Can We Know Too Much?
by
Gernot M. R. Winkler

Introduction.
                               
It is usually asserted that too much knowledge has never hurt anyone. My thesis is that this is not true. That it is a superficial, indeed a seriously misleading belief. Too much knowledge, if not matched with the skill of how to apply it, causes many disasters. It is a chronic problem of a large number of people who have been unable to complete their education in the proper meaning of the term.

It is interesting to note that in the old German General Staff, an academic background such as a PH.D. was extremely rare because it was assumed as a matter of common sense that too much learning would be an impediment to quick decision making. Advanced intellectual training was, for this reason, considered a hindrance and such persons did not well in the promotion process. Of course, it is evident that if you know little, and have no idea of the range of possible consequences of your decisions, you will decide easily and quickly. If, e.g., you have only two possible choices, your chances of  doing the right thing are still better than 50%,  because on the average, with some experience and any valid information input, the result should be better than a random decision by throwing a coin. The question is how much better than 50% we can hope to do -  and at what price? On the other hand, if you get less than 50% success, this would reveal that you are living with wrong assumptions which bias you towards error.

However, with more knowledge and more foresight for possible consequences,  more and more time will be taken and the decision maker will eventually suffer agonies of doubt unless, at the highest level, he has developed the insight, experience and discipline of the master. He will also be extremely cautious (the caution of the old, the tested, and the chastened). From Napoleon's point of view [1], the German military bias against high intellectualism would have been a mistake - they confused time for decision making with procrastination in the execution. But there is more to consider because we need also firmness in the execution, which requires strength of character. Regarding this, if we assume as a “typical” PH.D. someone with a high degree of abstract learning, as he used to be known as “Egghead”,  the German military had their doubts. Have they been justified? We believe that they had a point.
For Good Decisions, one can consider too many details!

We question why one can know too much - which is not a trivial question! The surprising answer has been demonstrated variously, e.g., by the results of prediction  in the practice of psychology. As much, if not more than in other sciences, one investigates the validity of the predictions that are being made on the basis of tests and it turns out that estimates of future academic success made on the basis of four predictors have been consistently better than those made with twenty-two predictors which, presumably, used much more information (Bartlett & Green (1966), Clinical Prediction; does one sometimes know too much? J. counsel. Psychol. 13). These findings seem to go flagrantly against common sense and also against the belief that one can never know too much. However, the results of the research just mentioned are in perfect agreement with what one should expect on the basis of a most general principle, Ockham's Razor [2]! Indeed, in extensive computer studies, the statistical predictions on the basis of the simplest algorithms have been found better than the results of the most complicated (Makridakis & Hibon, 1979, Accuracy in Forecasting: an Empirical Investigation. J. R. Stat. Soc. Ass. Vol.142, Part 2, pp. 97 - 145).

A mind is, however, not a computer and the question must be addressed how can too much information impede the decision making process? Is it not important to enlarge the Universe of Thought as much as possible? This “universe” is what we keep in the back of the mind regarding possible consequences, and with more ramifications that we can envision, we will be more careful. The answer is, of course, yes, we should have as much background information available as feasible. However, there is a necessary condition attached to this requirement: We must also have worked out a scale of values against which we set our priorities. Not all details are equally important, or may remain important in the future, and not all can, or should be considered. Usually we also face conflicting requirements, therefore we must find the order of their importance, and if we can, with  numerical estimates or weights to assist us with the decision making. Unless we can do this, too many data, indiscriminately considered, will seriously mislead the unwary.

In this case, we fall prey to the confusion that is inevitable if we try to handle too many items in the mind at the same time, without following Descartes' advice to insist on “clear and distinct ideas”. Jugglers are limited by the number of pieces they can hold in the air. The brain has a physiologic limit when keeping itself conscious of different thoughts. We have a short time window of comparison (“at the same time”) and for most of us, the above number will be much smaller than seven. Disregarding this creates a great problem and is the main reason why highly intelligent people can commit the most outrageous blunders. They disappoint us when they come up with stupid judgments which they probably have based on confusion of some unimportant or ephemeral detail (if not personal bias due to some doctrine), while simple persons maintain more easily their common sense.

In addition to confusion, failures in complicated judgments are therefore due to the absence of a Scale of Precedence in the mind. The importance of details must be kept firmly in the mind. These failures are the major reason why so many of the intelligentsia are failing in their proper function. Instead of teaching the eternal values of mankind, i.e., truth, justice, and objective reason - which would be the right mission that only they can do  -  they like to engage in all kinds of pet ideas and ideologies, leading even to deception if it benefits their simplistic  “cause”. This narrowness of  interest has been noted several times.  Julien Benda called it in his great work The Betrayal of the Intellectuals (La Trahison des Clercs, 1927). Infatuation with simplistic ideologies that push utopian ideas is an endemic problem among poorly educated “intellectuals”. Hence their uncritical love for all things left or right. With these various extremists, the motivation may be different, but the crippling intellectual weakness, i.e., the lack of a sound scale of Precedence, and the acceptance of vague, confused ideas, are the same.

The notorious judgment call.

In the execution of a project one depends on organizations where a multitude of secondary decisions have to be made. Every one of them will be made by someone who judges the situation and, based on his judgment, decides what is to be done on his part, how - and when! This is necessary even in the typical bureaucracy where much time has been spent on writing regulations that supposedly will remove the need for judgments by untrained people. However, this does not take care of this problem - it makes it much worse because the real problem is now obscured. In war, of course, the problem is immensely magnified and the life of many depends routinely on the more or less intelligent decision of individuals.

These people, however, whose individual judgments are so important, are not always sufficiently prepared for their task. Many are overwhelmed by the information and are not used to manage this. Moreover, since it is not a personal affair for them, they will not be driven to use the utmost care in their thinking. Being vaguely aware of their weak judgment, they will cling closely to whatever rules they can construe as applicable, and we get the typical bureaucratic performance by blind regulation; a regulation that becomes inevitably more detailed and rigid over time. Moreover, excessive reliance on the system leaves the workers less involved with their work. The worst failure in decision making happens, of course, if his weak character leaves the decision maker vulnerable to motives of possible personal advantages that can be influenced by his decision. Such advantages can look quite innocuous, such as coauthor ship of publications, etc. We can expect that in such cases, the decision will most likely not be objective.

Given this situation, what can be done about it? People will find some comfort in the pious hope that one can rely on common sense and that common sense will prevail. Well, good luck!  But is it not much better to take up the task of training people and preparing them for correct reasoning?  If it is done in a systematic way and regularly, it will over time, have a profound effect: Not only by producing better decisions;  it will make people more aware of their responsibility!

Anyway, it will be wise to think about training people in the art of decision making: to do it in a more formal, fully conscious way and not quickly “by the seat of the pants”. From the comments made by those who had to explain after a disaster what they have done and why, it is clear that most people are fuzzy in their mind about the need for clear criteria (and which criteria?) as guides for their judgments. They are also not aware of scales of precedence (which is perhaps even more important) to be used for weighing between conflicting factors. In other words, it is necessary to prepare and train people in the important art of Judging the Situation and Decision.

Spell out the values and precedence to be used in decision making

Depending on the level at which people operate, their decision making training has to be more general or more specific;  more sophisticated or simple. To demonstrate it, we take an extremely simple example which should make things obvious. It will be up to you to translate your requirements to take the same steps as in the example.

Suppose you have to train a person to procure brandy for a restaurant chain! If you take any unprepared apprentice, he or she will judge a sample most likely by the strength (proof) of it and will simply not be aware that there is a world of taste sensations that needs to be probed. To make our job of demonstrating easy and convincing, we should use a test set-up with three or four samples of brandy of various quality and with their identity hidden  Now, after the first round of tasting, you let the applicant put his (her) judgment about their findings on paper.

As the second step, you explain the tasting process: The different tastes to be experienced, the softness, the aroma and its strength, the temporal sequence of how fast the taste dies out, whether it is interrupted or lingers on, a possible after-taste and where in the mouth the taste is sensed, etc.; -  all this has to be conceptually separated and elevated into full awareness. In other words, you alert the person to the expected variables and in this step you establish on paper the value factors and their importance (precedence). Then you repeat the tasting with a new set of samples, again with hidden identity. The result of the quality judgment that the applicant makes, after the criteria and their precedence have been given, is an indication for the decision making capability of the applicant.

In our case, the training can be specific and matched closely to the requirement of the envisioned job. A test and training for a more general judging capability is more extensive and more difficult to prepare. It can be sophisticated with numerical value and probability assignments and other details of decision theory (with utility functions and various tools of Statistics) but it must always have the same essential features:

        First step:  Let the applicant make a judgment without giving the criteria.
                        The value of this step is in convincing him that he needs to learn.

