research and perhaps for administrative computing. In some cases small colleges may be so remote from existing or new centers that they may have to establish and use limited computer facilities of their own. Finally, as we have noted, some needs for educational computing may be met through centers established by groups of universities, or through private services.

Problems of Educating Faculty

Obviously, the faculty plays an important role in determining the rate at which computing is introduced into undergraduate courses. We feel that an intensive effort will be required to show the faculty the advantages and importance of computing and to help them to learn to use computers effectively. A number of suggestions to this end are contained in Appendix D.

There are basically two types of faculty education required if educational computing is to find a useful place in a college. First, the campus must have at least one faculty member who can teach a good basic programing course. This faculty member can be in any department-his prime attribute can be enthusiasm. Many young faculty members are already reasonably well qualified to teach such courses; others will need to participate in summer institutes or other special programs.

The second type of faculty education which is needed is more difficult to provide but is not unfamiliar. It is the basic education associated with any substantial revision of course material. To take maximum advantage of the computer it is frequently necessary to integrate new problems into a reorganized course. Planning problems and preparing the course revision require a large amount of time and effort from the faculty members involved and a consequent reduction of available teaching staff. Once done, of course, the normal amount of effort required from faculty members to keep abreast of a field is sufficient to make such courses generally useful by all members of the department. The examples in Appendix J illustrate the results of this kind of problem planning at several schools. The final report on “The Use of Computers in Engineering Education" * contains a list of 66 problems for various engineering fields. The rate at which problem planning and course revisions can be carried out depends heavily on faculty interest and on their understanding of the power of the computer as well as on the availability of facilities.

Because of the great importance of this faculty education, we recommend an expanded faculty program of education to provide adequate faculty competence in the use of computing in various disciplines. The Government should provide all funds necessary to support such a program.

"The Use of Computers in Engineering Education," final report of project supported by the Ford Foundation, College of Engineering, the University of Michigan, Ann Arbor, Mich., Jan. 1, 1963.

The Problem of Controlling Computing

It has often been proposed that computing should be an overhead itemthat it should be supplied to students and faculty without any formal procedure of allocation, as is library service. We believe this to be unrealistic. It is perfectly possible for one user to write one program which justifiably (although more often unjustifiably) will run one or more days on the most advanced computer available. Such a monopolization or preemption of a university library never occurs and thus does not pose a comparable problem. Although most computer runs are short (less than a minute), to arbitrarily limit the time any program can run would preclude some important and legitimate uses of computers. Further, in the absence of adequate control, students can fritter valuable time away in meaningless computer use even when the running time is short. Therefore, control cannot be exercised by just limiting the maximum length of any run, but instead requires that cumulative use records be maintained for each user. We find that institutions which have a well-run computer center keep careful records of computer costs, and keep careful records of computer usage by assigning job numbers and charging computer time against jobs. This accounting will be more complex but no less essential for remote consoles.

Some universities (Michigan, for example) allocate computer usage among various departments on the basis of time. Others (Harvard, for example) allocate computer usage on the basis of dollars. We believe that measurement in dollars is more meaningful to users and is more useful in comparing the value of computing with the value of, say, laboratory apparatus or special-purpose computers than is measurement in terms of computer time. Regardless of whether the unit is time or money, it is important that some allocation procedure be used which provides effective control.

Measurement of computer usage in dollars need impose no extra burden on faculty members, and might indeed eliminate the burden of arranging or begging "free" computer time here and there, as some faculty members must do at present. For example, faculty members can have budgets (the source may be university department funds, contract or grant funds) from which their students' computer costs for basic instruction, course problems, or research work can be supplied.

Keeping just account of costs of computing can be a powerful tool in university and agency hands in avoiding unprofitable research programs. The cost of extremely lengthy computations in various fields, including medicine, information retrieval, and artificial intelligence, should at all times be clear to the research worker, the university administration, and the funding agency. This does not mean that a uniform accounting system should be imposed on universities, but merely that the accounting system used by each university should be clear and should apply to all computing services.

We recommend that colleges and universities develop and use accounting procedures which accurately measure the cost and utilization of computer services. With such information, the allocation of computer time for research and education and the anticipation of associated costs should be made on a realistic and measurable basis. This should encourage a more balanced allocation of funds for research and educational uses.

III. THE COMPUTER SCIENCE STUDENT

We have noted in the introductory section of this report and in Appendices A and I the magnitude and rate of growth of computing in this country and the need for more men trained in computer science.

Computer science advances and changes as rapidly as all of the computer art. Thus, it would be futile as well as highly undesirable for us to try to describe or prescribe in detail what computer science is or should become. That is a matter best left in the hands of the academic community, to evolve through interaction with the computer industry and users of computing, academic, and nonacademic.

Computer science had its academic origins in designing and building computers in universities. Indeed, the very first electronic digital computer, the Eniac, was built at the Moore School of Electrical Engineering at the University of Pennsylvania. Today, it is no longer appropriate for a university to build a large-scale machine to provide its routine computing service. We also believe that the time is rapidly passing when it is appropriate for every university to develop a large-scale system program for its routine computing center. These aspects of hardware and software will be largely the province of commercial enterprises because the great effort involved in such developments cannot be carried out without excessive delay in the universities.

Yet men must be educated to understand hardware and software very deeply. By providing such education, the universities cannot only supply the computer industry with needed expert manpower; they can strongly influence progress in both hardware and software. Progress requires both university research and commercial enterprise. In general, universities will work on special hardware and software necessary to exploit computers in new ways for new or more efficient uses. This will include the design and construction or adaptation of special peripheral equipment, the development of the necessary software, the devising of new program languages, and the derivations of special procedures or algorithms for obtaining desired results.

Education for such research in computer science includes theoretical studies of machines and machine organization, the study of software and languages and their relations to a wide variety of disciplines, the study of hardware, and appropriate background work in mathematics, physics, and other fields of engineering and science.

This work calls for access to and interaction with a good computer center. Since many computer science departments also grant a master's degree, it is difficult to separate the undergraduate use from use at the master's level. Graduate work in computer science calls for substantial use of computer time in carrying out research on software and toward new computer applications. Such time may often more appropriately be paid for out of research project funds rather than as an educational expense.

The demand for people trained in computer sciences exceeds the supply. In fact, we have noted previously that one argument for supplying educational computing remotely from centers which serve many schools is that there is not enough trained manpower to establish and staff good computer centers in all colleges. In trying to evaluate the needs, we have been unable to find adequate data on the number of men with various skills now employed in the computer field in industry, government, and schools, or any meaningful estimate of the number who will be needed in the future. (What seem to be the best estimates available are given in App. I.)

We strongly recommend that the Federal Government collect meaningful data concerning computers and the jobs, personnel, and educational facilities associated with them, and endeavor to make useful annual forecasts. We caution that because of the rapid evolution of the computer art and its highly technical nature, useful studies must rely on well-informed and astute knowledge of the state and evolution of the art as well as on statistics.

Despite the importance of instruction in computer sciences, the total amount of computing connected with such instruction will certainly be small compared with the total amount of undergraduate educational computing which we have estimated earlier in this report because there are so many fewer computer science students than there are college undergraduates. Thus, if the deficit in undergraduate computing is made up, as we propose, an adequate amount of computing would be available for computer science education. It is of course important that such use be recognized as a part of the educational use of computing.

We must not, however, overlook the quality of computer facilities necessary for good education in computer sciences. In order to provide the computer experts who are needed to produce the new computers and computing techniques which are so vital to our national defense, to increasing our productivity, and to improving our standard of living, we must have excellent computer science departments at a number of schools. Though computer science education and research need place only modest demands on a large computing center, the quality of the center is of utmost importance.

It is hard to see how a master's program in computer sciences can be conducted without use of a center of such quality that costs would be about $1 million a year. Of course, most of the yearly cost of the center would be covered by charges for uses other than computer science education. For