Behavioral science research; Computers as simulators

The following program is produced by the University of Michigan broadcasting service under a grant he made from the National Educational Television and Radio Center in cooperation with the National Association of educational broadcasters computers as a simulator. It's a program from the series human behavior social and medical research produced by the University of Michigan broadcasting service with special assistance from the Mental Health Research Institute of the University of Michigan. These programs have been developed from interviews with men and women who have the too often unglamorous job of basic research. Research in medicine the physical sciences social sciences and the behavioral sciences locational you will hear what may seem like strange or unfamiliar side. These are the sounds of the participants office is a laboratory or clinic where the interviews were recorded. The people you will hear today are Professor Russell a cough who is director of the operations research group at Case Institute of Technology in Cleveland and a

portion of a roundtable discussion with a group of experts from the Mental Health Research Institute from the University of Michigan campus. And my name is Glenn Phillip. Computers and simulation are imposing words by themselves. Put them together and one feels a bit overwhelmed. However during the following discussion we hope to be able to shed some light as to what takes place and how the future use of computers will greatly aid scientific investigation. What then is meant by computers as a simulator. Professor Russell takeoff of Kay's Institute of Technology Explained I think it's necessary first to understand what we mean by simulation. And this requires first of understanding of what we mean by a model as this this concept is used in science as a model as a reputation as a representation

of a process or a phenomenon that's under study which has the characteristics that it can be manipulated by the investigator so that he can ask questions really modeled it. Time might also be asked of the real thing but which he might not be able to get out of reality. Not the model as it normally exists and science is a static representation. It does not behave it's merely there and we ask questions of it. So recently it's become increasingly obvious that we want to study the way the various types of systems and phenomena behave over time. And as a consequence we become increasingly interested in representations which behave like the things which they represent in effect models which imitate the things that they represent. Now such a model one which has

behavior that has properties in common with the thing that it represents is a simulation of that thing. What we have learned to do with computers essentially is to program them in such a way that they will display characteristics similar to those of various types of systems which we encounter in reality. And as a consequence we can address to these representations are imitations of reality questions and get answers whereas in reality we're not free to manipulate the system so as to observe it under controlled conditions. Professor Raycroft used to more turn program and programming and these are two more terms which leave many of us wondering then what does one mean when using these terms. Well a program is an explicit set of instructions to something as stupid as a computer. This set of

instructions must be so complete. Someone ambiguous and so exhaustive that in any situation which might arise the computer has an appropriate instruction on which you can call which tells it exactly what to do. So in a sense then the program is simply a very complete and effective set of instructions. I asked Professor a cough what he felt was the greatest use to which the computer has been put. This was his answer. Well recently I think there have been two types of development in the use of the computer. One in what we might think of as basic research in the other and applied research areas in the basic research area. We're making progress in extending the class of problems on which the computer can be used for obtaining satisfactory solutions to fairly recently we've been

restricted in the use of the computer in problem solving to problems which could be analyzed sufficiently so that we could develop equations which essentially represent the choice situation confronting a decision maker and the consequence of these choices. The types of outcomes that are possible. The values associated with these various outcomes. But there have always been a large class of problems which are so complex that we've never been able to completely specify their structure in either one or a set of even very complex equations a simple example of such a problem would be the problem confronting the chess player in a game of chess. As a consequence it has been difficult by the use of the methods that have been available to us in the past to use computers for solving problems of this sort. Another example are the problems which arise in

geometry are in logic where we're given a set of axioms and postulates and are asked to derive certain theorems up to date nobody's been able to develop an explicit procedure which would when applied lead someone directly in the shortest route from a set of axioms and postulates to the desired theorem. There's a process of trial and error and exploration that's necessary. This type of problem solving has been called your mystic problem solving and recently as a result of work that's been done at the rain Corp and Carnegie Institute of Technology some considerable advance has been made in programming computers so as to solve what are called unstructured problems specifically the computer has been used to the duces the erms in certain classes of logic particularly the logic that's presented in Principia Mathematica. It has also been used to play a reasonably good game of

chess and is being used increasingly in problem areas that cannot be adequately treated by a normal mathematical methods. On the practical side I'd say one of the most exciting advances and uses of the computer has been a simulation of large complex systems because we can get adequate representation of systems inside a computer. We can conduct experiments on these large systems vicariously. Whereas before we couldn't manipulate the real system. So as to conduct experiments and learn about a structure in the way that it operates as a consequence today by the use of the computer and frequently in combination with human beings as in the procedure that we call operational gaming. It is possible to take complex systems and collapse the time required to observe their operation and study their