        Second step: Explain the detailed criteria and their precedence on paper.

        Third step: Repeat a test and have a brief written report prepared.

        Last step:   Evaluate and discuss the test results in view of the requirements.


Conclusions.

What should we conclude, should we stop collecting more data to avoid confusion; is it wasteful? No, not at all. We must have as much information as we can get with the available resources because how else could we decide what is most and what is less important, what is permanent and what is only a transient? We need the information to make effective use of Ockham's Razor [2]!

What all this really means is that, in the preparation for a critical decision, the important point is not so much to gather as much information as possible and to exhaust your resources just in doing this, than to make sure to recognize what is important and what is less important. In other words, we must not only be made fully aware of, but must be trained to become experienced in the sifting of concepts and the use of a scale of values and precedence.



Notes and References.

[1] Napoleon writes in his Memoirs: Il faut être lent dans la déliberation, et vif dans l'exécution. To be quick in execution could hardly have been said by a greater authority. But slow in the deliberation? Today, a hesitating executive would be accused of procrastination by people who have never felt the heavy weight of responsibility. Procrastination is a frequent problem, indeed, but some of it is highly beneficial! To find the best solution to a problem can be so very difficult because it is extremely hard to be completely objective, as we can see every day in the media. Even without the problems of bias with distortion of our judgment by emotional and character factors, (and firm pre judgment on top of it); -  deep understanding requires very much time.

Therefore, during deliberation it is advisable to take as much time as possible, not just to avoid hasty decisions, but to have time to ruminate the problem thoroughly. We must ponder, and reconsider until we understand. Some minds find it conducive to think in company, with advisors and friends. Others do not. Indeed, discussion is likely to bring into consideration more viewpoints and details and for this reason some discussion should be included. However, once the factors have surfaced, undisturbed solitude while strolling around will allow even the most subtle considerations to come into their proper weight. Unless there is real urgency (which, of course, is usually the case in combat), and if we can afford it, the best idea is to wait until clarity is reached, until the decisive factors have revealed themselves, or until the problematic situation has changed or gone away for extrinsic reasons.

The second part of Napoleon's doctrine, to be quick in the execution, is generally considered to be a major reason for his military successes, and indeed, it is important to be fast everywhere! For various reasons, most of them really weak excuses, America seems to be in the process of forgetting the value of time (or suffering from an epidemic of "tachyphobia"!). There are great exceptions, but one cannot doubt that on the average, things are going more slowly today than they were forty years ago. One could think that the imposition of the federal 55 miles speed limit in the seventies had something to do with it, but this was only one of the symptoms for the dropping values of time and speed versus a host of other issues. Anecdotal evidence can be seen everywhere.

It is ironic and significant that the disregard for the speed principle started with Napoleon himself! Napoleon's greatest blunder which sealed his fate, was that he did not follow his own prescription: be fast in the execution. During the decisive battle at Waterloo that cost him the empire, he delayed the opening of his attack on Wellington from morning until midday, reportedly to allow the ground to dry. This is difficult to understand and appears as an excuse for indecision, but the delay allowed Blücher's distant troops to reach Waterloo and support Wellington! This turned the tide, and Napoleon's forces were routed by the combined allied armies.

[2]  William of Ockham (d. 1347) stated the doctrine that entia (principia) non sunt multiplicanda praeter necessitate, i.e., one must not introduce more concepts (abstract entities, details, elements, assumptions) in theories than necessary. This "Razor" (to separate the unimportant from the important) has been extended by Ernst Mach in his famous, but misleading, principle of "Economy of Thought" as the criterion for a good theory. George E. P. Box called it the "Principle of Parsimony" in statistics.

In its formal generality, Ockham’s Razor is a principle of the very highest importance - it should be a fundamental strategy for all intelligent analysis: Use as few assumptions, as few concepts as possible in your models to be safe on the average. The inclusion of more details than absolutely necessary renders a theory increasingly vulnerable to error because some of these details will very likely be transients or purely accidental because of your limited information. Of course, in the case of data modeling, within the range of the available data, a more complicated model produces a better fit with smaller residuals, but the price that one pays is a reduced confidence in extrapolation (over fitting). In each case, there is an optimum number of details that must be considered for acceptance (e.g., terms in a polynomial fit), but we do not know reality sufficiently to be able to decide what will be of importance in the future, in other words, what is systematic as opposed to what is temporary and merely accidental. In any case, the doctrine advises us to be on the safe side. The eminent British savant Sir Harold Jeffreys states in his noted work on Probability that  . . variation must be taken as random until there is positive evidence to the contrary. Therefore, robustness is our aim and not, as it was suggested by Ernst Mach, economy of thought,

Say, you want to predict the decision of an acquaintance. You do not know him well, but have met him and gained some idea of what kind of a person he is. It would be a mistake to be too much influenced by a single act that may be an accidental reaction to some one-time stimulus. Important are his character, values, beliefs, habits, inclinations and principles for action - data that must be derived from his past. These “patterns” are more reliable predictors for his future behavior. Understanding and, based on it, good prediction depends on the ability to distinguish the systematic from the accidental happenings, the features of lasting importance from the transient effects.

Wm. Edwards Deming, the great American management expert who was instrumental in bringing Japan's industry up to world standards after WWII, placed great emphasis on the need to avoid treating accidental failures as systematics, i.e., as something which would cause the manager to take corrective action - for a failure which is purely accidental and not a lasting problem in the system. Such an action Deming calls tampering. The opposite error is in ignoring systematic problems, i.e., treating them as transients (or harmless), and doing nothing. We may call this an error of ignoring.

While the correct diagnosis must be our goal, with limited experience we can only obtain suspicions, at most probabilities, and it will not be easy to decide right away if a trouble is transient or systematic. We need time (more data) to decide this. If we apply Jeffreys’ advice in our situation, we should lean toward the second error as the less expensive on the average, i.e., temporarily ignoring until we know better what the situation requires. (Of course, the proverbial bureaucrats are extremists in this regard). Otherwise, by tampering, we disturb the system. But, try and tell this to the eager beavers (also known as castor studiosus, with a favored habitat around Washington, DC)!  Something must always be done right away! If you do not want to do it, it might be wise to serve this need in some other way.

During the 1992 election campaign, the U. S. Government took a waiting attitude regarding the economy because several data suggested that the economy was already on a rebound, as later data showed that, indeed, it was. But this was not the picture as seen by the public, and the Executive's and the Federal Reserve Board's reluctance to act on the system were seen as insensitivity. It probably influenced the election ("It's the economy, stupid!"). However, since overheating was avoided, it allowed a longer than usual period of stable economic performance and sustained growth. A safer strategy would have been to do something innocuous to avoid the appearance of insensitivity while in effect not tampering.

A special kind of tampering is an organization wide regulation, if it is issued because of an individual misdeed. This happens almost as the rule in organizations that are managed by insecure bureaucrats. But a personal problem is not a systemic (system wide), or a systematic (lasting) failure, unless it happens by several individuals, indicating that the wrong people are being hired, or that some necessary control is missing. These things are lasting and must be corrected. Otherwise, minor events must not cause a systemic reaction, but need an appropriate personal disciplinary measure. By changing the system because of a single problem, we needlessly introduce a disturbance and the likelihood of additional trouble or, as a minimum, we add overhead. In electronic systems, a component failure requires exchanging the component, but we do not immediately change the design of the system. However, if it happens several times we have to look for a different type of component (or a different supplier), or avoid its use by changing the design.

The Razor tells us to avoid splitting hairs (not to go too far in abstract analysis), because if we do, we increase the likelihood of becoming totally unrealistic. Highly  abstract philosophical systems originate this way (and also some of the super advanced ad-hoc theories in “ironic” science, a term coined by Horgan in his The End of Science),  which is why these great activities of the mind can acquire a bad reputation. Therefore, we should take Ockham's Razor as a key guiding principle for the practical problems of decision making. Even the military have adopted the razor tacitly in their KISS recommendation (Keep It Simple, Stupid!). In our times of information overload, the Razor and the lesson sketched above, assume a vital and increasing importance.

We must stress the great importance of the Razor  because its vast relevance is not sufficiently appreciated. It is seen merely as a medieval curiosity. An example for this is that for superficial reasons, it has been deleted as “unnecessary” from the otherwise excellent listing of "What Literate Americans Know.”  This happened in the Vintage edition of the highly recommended work by E. D. Hirsch (1988), Cultural Literacy, Vintage Books, Random House, N.Y.  It is advisable to mention this work here in the present context because it addresses the opposite (and today more frequent) problem for the mind: not an excess, but an insufficient set of concepts - which is caused by a deficiency in education..
Copyright © 2003,   Gernot M. R. Winkler          Last Correction:     05/14/2009