Graduate Instruction and Research

K.E. Iverson
 

The large-scale digital computer was first conceived and built solely for the purpose of scientific computation. In the short space of ten years since the completion of the first computer, these machines have reached such a high state of development that universities need no longer concern themselves with the scientific computer as a subject of research. For example, when our chairman, Dr. Jacobson, required a computer, he wisely procured a suitable unit without embarking on an extensive development program.

At the same time, educational programs in numerical analysis and the use of computers have been developed to fill the needs of scientific computation. Without in any way minimizing the continuing importance of such programs, I should like to point out that their pattern is now fairly well established, and furthermore, that there are solid reasons for relegating a considerable part of this work to the undergraduate level. For these reasons I wish to exclude consideration of scientific computers and computation and confine my attention to the educational needs arising out of new applications of automatic computers.

The development of the scientific computer has stimulated great public interest in applications which go far beyond those originally foreseen. Examples of these applications are familiar to all of us and range from business accounting and the control of industrial processes to language translation and the automatic compilation of concordances. Industry is applying the computer to a great variety of jobs, including payroll accounting, customer accounting, and automatic inventory control. A pioneer example of the latter application is furnished here in Detroit by the J.L. Hudson Company.

It is, of course, the need for processing data which is common to all of these applications and which makes computer techniques valuable to each of them. In principle the scientific computer is capable of filling the needs of any such application. However, in most cases a great deal of investigation is required to discover practical ways of applying computer techniques, and present progress in this direction is slow and unsatisfactory to all concerned.

There are many who would disagree that the present rate of progress is unsatisfactory. In fact, there is a school of thought which appears to hold the position that the scientific computer, or some small modification thereof, is a satisfactory tool for nearly all applications, and that all that remains to be done is to train any army of programmers to apply existing equipment to the jobs at hand.

It is possible that this attitude is correct. In order to establish this, however, it will be necessary to evaluate carefully the possibilities of machines designed specifically for each particular type of job. This is a task which should be undertaken promptly, but it will require the attention of persons having a much better educational preparation than that required for programming.

The reasons for the slow progress are apparent. On the one hand, those who could profitably use automatic computing equipment in their work have accumulated a vast experience in the present practices of data processing but possess little or no knowledge of the capabilities and limitations of automatic devices. On the other hand, engineers with the necessary knowledge of computers lack knowledge of the special fields of application. There is, in short, a scarcity of individuals possessing experience and training in all of the disciplines involved.

There are presently three distinct efforts being made to correct this situation. They are:

1. Cooperation between members of the several disciplines;

2. Short term courses in computers offered to operating personnel in the field of application;

3. Professional education on a graduate level.

All of us take part in the first of these efforts, through personal contacts and conferences, and many give assistance in short term courses such as the one just concluding here at Wayne.^[a] These efforts have been and will continue to be extremely valuable in providing the operating personnel required. Nevertheless, cooperation between members of several disciplines cannot hope to match the single individual with a mastery of all the disciplines involved, and short term courses cannot be expected to produce individuals with the fresh outlook required for significant advances in a difficult field. Only by professional education on a graduate level can we expect to develop people who will approach the subject from a research point of view and provide the leadership required.

Although it promises rich returns on a long-range basis, the third effort has unfortunately received the least attention. Because it has received so little attention to date, opinions as to the content of a graduate program are still in formation. Rather than attempt a survey of such opinion, I plan to present a single viewpoint, as embodied in the graduate program in data processing to be instituted in Harvard in the coming year.

Although the program was conceived in the Division of Applied Science, it was quickly recognized that such a task could not be undertaken by engineers alone. Accordingly, the cooperation of the Department of Economics and of the Graduate School of Business Admininstration was enlisted. It is firmly believed that close cooperation between groups of this nature is essential to the success of any such program.

The master’s degree program will follow the normal pattern at Harvard, requiring eight half-courses and no thesis, and usually will be completed in one year. The recommended program consists of the following courses:

1. Two half-courses in data processing. The first half-course is concerned with the use of computer techniques in the storage and handling of information, with applications to automatic record keeping and automatic control of industrial processes. Topics will include the representation of information, problems of access to information, a systematic treatment of the problems of sorting, logic and switching theory. Machine programming will also be discussed and illustrated by demonstration problems on a large-scale computer.

The second half-course will be devoted to the organization of data processing systems and case studies of existing systems. These courses will be presented jointly by a faculty member of the Division of Applied Science and a faculty member of the Business School.

2. Two half-courses in electronic control and calculating circuits. The first half-course deals with timing circuits and pulse techniques with illustrations drawn from a number of applications. The second half-course deals with timing circuits and associated apparatus applicable to modern data processing equipment.

3. One half-course in accounting. This course is designed to acquaint the student with internal accounting control in business enterprises. Methods of cost accounting, inventory control, wage accounting, and distribution cost accounting are among the topics covered. This course will be presented by a faculty member of the Department of Economics.

4. One half-course in operations research. Operations research is one of the most promising tools recently made available to management, and in discussing its place in the present program it will be helpful to distinguish between certain functions of management. One obvious function is the direct control of office operations. If office work is viewed as a routine job of information handling, there appears to be no more reason for management to take a direct interest in the details of office operation than in the engineering details of other production line operations. Nevertheless, because the office serves as the channel through which management operates, as well as the prime source of information, a direct interest in office operations has always been necessary. The mechanization of offices operations by the application of computer techniques may simplify the problems of management in this area but will have no direct effect on the second and more essential function of management, namely, the effective scheduling of interrelated activities.

It is in this area of scheduling interrelated activities that operations research is of value. The projected course will be presented by a member of the Department of Economics. Topics covered will include input-output economic analysis, theory of games, and linear programming.

5. One half-course in machine components. This course is a survey of the various storage and switching devices finding application in data processing systems.

6. One half-course in statistics. There should be no need to establish the importance of statistics in data handling operations. Nevertheless, the following comments may be in order. In assessing the requirements of data processing systems, it often appears that the collection of the required statistical information looms rather large in point of equipment required. For this reason a careful look at the possibility of using carefully chosen samples rather than a 100 per cent coverage may be very rewarding.

It will be noted that courses in numerical analysis have been omitted from the foregoing program. This is not to imply that work in scientific computation is no longer of importance, but rather that such courses have now become established and that the new program is oriented toward business applications. Courses in numerical analysis are, of course, available and may be elected instead of some of the recommended courses.

Especially on a graduate level the variations of background preparation require the possibility of substituting other courses for some in the recommended program. In addition, courses in the theory of switching, servo-mechanisms, and analog computers will also be available.

While the normal program will emphasize applications to business operations, other specialities will be recognized. For example, a specialist in the field will lecture on the problem of automatic language translation. The automatic processing of information of a continuous nature such as is encountered in the control of industrial processes will also be considered.

An essential part of such a program is the provision of facilities for research. These facilities include two large-scale computers, well-furnished scientific libraries, and the extensive facilities of the Business School library.

Especially in the field of accounting are we fully aware that this type of endeavor has no precedent on a graduate level. The need for constant research by the instruction staff is especially great in a new field which has little or no literature. Accordingly, strong research programs must be established and should serve the stimulate the interest of students.

At present, a detailed study of any one of a number of business applications presents a challenging field of research. Until general principles begin to emerge, each type of business application will present a unique problem, and it is doubtful whether the art will mature as rapidly as was the case with scientific computers.

One result of research should be a continued improvement in the course program itself, and several years may be expected to elapse before a stable and satisfactory solution is reached. Nevertheless, in view of the urgent needs which were pointed up by earlier speakers, it is clear that a start must be made as soon as possible.

Editor Note: