pursuing the line I was following, I stood a good chance of discovering some alternative to refutation. For the situation where we could build things, despite the logical problem of not being able to demonstrate the veracity of any proposition, as argued earlier, was unbearable. (This is the great hope offered by inductive procedures - that given sufficient confirmation it should be possible to say of something that it is true). The world is real and it seemed inconceivable that we should be able to live in it if there were not some "truthful" correspondence between what one thought and what actually happened to be the case.

My friend's position was identical to mine though reversed. He could not, as I see it, accept refutation because it conflicted with the facts as he saw them. Induction is not satisfactory but accords better with the facts as he sees them. Whilst not being perfect, no more than refutation is for me, there is always the chance of something else turning up, just as in my case. And of course this is the only position one can take. One examines the arguments but in the final instance, life always entails choices, what line of thought to pursue. One cannot pursue every crazy idea that happens to come along. Time is too short. And at these points of choice, credibility, hunches, beliefs and the like are what will decide the matter.