Education or Training: Reflections of a Life in Computing

This year marks the 40th (the tyranny of base 10 numbers) anniversary of my first introduction to computing.

I have been much too busy over all those years to take the time to consider systematically the dramatic changes and the lasting trends that I have observed. However, when I was asked to do so recently, it seemed to provide the excuse I needed to undertake the common indulgence of those who have seen this many years -- to reflect on them.

The Evolution of Computing

In 1959, I was 17 years old and a sophomore at Haverford when the college acquired its first computer, an IBM 1620. As part of that acquisition, IBM included a six-week course for any member of the college community (remember this was before the "unbundling" decision). I was a chemistry major, but chemistry then at Haverford College was not particularly quantitative and that was long before the use of computers for graphic modeling of atoms and molecules. I had taken Introduction to Sociology and was planning to take Social Research Methods, which entailed some statistical analyses. Also, at that time I was planning to teach science and to train other science teachers and I wanted to learn as much and as varied science as I could. Computers appeared to have some (small) potential for use in the sciences, so I decided to take the course.

The course was devoted almost entirely to teaching us the arcane fundamentals of a language called SPS (for Symbolic Programming System), which was the assembler language of the 1620. I have several memories of that course. It was my first encounter with what I call "training," as opposed to "education." I did not understand that the intention was not to provide a broad understanding of computers and programming, but simply to enable me to actually write some functional code. Training as opposed to education is an important concept and one that is quite relevant today, particularly in light of the urgent calls from governors and legislatures to prepare students "for the work force." I will return to this issue several times.

My second memory of that course is that it was a non-credit activity that had to take second place to my regular course work. At that time I had no expectation of a career in computing, so I don't think that I got as much out of the course as I could have had I been able to devote myself to it more fully or had I seen it as a key foundation for my future.

The third point that remains vivid for me even after all these years is how impressed I was that the instructor was a woman who was probably not more than 10 years older than the undergraduates. I did not expect a woman instructor -- even though my parents were intellectual equals who often worked and wrote together, and my mother's books were generally on scientific subjects. However, outside of my home, women were much less likely to be in such a position; mathematics, the natural sciences and computing were far more male-dominated then than now. But I am happy to recall that IBM had made a concerted effort to recruit and retain women long before other companies and had, thus, benefited from some of the brightest young minds in the field. I think I learned more in that course about the equality of people than about computing -- a lesson that I hold as far more important.

Early Uses of Computers

I employed the computer a bit as an undergraduate, but only in my social (not natural) science research. It wasn't until graduate school at Teachers College, Columbia University, that I became deeply involved with and somewhat proficient at using computers. They also had an IBM 1620, so I felt at home with it already. I was studying tests and measurement and evaluation, which entailed considerable descriptive and inferential statistical analysis. I decided that I should take the one computer course offered. It was a FORTRAN programming course taught by the director of the computing center, who was also a professor of educational psychology. That course trained me to do some programming, but more importantly it educated me regarding some basic programming and computer concepts. I was hooked! I served as lab instructor for the course the next semester and took over teaching it after that. As have thousands of other graduate students since then, I got a job in the computing center, first as "assistant," then as assistant director, and then as associate director.

My assignment was to run the center single-handedly from 5 p.m. until midnight. It was an excellent learning experience. The center was magically transformed into a research facility at 5, when the administrative computing staff went home. In the evening, the faculty and graduate students came to do their computing, almost exclusively statistical in nature. I was forced to learn all aspects of computing -- hardware operations and maintenance, systems and applications programming, statistical analyses, research methodologies, and dealing with users, vendors and service technicians. I also wrote all of the programs for the analysis of my own dissertation data. Besides being trained to perform certain functions, I was being educated seriously regarding computing.