operation over a period that would be extended in real time but is actually quite short in simulated time. For example this is being done quite extensively now in studying the management of large industrial enterprises or the operation of complex weapon systems. Wondering about some everyday problems perhaps a political election. I wondered if it would be possible perhaps to predict certain trends. To this question professor a cough replied. Well I think it will be in fact I think at least as far as political elections are concerned. We've had some demonstrations that it's possible. But I don't think the critical question hinges on the concept of computation but on the concept of prediction the computer will only do what we tell it to do. The difficulties at present are developing the adequate predictive procedures. Which once developed can almost certainly be put on a computer. There may

be limitations in the computer in terms of the size of its memory its capacity to retain data and the speed with which it will operate but these are mechanical or electronic problems which I'm sure will be overcome. The much more difficult problems involves developing a sufficient understanding of the social processes to be able to represent a meaningful meaningfully in equations which are related to the structure of these processes so that we can get effective predictions under changing conditions. In the pure mechanics of operation what must an individual do in order to prepare the facts and the data in order to feed them into the computer. Well because of the complexity of getting the facts into the computer are making them available inside the computer. The problem is what the computer should do with these facts once it has them. You have to really step back a bit and ask the basic question of what facts are relevant for the

prediction that you've got in mind and this already presupposes a great deal of analysis. The concept of the relevance of facts involves one of two types of approaches to the problem. Either you have to understand why certain things happen to be able to explain the interactions between events and objects and develop a theory which explains certain events such as we have in the physical sciences commonly but very young commonly in the behavioral sciences. If you have such a theory then the theory itself tells you what kind of information you need in order to predict an outcome in complex areas of social and human behavior. Normally such theory is lacking and as a consequence we have to resort to a different type of prediction. Normally what we do is use intuition and collect information on a lot of related variables or variables which appear to be connected in some way with the event that we're trying to predict. And by

studying the past and the use of methods of regression and correlation we try to find out what combinations of variables tend to move in the same direction or in opposite directions that we use essentially look for signs or indices. And so for Sometimes we can predict events even though we can't explain them because we know that one thing is likely to occur in combination with others or after something else just as we say smoke in the sense predicts fire. So certain things which happen in the political arena may predict which party is to be elected in a given election. But the difficulty again here comes in to watch done before that you were ready to use the computer in determining what data is relevant and then secondly what has to be done with the data. Once you have collected it assuming that it is relevant. The idea of computing all facts lead some to feel perhaps that human minds would not be needed for decision making that the machine would

make these decisions for us. Isn't it impossible to ever conceive of doing away with human decision making. This was the reply of Dr a cog. Well I haven't heard anybody seriously propose that the computer was ever going to do away with in the sesame for human decision making. I certainly think it's true that a large volume of the types of decisions which are currently being made by human beings can even now be better made by computers. If they're properly programmed. But this is only true because at least some people in society know how to make these decisions better than most people actually are making them. And we can tell the computer to make them in some desirable way. But every time we've developed a procedure for solving one class of problems by the use of the computer we simply free the human being to turn to another class of problems which he's been ignoring because of the pressures of time and the lack of the leisure necessary to analyze and come to

understand these other classes of problems. I don't think there's any danger that the AI computer is going to replace the human being to the contrary I would say he's going to elevate the human being in the sense that the first industrial revolution resulted to a large extent in the replacement of human beings as a source of energy. I think the use of the computer and the second industrial revolution will replace the human being as a source of control of the machine processes and as a consequence this will all of a to human being into essentially a programmer for routine types of decision making. And this will make intellectual demands on him far in excess of the ones that are currently made on him. Will a greater degree of intellect be needed in the future perhaps than it is even now required as these comments would indicate. Professor takeoff said only that again this is simply a continuation of a trend that's already

begun. If you compare the educational requirements of the manager of an industry or an industrial or a military general today as with what was required of them 25 years ago you find it tremendously increased technical demands are made on and he must know a great deal more about economics about finance about psychology about the nature of the physical process that his company is involved in or in the structure of the weapons with which he deals if he's a general. And this. Continuous increase in the level of education and training required on the part of people who are responsible for the conduct of organized behavior is going to continue although I think it will probably be considerably accelerated by the development of the computer. The question of ethics often arises when small or selected groups reach decisions for the masses. I asked Dr a cough if he felt there were questions of ethics

