Let's not waste time David and Sarah Kerridge This is a commentary on the ideas on the Theory of Knowledge which Dr Deming includes in his seminars, and "The New Econ- omics". Phrases in italics are quotations. It takes into account the things he says elsewhere, and the general Shewhart-Deming approach, rooted in CI Lewis. Systematic learning The Theory of Knowledge is concerned with learning quickly and efficiently. It can be summed up in the one sentence: Let's not waste time. Scientific method is the way we apply this theory. Of course, people learned things before they knew about science. They mainly relied on common sense, accidental discoveries, or just trial and error. It worked after a fashion, but it was slow and uncertain. The belief that madness was caused by the moon lasted for thousands of years: hence the term lunatic. Learning was slow and uncertain because it was unsystematic. The PDSA cycle is a system for learning. It depends on choosing a theory, using the theory to predict the effect of changes, and observing the success or otherwise of our actions. We then use this information to improve the system, and to refine or change the theory. Each stage of the PDSA cycle can be performed more effectively if we understand the ideas behind it. This means dealing in Philosophy, which most people do not think of as a practical subject. But the branch of philosophy we need is simply the study of how it is possible to know anything. So understood correctly, and used sensibly, it is highly practical. It may seem surprising, but managers need the theory of knowledge far more than pure scientists. The pure scientist in his laboratory uses advanced technology to observe under ideal conditions, ignores what he can not observe, and disbelieves what he cannot prove. The manager faces problems of far greater complexity, because they are system problems, and has to observe under the difficult conditions and pressures of ordinary work. What is more, he must not ignore what he can not measure, or can not prove. This is just to say that the manager must be responsible in a way the pure scientist is not. Theory This does not mean that we all have to become philosophers: It is not necessary to be eminent in any branch of Profound Knowledge..... All that is needed is a grasp of some straight- forward ideas. These are simple, but subtle, like everything else in the Deming Philosophy. The first thing is to know what we are talking about. A theory can be anything from a vague hunch to a clear prediction. Vague or definite, it always points to things not yet known, though it must agree with all that is known. The Theory of Knowledge is a theory in precisely the above sense. It predicts that by following its rules we will gain knowledge faster and more certainly than by not following it. The theory of knowledge does not deal in absolute truths. Ultimate reality is unknowable, by means available to science. There is no such thing as a fact. There are, however, things we can agree on, and things we can predict. The Theory of Knowledge deals with learning efficiently, and improving our ability to predict, so that we can achieve our aims with the least possible waste of time and effort. Explanation An explanation refers to the past: "That happened because..." It may suggest a theory, but is not itself a guide to action, and can not be tested. We sometimes have several different explan- ations, but the predictions they lead to are the same. It is a waste of time to argue about which explanation is best under these circumstances. Look hard for situations where the predict- ions are different. An explanation makes us feel comfortable with what we see: it suggests a "reason", or a pattern relating otherwise unconn- ected events. This is useful, because it is much easier to learn new things if we can see a pattern in them, and especially if we can see a relationship to what we know already. Perhaps our desire for explanation is a response to the threat of otherwise overwhelming complexity. But feeling comfortable is not know- ledge, and may hinder the advance of knowledge. The value and limitations of logic Logic and mathematics can not, by themselves, predict the future. They help us to see whether a theory or an explanation is self- consistent, or consistent with other knowledge. They can clarify a question, but not add to knowledge. Logical analysis can also help us to uncover the hidden assumptions that we make, when we put forward an explanation: but the explanation is not justified by these assumptions. In so far as any explanation can ever be justified, it is justified by the success of the theories it suggests. Logic and mathematics are also useful in showing the consequences of a theory: in formulating the predictions of the theory as precisely as possible. Definitions and meaning Knowledge is wasted unless we can pass it on. This usually means expressing it in words, or some common language of pictures and symbols. It is a waste of time to argue about the meaning of a word: we should agree on a meaning and stick to it. When precision matters, the meaning must be defined in terms of action (Operational definition) as this is the only way we can test whether we agree on the meaning. We should not define to greater accuracy than is useful for the purpose in hand. Don't split hairs - we may fossilise ideas. Prediction is the test of knowledge It is a waste of time to argue about anything that does not affect future action, or a theory that can not be tested by observation. A theory can only be tested