I was creating a standardized test and the norming tables were several hundred pages long. I had no desire to retype those tables for inclusion in my dissertation, but printers then used only pinfeed paper, not the 20-pound bond required by the university. I printed my tables on the computer, took them to the Xerox office in midtown Manhattan where they were transferred to microfilm, and then had the microfilm printed by Xerox on 20-pound bond. When the administration was assured by Xerox of the permanence of the ink, I was permitted to include the pages directly in my dissertation. To the best of my knowledge, that was the first dissertation at Columbia University and probably one of the first in the country printed by computer. What is today commonplace with a laser printer was then a major revolution.

The IBM 1500 System

While I was associate director of the Teachers College computing center, several professors in Special Education obtained a large grant to investigate, among other things, the potential of computers for studying and educating children with special needs. Grant funds were used to acquire an IBM 1500 system to replace the 1620. The 1500 system was years ahead of its time. The innovation and power of the system were in its peripherals (unique for their day), its multi-user operating system (one of the first of its kind), and its COURSEWRITER II programming language (an assembler-level language designed specifically for writing computer-assisted instruction modules).

During the early 1960s, IBM researchers had experimented with computer-based instruction, using earlier computers and an assembler-level language they called "COURSEWRITER." The success of these experiments, together with the demand for better education fueled by the Russian Sputnik successes and the programmed learning research of such psychologists as B. F. Skinner, led IBM to develop the 1500 system. The system was purchased by about a dozen other educational institutions (including school districts in Montgomery County, Maryland, and Kansas City, the Ohio State University and the University of Alberta in Canada).

In effect, we all were paying IBM for the privilege of doing their market research. The questions in their simplest terms were: "to what extent is computer-assisted instruction feasible, effective and affordable?" The answers in their simplest terms were that it was clearly feasible and could be quite effective for certain subjects under certain conditions, but it was very expensive. In arriving at these answers, we all learned a great deal about education, about computers, and about combining the two. Unfortunately, too often many of these lessons are unknown or disregarded today by the creators of computer-based instructional materials.

Those of us who were working with the 1500 system at Columbia and elsewhere proved that computer-based instruction is extremely expensive unless it can be amortized over a very large number of students. The hardware cost more than $250,000 (the equivalent of more than $1.25 million today) for a system that could accommodate a maximum of 16 (the tyranny of base two numbers) simultaneous users, about $16,000 ($80,000 today) per user. Modern computers with similar peripherals today cost in the neighborhood of $1,500-2,000. If we add the expense of a network and the requisite servers, the total cost is much higher. However, a modern network can serve far more simultaneous users, thus lowering the cost per user significantly.

In contrast to the hardware, the cost to develop computer-based instructional materials was and remains very high. In the early days of developing routine courseware for the 1500 system, we estimated 100 to 300 hours of development time for each hour of learner time. For courseware that included specialized or innovative features, the development time could be three or more times greater. I believe that these estimates are still reasonably valid today, even with the advent of many application programs that assist authors with the technicalities of coding. The bulk of the time commitment remains in the intellectually challenging instructional design activities required to create effective and engaging materials.

We and other developers of 1500 system computer-based courseware found that learning many subjects that are generally based on rote memorization can be enhanced by the use of computer-based materials. For example, learning vocabulary in a second language has often been the subject of computer-based materials. I heard that some such courseware developed at the University of Alberta for their 1500 system was used there quite successfully well into the 1980s (they amortized the development cost over enough students to make it truly cost-effective). It was finally abandoned not because better materials became available, but because replacement parts for the 1500 system could no longer be obtained.

Here we come again to the distinction between what I have termed "training" and "education." The "routine" courseware mentioned above can be quite effective for training -- that is, the process of teaching someone a specific set of facts (such as addition or multiplication tables, or second-language vocabulary) or a specialized routine (such as how to assemble a particular widget). Education, however, entails far more -- it is the process of helping someone to understand a principle or a concept, to perceive the interrelationships among various concepts, to be able to analyze situations and problems, and so on. Computer-based instructional materials designed to educate, rather than train, require far greater sophistication and development effort.