involved in the philosophy that some people can make decisions better than others. He said Well there are but these are certainly questions the recognition even that a complete democracy of the existence of experts. This is a is an admission that certain people are better equipped to make certain classes of decisions than others are. As a consequence the medical profession has certain progressives with regard to decisions concerning health for example inoculation against certain classes of diseases which the average person isn't allowed to object to. Similarly legal decisions are made of this type. And decisions concerning the structure of a house the individual man who's building a house isn't free to decide however he wants to how to build it. He must make his decisions if he insists on making them conform to criteria which have been established by experts in this area. All I'm suggesting in my earlier remark is that there

are experts on decision making processes in areas which have not been generally acknowledged and that with the development of the computer. These types of expertise are becoming more important. And as a consequence I think we'll have the development of larger and. More significant groups of experts operating in society in order to arrive at solutions to this and to other major scientific questions. I wondered if Dr. A cough would agree that more interdisciplinary study would be required in the future. To this he replied. I certainly think so I think we're in an era scientifically that might well be called in the future of the era of interdisciplinary research. I think that the Advancement of Science by the increasing specialization of scientific disciplines and by training people to know more and more about less and less. But this development has

slowed down considerably and is incapable of handling the kinds of problems with which society is most concerned today. Reality doesn't divide itself conveniently into the various disciplines which universities found it convenient to divide themselves into for administrative purposes. The problems aren't physical chemical psychological biological economic or any of these things they're problems and the various disciplines simply represent different ways of looking at them. It turns out the problems that are most important today can be studied simultaneously by many of these different ways. And the critical research question is which of these ways or what combination of them. That will give us the kind of knowledge that will allow us to gain most control over our environment and the events that occur in it. All the indications to date are that the greatest progress can only be made by fusing concepts and techniques developed in the various specialties of science.

The one big advance that we've made it seems to me in this direction in the last 25 years is that whereas 25 years ago people talked about the scientific generalists that is training a single man who would have knowledge of a number of different disciplines and who could do this job by him self today we realize this is impossible. As a consequence more and more we see the use of teams of scientists made up of people drawn from different disciplines and gradually were developing it sufficiently common vocabulary so that cooperative endeavors between people trained in different disciplines is becoming increasingly effective in problems of this sort. Such a group of interdisciplinary scientists to which Dr takeoff has just referred is to be found on the University of Michigan campus at the Mental Health Research Institute. Here these men representing varying disciplines are gathered. Four of them Dr. James Miller Dr. Ralph Girard Dr. Anatol Rapaport and Dr. Merrill

flood. Where in this round table discussion and they discussed the questions What role will computers fail in research and what will be the practical applications in the future. The first gentleman to speak is Dr. Rappaport. There are several Cicero's are they from plague. First it is what the name implies it was a computer is a mathematical machine that is it does calculations for you. Much more rapidly at a fantastic rate than can be hoped for by any human brain and there are civil liberties but there are of course other uses to which computers can be put. These are called simulation situations on the computer it is a very interesting development of recent years of course simulation is not new. Engineers have known simulation for a long time in aerodynamics if you want to know where the behavior of an airplane will be and before you go up in the air you take it through a

wind tunnel where essentially the aerodynamic conditions are the same as they would be and the air so that in the way the wind wind is assimilating device for the atmosphere and also you can simulate the airplane itself by playing with a smaller model before you build a big one. Not a mathematical simulation. It does the same thing if you really love all the others you may have certain ideas as to how a certain situation can be described mathematically in terms of mathematical model and simply it is not possible with the present state of mathematical techniques to draw all the implications that model by deductive processes alone. So you simulate in the computer that is you program computer in such a way that the computer becomes a replica of what you suppose a situation to be and what you read out of the computer is the resulting behavior which then gives you the answer. If the situation was as you had pictured it that is if the model is approximately correct. This would have been the behavior which you can then compare with the actual behavior and see to what extent your original model was corroborated this corroborated. And if the latter of course you

modified That's one use to which computers haven't but can you predict behavior by the use of the computer. What you can predict you see is what the behavior will be if the underlying model is as you put it to be. I think you might be interested to know that right now Dr. Fine is conducting a research in which he is developing a program for a computer which will in one sense predict the behavior not of human beings in this case but of animals in a sort of maze in the laboratory. One reason for starting with animals is that we think they're simpler and easier to predict. And if we can learn to predict them the same method would apply in the case of the person and even the persons and groups. So that the aim of this research is to develop a method for predicting the learning behavior in animals. Eventually of rats and then in particular people now adjustments of personality which reflect themselves in their decisions learning and behavior. And this directly would contribute to