when it makes a prediction that could in principle be proved wrong by an observ- ation not yet made. Explanation is comforting, but prediction is disturbing, because we may be proved wrong. Only by taking the risk of being proved wrong can we make progress. However many times a theory predicts correctly, we can not say it is "true". It is only useful, as a sound basis for action. There is no knowledge without prediction: but the purpose of knowledge is action. So a useful prediction is one that enables us to say that one action will be more successful than another. This implies that there is a value system by which success can be judged. There is no point in arguing about values, they are a personal matter. The most we can do is to avoid inconsistency in applying them. The aim is a value judgement. This applies just as much to our system for learning, as it does to any other system. Stability Prediction is meaningless unless we know what makes the relation- ship it depends on stable. Pure scientists concentrate on inher- ently stable processes in chemistry, physics and biology, and call these "fundamental". The applied scientist must achieve stability of the conditions that affect a particular prediction. It may not be necessary or possible to achieve stability of the whole system. In fact, no system is ever completely stable. It is sufficiently stable when the predictions we make work well enough to be useful. It is a waste of time to use an unstable measurement system to test a theory. Improvement of the measurement system to the point where the errors in it are small compared to the errors in the system under test is vital. Selecting a theory Prediction, and the rules of operational meaning, are only one side of the scientific process. They tell us how to test, and so improve, a theory. The other side of the theory of knowledge deals with insight: knowing which questions to ask, and hence which theories to test first. Here are some guidelines: Test the simplest theory first. This is not because it is more likely to be "true", but because a simple theory is easier to use. A simple theory is also easier to communicate, to disprove, and so improve. Every theory has a limited range, or domain within which it applies. You do not understand a theory until you know when it will not work. Study extreme cases to find the limits and increase knowledge. A theory with a wide domain is a better starting point for testing than a localised, special theory: sometimes even if the local theory gives more accurate predictions. This is because a theory which links previously separate things, like Newton's theory connecting falling apples with the motion of planets, generates new opportunities for testing, and so leads to faster progress. This still leaves a wide choice of ideas that could be tested. We can avoid wasting time by learning to ask the right question. The questions are more important than the answers. Learning to observe. We can not observe without theory: we do not know what to look for. An operational definition must be designed to achieve sufficient agreement for a stated purpose, and this purpose implicitly relies on theory. But more generally, we must learn how to observe, to know which things to ignore and which to concentrate on. This is why any new situation is so confusing. The following story illustrates the idea of learning to observe. Charles Babbage, the first writer on computing, and a founder of the Royal Statistical Society, visited Herschel, the astronomer. Herschel asked if he had ever seen Fraunhofer's lines (the dark lines in the spectrum of sunlight). Babbage said that he had not, but would like to. Herschel said "I will prepare the apparatus, and put you in a such a position that they shall be visible, and yet you shall look for them and not find them: after this, while you remain in the same position, I will instruct you how to see them, and you shall see them, and not merely wonder that you did not see them before, but you shall find it impossible to look at the spectrum without seeing them". Babbage records "The prediction of Mr Herschel was completely fulfilled." Learning to observe applies even more strongly to patterns that the mind appreciates, than patterns that the eye sees. The Deming Philosophy trains us to "see" things that were always there, but were invisible to us. Overcoming our psychological limitations One way to waste less time in finding the right question is to understand the ways in which our view of the world is warped. Usually we oversimplify. The four aspects of Profound Knowledge illustrate this. Systems We get the things closest to us in time and space out of proportion. "Out of sight, out of mind." Hence short-term thinking, and sub-optimisation. Hence also the neglect of the unmeasured and unmeasurable. Variation We think we can explain variation, even when it is common cause variation. We imagine that trends or patterns, even when not random, are more stable than they are. Our mental judgement of probability is very bad: we worry unduly about unfamiliar small risks, and ignore familiar large risks. Knowledge We always imagine that we know more than we do: that what worked in the past will work in the future, even though things have changed. We also tend to avoid thinking about anything we can not explain: often to the extent of not seeing it. We can learn even from a wrong theory. A wrong theory, used in a disciplined way, forces us to look at things afresh, and so learn. Any theory, even a right one, is dangerous if we use it blindly. Psychology We forget that other people differ from us. We are quite unable to be objective about ourselves. Fear and "incentives" distort every perception. More subtle is the effect of the system we are in. A system cannot understand itself. But when we are inside a system, we are part of it, and cannot understand it. This is why Profound Knowledge has to come from outside. Insight is more important in applied science than in pure science, for two reasons. In pure science a wrong theory will always be corrected eventually, as every other scientist will be delighted to disprove it. The manager has no check of this kind. Even when an idea fails, it does not mean the idea was wrong: often outside circumstances overwhelmed it. Alternatively, with some change in the way the idea is made operational, it may succeed. A new idea is always at a disadvantage compared to an old, well tried idea, so we must be prepared for initial disappointment. Another reason why insight matters more in business than pure science is cost. The costs of being wrong are higher, and the costs of investigation are higher. Finally, the costs of waiting for the right answer to emerge are much higher. Applied science deals with immediate,as well as long term needs. Insight only enables us to see which theory to test first, or how best to test it. It makes the PDSA cycle work better, never replaces it. Developing a balanced view We must learn to balance reductionist and holistic thinking. Pure science is all reductionist: chopping knowledge into easily studied separate pieces. This is why almost all scientists are so specialised. Our need to increase our ability to think about complete systems must not blind us to the value of reductionist thinking in the right place. For example, in the PDSA cycle, we first test an idea if possible on a small scale which is in a sense reductionist. Then we Act on the whole system as something separate. Another example arises in SPC. Where possible, we move upstream to study the process, not the outcome. This too is reductionist, because we are assuming that an improvement in one part will improve the whole process. But we check, whenever we can, that the whole system benefits. Perhaps we should regard system thinking not as the opposite to reductionist thinking, but as concerned with the correct balance and interplay between reductionism and holism. Overcoming defects of the education system Many wrong ideas stem directly from good scientific train- ing, applied in the wrong place. In experimental science theory is tested, not used. So anything that can not be used to test theory is ruled out. This means that the "scientific manager" is suspicious of the unmeasured and unmeasurable, and concentrates on "objective" final outcome, that is the "bottom line". Unfortunately "bottom line" figures, however concrete, usually come from a highly complex and unstable system. This avoids one trap and falls straight into another. The rule is: Test theory whenever possible: that is, when the system is reasonably stable and results measurable. Use theory so tested in all other cases. Other examples of bad effects of education are that we have been trained to think that the real world is "continuous", that small changes have small effects, and that everything has an identifiable cause, if we look hard enough. The fashionable subjects of "Chaos Theory" and "Catastrophe Theory" are valuable because they release us from this unconscious mental trap. It should have been obvious, to anyone who deals with the real world, that catastrophes do happen, and that chaos is usual, but now pure mathematics has made them legal. Other modes of learning The PDSA cycle is not the only way to learn. We also learn by monitoring a system. This means observing and recording in a disciplined way, often using a control chart. Here we have no specific theory in mind, nor are we experimenting. We are waiting for the inevitable accident which will force us to see something new about the system. Another mode of learning is by stressing a system. We push things to the limits, not because we believe that the stressed system will be better, but because we do not know in which way it will fail. When we discover this, it adds to knowledge. The pure scientist does this by examining his theories under more and more extreme conditions: hence the need for the massive and expensive equipment of particle physics. The idea of stressing the system is less often applied in management, but the Japanese certainly use it. They sometimes deliberately speed up a production line until mistakes occur. Then they analyse the mistakes, and see how they can be pre- vented. Mechanical systems are often tested under far harsher conditions than they will meet in practice. Testing to destruct- ion under extreme conditions is called "accelerated life test- ing". This is often just a quick way to estimate the life of a component under normal conditions, but can also be a good way to increase reliability. Why is it difficult to be reasonable? All of us know how difficult it is, at first, to understand many of the ideas in the Deming Philosophy, even though they are so reasonable. Insight develops slowly. But perhaps the difficulty is that the ideas are reasonable, where we are not. This is not a comfortable idea, but is in line with a great many well established findings of experimental psychology. A recently published book summarises these findings. "Irrationality" by Stuart Sutherland, Professor of psychology at Sussex University: published by Constable, November 1992