Enter the Web

It seems to me that this basic lesson needs reinforcement as we undertake Web-based instruction. While there are some superb materials that have been created, there are also many instances of exercises that have simply been moved from a workbook page to a Web page, and some are no more than automated page turners for text and graphic materials that may be better referenced on a printed page. Educators need to encourage governors and legislators who recognize the potential of providing distance learning opportunities for citizens to take into account the distinction between training and education.

There is a need and a place for both education and training. To create excellent Web-based training is possible and may even be cost-effective if the learner base is large enough. Web-based education is in its infancy and requires interpersonal interactions that the Web does not now easily facilitate. I am not referring to "collaborative learning," which can be achieved through electronic communications such as e-mail, chat rooms and bulletin boards. I am referring to the personal discourse, group and classroom discussion, and mentoring, needed particularly by less mature and experienced students. I am also referring to the personal interactivity required to help a learner draw upon experience and intellect to understand and internalize complex concepts and constructs.

The report of the Boyer Commission on Educating Undergraduates in the Research University, Reinventing Undergraduate Education, made this same point very well: "as [technological] innovations multiply, so do dangers: in many circumstances, casual over-use of technological aids already increases the real and psychological distance between living faculty members and living students. Technological devices cannot substitute for direct contact."

Alfred Bork and David R. Britton, Jr. (Computer, June 1998) came to much the same conclusion as I regarding the current capacity of the Web for education. They surveyed a number of existing courses and did not find the capability for high-quality interaction and other requisite activities. On the other hand, Patricia Cravener (Computer, September 1998) offers a rejoinder in which she describes several "Web-centric integrated distributed learning environments (IDLEs)." It seems to me that both views are valid: the Web may be capable potentially of supporting some forms of education, but presently it is really suited primarily for computer-based training.

I see a parallel to the recommendations sent recently by a blue-ribbon presidential advisory commission to the Clinton administration regarding funding priorities for computer science research. That report called for an increase of $1 billion over the next five years to support long-term basic research of the type that has given us the Internet, rather than short-term, product-oriented research that addresses a limited, focused question such as which phosphors produce a brighter green. Looking at the long term, studying the underlying principles, is analogous to education. Product-oriented studies are analogous to training. The report concludes that both are essential.

A related lesson learned from the 1500 system and subsequent computer-based instructional projects is that computers enhance learning most when they are used to perform activities for which they were designed and which would be much less efficient or even impossible otherwise. This means, for example, that using a computer to present a computationally-intensive simulation that illustrates a concept can be of great benefit. An early example of such a simulation was developed to demonstrate Kepler's laws of planetary motion. If the equations for the laws are solved for various bodies, their motion can be plotted, but such calculations are too tedious to perform by hand. Using the computer simulation, however, students could quickly plot orbits for many different bodies. They could also alter the laws slightly by changing the constants and powers used and then observe the resulting orbits. In this way, students could gain a "feeling" for the laws that went beyond merely memorizing their equations or being able to recite their prose formulations.

The converse of this lesson is that it is extraordinarily expensive and inefficient to try to use a computer in computer-based learning situations to perform an activity for which it is not designed. The best example of this from the 1960s and early '70s comes from attempts to have computers teach medical students how to take medical histories and develop diagnoses. When the developers tried to add natural language input and output to the courseware, the cost skyrocketed and the results were inefficient and inconclusive at best. Some systems achieved about 90% recognition of natural language input, but remained quite unrealistic and cumbersome. The situation remains essentially unchanged today. Unrestricted natural language interactions for computer-based learning are still essentially unrealistic, although natural language dictation, particularly with a somewhat restricted vocabulary, is almost commonplace.

What has now been learned is that computers in general and the Web in particular have a major contribution to offer in the field of diagnosis. They make enormous quantities of information available for almost instantaneous recall by a diagnostician, far more than anyone could be acquainted with or memorize. Clinicians and researchers can be immediately in touch with the latest developments in almost any specialty or subspecialty, but in written and graphic formats, not in natural language. Using the computer for what it does best (in this case, high-speed information storage and retrieval) helps to maximize the humans' abilities and capabilities.

Computers are the Means, Not the End

Using a computer in education for activities for which it is designed is another way of saying that we should use a computer as a tool to enhance education, much as we use it as a tool to enhance research and practice. The uses of computers made by my own children during their educations are illustrative of this observation. My sons, now 30 and 33 and both computer professionals, used computers in two basic ways: as text processors and as vehicles to learn about programming and engineering. They learned to program and to interface various devices to our home computer. They did not have computers in school. My daughter, now 13, also uses her computer as a word processor (albeit, a much more sophisticated one than was available to her brothers), but she has engaged in some other educational activities. At school, her accelerated reading program administered quizzes and kept track of students' progress by computer. In her classroom, there were a few occasions when connections were made to outside resources, but they had very limited communications capability. She has used computer-based references, such as electronic encyclopedias and Web sites, in the preparation of papers, and loves to draw and play games on her system. For each of my children, their basic education resulted from reading, study and interpersonal interactions; the computer activities were supplementary and supportive.

I have spent my career using computers and helping and teaching others to use computers in support of their endeavors. When I have taught basic computer theory, regarding the simplest of computers, I have explained that all one needs is a circuit to combine two bits (either arithmetically or logically) and another one to compare two bits and branch accordingly. Any computation could, in theory, be accomplished on such a device, although it would probably be very inefficient to do so. I have restated this theory as Tannenbaum's first law of computing: "You can do anything on any computer, if you try hard enough." (My second law is a bit more cynical: "On a network, you are always lowest on the totem pole," meaning that there always seems to be someone with a higher priority who is slowing you down.)

Applying my first law to instructional computing, I emphasize the second clause. It is sometimes so hard to do something that it is not worth even trying. Recall Cravener's rejoinder to Bork and Britton. She is right: it is possible to provide education via the Web. However, the required investment may not be justifiable. We have limited resources for education. The greatest need is to employ these resources in the most efficient manner to achieve the most effective education of our students. Training for large numbers of students can often be developed and delivered efficiently using computers. Education often cannot. We need to be sure that we understand the difference and allocate our efforts accordingly.

I believe that we have before us a wonderful opportunity to redress some of the severe social inequalities that plague our society as a result of unequal access to education. The Web offers the potential to reach countless people who otherwise are effectively cut off from higher education and other sources of training. For example, single parents working to support a family can study at their convenience; rural citizens can study without having to travel unrealistic distances; people in urban settings can study at hours when public transportation might be unavailable or when travel might be too dangerous. Also, via the Web, people can access courses from anywhere in the world, not just from their local institutions.

Realizing the Potential of the Web

The Web will remain just a "potential," however, unless society is willing to invest the resources necessary to make it accessible and to make appropriate learning materials available. The two major issues are connectivity and courseware. In many rural states, an average of 17% or more of families do not have telephones; a much higher percentage do not have computers or modems; and many of these families are not within a reasonable distance of a public facility such as a library that has any computer, let alone one available at an hour when they need to use it. The same is true, perhaps to an even greater extent, for many inner city dwellers. Therefore, the first requisite that must be filled in order to make the potential of Web-based instruction a reality is that we find a way to provide universal connectivity. Otherwise, poor, rural and inner city families will remain cut off from educational opportunities.

The second requisite is that appropriate instructional materials must be developed for the Web and maintained. The expense of establishing connectivity will be very large, but it is dwarfed by what will be required to produce and provide high quality Web-based instruction of a nature suitable for all of our citizens. Even if all we provide via the Web is training, that will require a massive investment; to develop educational materials is beyond any realistic budget that I have seen proposed. Educators and other interested citizens need to convince our governments of the importance and value of providing both training and education to help eliminate inequality. But we also must be sure to make clear the real costs as well as the expected benefits.

Robert S. Tannenbaum is director of Academic Computing Services at the University of Kentucky. [email protected]