mental health. Physically I might say just one more aspect of computers and that is the computer as I see it today plays the same part on the functional theoretical level that the microscope played generations ago. The level of simply describing a structure of some material entity you look through a microscope in order to see things that you cannot detect with the naked eye. You put your data to the computer in order to see relationships which you cannot see with the naked commonsense. You might ask for example is this fact or relate to this one. Is there possibly a cause and effect relationship between such and such and you cannot do so by simply viewing data because there are too many complications. But the high speed with which computers are able to perform statistical processes carry wishes factor analysis and so on and they was used sometimes to see relationships as if through a microscope. I'd like to make one remark about the question of the shortage of people. Some

think that the computer will displace researchers. It's exactly the opposite. One large computer may run for one million dollars a year and the rule of thumb is that it takes that many additional people to people with information. So as with the electron microscope in the research field as you had a powerful tool by having that stronger aptitude in the field of A for doing research you need many more people to do it. But also what can be accomplished by the computer plus this group of people is fantastically greater scientific isor upgrade. The effort the same way you do when you take that big power shovel and have a man run out you can dig many more ditches. I got some very important point that always advances my give trenchant frictional dislocations for the long haul I don't use upgrading human achievement certainly KSAT there's one other aspect of the use of computers I don't think we've mentioned and that is that both brain physiology

and also the social sciences dealing with particularly the study of the complex relations of people in societies like the United States found themselves a few years ago unable to cope with the complexity of their problems because they didn't have tools like a computer to analyze to Stickley a quantitative way. Such complexity now we do have instruments which give promise of being able to cope. And once again quantitative methods can hopefully be plot applied in the social sciences and in brain physiology and elsewhere that ten years ago just couldn't be conceived. And for a lay individual could I ask this question. You have talked about the use to which the computer can be put scientifically. How may it benefit mankind. Well let me tell you one application I heard reported at a symposium last week by one of the large computer companies. Therefore it was on work. The Rockefeller medical research institute

where they have a computer which is already doing a pretty skillful job at diagnosing and screening would expect that one large class of physical ailments. So that here is a case where the computer is doing much more than helping on scientific research into powerful diagnostic tool to go along with an active cardiogram and other such tools. Another example was a research that's been suggested by. Dr. Kornblum one of our group here in which he will actually have patients or subjects in the scientific experiments interact with the computer they will answer certain questions. Brought to them by the computer they will the answer will be typed back into the computer again and then the computer will do certain calculations and come up with another type of question which will make possible a sort of interaction with a testing after it is that it is not possible either with the ordinary pencil and paper

psychological tasks or with apparatus tests that don't involve computers. It's almost as if the human being were interacting with another human being. The difference being that is interacting with another apparatus or with an apparatus that is capable of measuring and quantifying its fantastic rate much faster than an individual human being could. On the fanciful side I like Dr. Seuss with our. Group also as for calls to the computer might someday be a direct therapist in some fields and to give credence to this Professor Skinner Harvard University as the inventor of the so-called teaching machine for the computer in the best sense of the word. Is the complete teacher for a student in high school physics high school. Algebra and the like so teaching machine the therapy the Hsin are things of the future a few things about this teaching machine in the first place. The student need never be ashamed to ask a question of a teaching machine. Teaching

Campi has no feelings of superiority for you to start Secondly it's infinitely patient. So there are these very very real. But the guys you have doesn't go any faster than the student can see both grade reste to recommended then for the school. I am becoming very convinced that there is another side that people are sometimes more. Regular stimulating to students the machines are likely just a way to get the integration of ideas and facts. We extend our thanks to the members of the staff of the Mental Health Research Institute of the University of Michigan to Dr. James J Miller Dr. Ralph Girard Dr. Anatol Rapaport and Dr. Merrill flood for their participation on today's program and a special thanks to Dr. Russell a cough whom we heard earlier for his participation on this program on computers as simulator next week you will hear Professor Harry Helsa and Dr. Jack W. Dunlap and Dr. Edward R. Joe Johns as they discuss man and machine on the next program from the series

human behavior social and medical research produced by the University of Michigan Broadcasting Service. Glenn Phillips speaking asking that you join us next week and thanking you for being with us at this time. This program has been produced by the University of Michigan broadcasting service under a grant in aid from the National Educational Television and Radio Center in cooperation with the National Association of educational broadcasters. This is the end of E.B. Radio Network.