This is one of several parts of my undergraduate thesis at Stanford entitled “Academic Revolution and Regional Innovation: The Case of Computer Science at Stanford 1957-1970”. It was submitted on May 17, 2011, and the text here remains unchanged and unedited since then.
The construction of new disciplines within the academy provides rich insight into the politics of knowledge. An academic discipline is the intellectual and bureaucratic categorization of a domain of human knowledge, and is composed of its practitioners, modes of communications (such as journals and conferences), and database of acquired knowledge.1 The modern American research university includes several prominent features, but two of the most important are its relatively decentralized structure based around these academic disciplines and the self-autonomy of the faculty members to determine the course of their fields.
These disciplines are not static, but adapt as new domains of knowledge are created. The classic quadrivium of arithmetic, geometry, music and astronomy has gradually expanded to encompass a large range of disciplines loosely grouped into the natural sciences, social sciences and the humanities. Today, it is not unusual for universities to have dozens of departments organized into several schools. While there has been a long intellectual history about the possibility of these disciplines eventually unifying, the general trend has been for further branching and fragmentation, with interdisciplinary programs acting to overcome some of these gaps.2
The terrain of fragmentation is constructed through a mutual shaping between the institutions that support knowledge production and the people that populate them. The peculiar features of the modern university create an environment in which social and political factors can play an instrumental role in the direction of this branching process. A new discipline may receive resistance due to an intellectual disagreement over its definition as a domain of knowledge, or simply due to the power held by administrators. Thus, the creation of a new field represents an important area to study the politics of knowledge, or the ways in which social and political forces shape the use, dissemination and discovery of knowledge.
History plays a crucial role in the development of new disciplines, including computer science. Starting in the immediate postwar period after 1945, there was an increasing supply of computation to researchers and practitioners due to the quickly developing power of computers.3 This increasing capacity allowed for a greater range of potential calculations, and notably affected the field of numerical analysis, the study of methods to solve mathematical problems for which analytical solutions are not possible.4 Computer science was partly constructed out of numerical analysis, and this derivation is particularly important in the course of its development at Stanford.
Starting in the late 1950s, there was a growing realization among computer researchers that a discipline existed outside of the currently accepted domains of knowledge. The pathways from the current disciplines to computer science varied by institution, but here we focus on the Stanford case. George Forsythe, the father of computer science at Stanford, was a numerical analyst and a faculty member in the Department of Mathematics. Almost immediately after joining Stanford in 1957, he set out to construct this new discipline, but not without difficulty.
The branching of computer science from mathematics created significant tension due to several concerns. First, as a new field, computer science faced the burden of developing the institutions and organs needed for an academic discipline. Computer science did not have the typical elements, like journals, that would formalize and rationalize the discipline, and thus, it was difficult in the early years to find support among faculty members, particularly during tenure cases.
Another source of tension was the intrinsic nature of computer science, which prevents the field from being clearly defined as either a basic or applied science. The discipline has constructions like the Turing machines, the discovery of which is one of the most important intellectual developments of the 20th century. At the same time, it has very prominent applications through its utility in programming computers to perform tasks. This divide is at the heart of the faculty debate presented in this chapter. Skepticism about computer science was common, especially among faculty in the natural sciences who perceived the discipline to be closer to engineering.
However, it was not just external faculty who debated the definition of computer science, but computer scientists across the country as well. One side argued that computer science should facilitate research in other fields, and should thus be considered a service to the university. This conception was common among researchers funded by university computation facilities. Others in the field argued that computer science was its own academic discipline and domain of knowledge, and thus should be considered an equal in the university. Unsurprisingly, this perspective came from faculty who desired an academic career studying in the area.
A fundamental component of this debate, and the politics of knowledge more generally, is the definition and application of academic legitimacy. Legitimacy within the academy is essentially the acceptance of a discipline by the faculty of other disciplines, as well as generally by the school administration. However, there are additional components to consider. To be a discipline requires practitioners, modes of communication and a database, and there may be varying disagreements on each of these parts. For example, should a memo sent through the mail be considered a “publication”? What should the qualifications be for a faculty member in a field that does not have doctorate programs?
This study engages with this broad intellectual framework by analyzing the history of the development of the Computer Science department at Stanford. Computer Science developed out of the Mathematics department, and it is here that the first frictions between the new field and established disciplines took place. As Computer Science gradually expanded in size and independence, critical attention from faculty in other disciplines became more palpable. Nowhere was the conflict over the field more evident than in the debates over faculty appointments in what was first the Computer Science division and then later, the Computer Science department. While faculty outside the field had little control over the autonomous administrative unit, their input was a critical element of the tenure process.
This chapter analyzes and contextualizes the conflicts over the legitimacy of computer science primarily by analyzing the tenure cases of John McCarthy, and particularly Marvin Minsky and William F. Miller. Due to the richness of the archival materials, these cases provide an excellent episode to analyze the politics of knowledge of computer science by placing different actors in dialogue with one another. The analysis will show that influential faculty were opposed to the development of computer science over concerns that its applied character did not belong in the School of Humanities and Sciences (H&S). However, organizational flexibility on the part of Stanford, particularly by the dean of the H&S school and the university provost, provided crucial support in the discipline’s formative years. Their “gamble” would eventually pay off as Stanford developed a reputation as one of the leading centers of computer science in the United States.
Numerical Analysis and Computer Science
Today, numerical analysis is defined as the study of algorithms for the problems of continuous mathematics.5 Its practitioners focus on the development of approaches for solving problems without analytical solutions. The emphasis on continuous mathematics is important, as continuous functions lend themselves to iterative methods of solution.6 These iterative methods, which include gradient methods and Newton’s method, are complemented by direct method approaches that can reach the precise answer in a finite number of steps. Such methods as QR factorization, the simplex method and the Householder transformation to calculate the singular value decomposition are among the most well-known tools developed in the field.
To get a flavor of the sort of problem that is solved using numerical approaches, consider developing a best fit line for a set of data. The line is linear, and therefore it only has two parameters — the intercept and the slope. The line can be translated vertically by changing the intercept, and the line’s direction can be changed by adjusting the slope parameter. However, there could be hundreds if not thousands of points on the graph, and thus, it is likely impossible for a line to touch them all. Numerical methods are used to build the line closest to all of the points by reducing the total combined error of the distance between the points and the line. One direct approach is the method of least squares, but we could also consider an iterative approach that would slowly vary the parameters to steadily decrease the total error.
The highly-developed nature of the field today is vastly different from the terrain viewed by numerical analysts in the immediate postwar period. Many of the common algorithms and approaches used today had yet to be developed, and there was difficulty in implementing many of the methods due to the limits of human mental faculties and time to calculate iteratively. The rise of computers would fundamentally alter this situation. The development of computation power in the 1940s and 1950s provided a new capability to researchers, allowing them to make quick calculations that are at the core of numerical analysis.
It was within this milieu that George Forsythe developed his career. He received a PhD from Brown University in 1941 and became a meteorologist during the war, a position that demonstrated to him the value of numerical solutions to problems like simulation.7 Starting in 1948, he began working at the Institute for Numerical Analysis of the National Bureau of Standards at UCLA. That position afforded Forsythe access to a computer, and he quickly developed a sense of the intellectual power that computation and programming could offer. As his experience increased, Forsythe began to encourage mathematicians to focus more on the development of computation, arguing that the possibilities of the technology were immense for the development of the field. He had a difficult time convincing mathematicians of that goal, a critical part of the later story at Stanford.8
This hesitation by mathematicians was partly due to the difference in stature of the two fields — mathematics and computer science — at the end of the 1950s. Numerical analysis is intimately connected with the development of mathematics. One of its core methods was developed by Newton, who also developed calculus. Despite the development of core theoretical concepts like the Turing machine, computer science lacked the storied history that provides a source for academic legitimacy. The scope of computer science is far larger as well, incorporating not just algorithmic development but also issues of theoretical complexity, software design, natural language processing, artificial intelligence and other domains. Numerical analysis thus also benefits from greater coherence due to its more restricted domain of research. That coherence was also a major concern of mathematics, as we will soon see.
Computer Science and Mathematics
Forsythe joined the Stanford Mathematics department in 1957 as a full professor, joining John Herriot as a numerical analyst. Herriot was among the first of the leaders of computation at Stanford, and the two men immediately began to consider approaches for developing the field of computer science on an educational and intellectual level within the department. In the years before 1961, there was no official administrative structure for computer science, although Forsythe did formulate a sub-field of sorts within the department by 1959. All decisions regarding the academic side of computer science in these years were passed through David Gilbarg, the chair of the department. Gilbarg’s interests were in algebraic number theory early in his career, but his work in World War II led him to focus on nonlinear partial differential equations and fluid dynamics for the remainder of his career.
When Forsythe arrived, the differences between the area now defined as computer science and the traditional field of mathematics were relatively few. Mathematics hired Forsythe to add strength in numerical analysis, and he strongly believed in the utility of numerical approaches, which he considered to be the next stage in the development of mathematics. He urged his colleagues that a mathematics education should include at least a basic background in using computation.9
The university administration was also enthusiastic about the new field. Albert Bowker, a statistician who was an associate dean in the School of Humanities and Sciences, discussed the formation of an autonomous division for the field from the very start of its formation within Mathematics.10 By 1961, the discussion had moved to the issue of logistics, and how such a division would be formed and operated within H&S. Gilbarg was not actively a part of these conversations despite being chair of the department, and felt with some surprise that the “thinking on this matter has progressed substantially, and much farther than I realized.”11
Gilbarg informed the H&S dean, the philosopher Philip Rhinelander, of his basic approval. Gilbarg was relatively enthusiastic about the creation of an autonomous division for computer science, arguing that it would provide coherence and would be easier to expand the faculty. Furthermore, he argues that if the field was to become a department, then moving to an independent division would be an appropriate first step. Even at this point, just two years after the creation of the sub-field, he notes that there was increasingly a divergence between mathematics and computer science, and that an independent division would fit the diverging nature of the two fields. “The new faculty contemplated for the Division,” he wrote, “would not ordinarily be appropriate as members of the Mathematics Department.”12
Not surprisingly, Gilbarg came down squarely in support of the approach taken by non-numerical mathematicians. “I admit to some qualms concerning the scientific quality of the work in Computer Science – at least compared with that in the traditional scientific disciplines. I refer primarily to the technological rather than fundamental character of much of the work.”13 Nonetheless, his more immediate concerns were pragmatic, consisting of requirements for transferring the faculty slots for Forsythe and Herriot into the new division and for separating the new budget. Tellingly, one of his conditions was that faculty appointments within the division should require his approval, creating an administrative review that would greatly increase the tension between these two fields in the coming years. However, there was little desire at this point to negotiate these requirements, and the conditions were accepted a few days later by Patrick Suppes, a philosopher of science and an associate dean of H&S leading to the creation of an official division within the Mathematics department.14
The heart of the issue is the unique economics of universities. Faculty slots are highly prized, since they are permanent budget outlays to a department and at Stanford, offered the possible benefit of tenure to their recipient. Since Computer Science was a part of Mathematics, the two programs shared faculty slots, and a zero-sum mentality developed over each new slot. In this case, the bureaucratic structure increased this divisiveness because the chair of the Mathematics department was the chief evaluator of faculty nominations from the division. The reviews succeeded when interests aligned, but Forsythe’s desire to expand into new areas like artificial intelligence would cause a permanent fracture in the relationship.
In 1962, one of the first computer scientists to be nominated for an appointment at Stanford was John McCarthy, a researcher in AI at MIT who had previously coined “artificial intelligence” at Dartmouth.15 Halsey Royden, a professor of mathematics and a member of the Appointments and Promotions advisory committee within H&S,16 commented on McCarthy’s qualifications to Robert Sears, who was by then dean of H&S. Royden wrote that McCarthy had previously been an assistant professor at Stanford, although he was not reappointed to the position in 1954 due to concerns that he lacked publications and an ability to find his own problems.17 However, Royden was excited about the future of artificial intelligence research, writing that “this is certainly a very exciting field at the present time, and I feel that it is very important for Stanford to move in this direction.”18
Like Gilbarg, Royden also perceived a difference between the area of computer science and mathematics. He commented that McCarthy was not strong as a mathematician, but that his sophistication was of “a higher order than is usually shown by people in the field of computer sciences.” Royden argued for the appointment by comparing computer scientists to mathematicians in the social sciences, where “there are very few people with established mathematical competence, but where it is important for Stanford to keep abreast of a developing field.”19 In Royden’s judgment, McCarthy was the best within machine learning, and thus Stanford should attempt to secure him. Finally though, Royden urged some level of caution, since McCarthy’s appointment as a full professor would likely cause some “unhappiness” among the faculty in Mathematics, who would feel that he is receiving “unmerited preference” over others in the department.20 With the support of Royden, Sears moved forward with the appointment, and officially brought in Gilbarg to the conversation.21 McCarthy would join the department as a professor later that year.
While there was minor concern over the appointment of John McCarthy in artificial intelligence, it would be the proposed appointment of a second faculty member in the area in 1963 — Marvin Minsky, a researcher in artificial intelligence who worked with McCarthy at MIT — that would lead to the complete separation of mathematics and computer science.
As computer science’s enrollments and research agenda continued growing in 1963, additional faculty slots were granted to the division. The university administration wanted to develop artificial intelligence at Stanford, seeing the possibility to strongly compete in the field against peer schools. Forsythe wrote in his budget request for the 1964-65 academic year that McCarthy was the strongest professor in the department and that he was transforming the division from “a modest mathematics-oriented group to a major role in computing.”22 McCarthy was in an excellent place to lead this movement because his research interests included both artificial intelligence and non-numerical computations. Forsythe argued that appointing Minsky to the division would complement this area by adding strength to McCarthy, as well as creating important connections to the Medical School.23
Joshua Lederberg, chair of the Genetics department, enthusiastically praised Minsky, writing that he was an “outstanding” choice for a position at Stanford. If appointed, Minsky would be given a joint professorship between Genetics and Mathematics, and Lederberg believed that Minsky could assist his department in its research in exobiology. Furthermore, Minsky would help with the “instrumentation crisis” facing biology and would develop new methods to combine the use of computers into the analytical equipment used in science and medicine. Furthermore, Lederberg placed the appointment into a grand narrative, connecting artificial intelligence with the search for a common paradigm in biology by comparing the adaptive processes of a computer to the evolution exhibited by cells. More strategically, Lederberg stated that the appointment of Minsky would allow Stanford to create a “concentration of talent” within artificial intelligence, allowing the university to become a national center, and asked the deans to dismiss “whatever controversy there may be concerning the actual present stature of this field of investigation.”24
Lederberg was not a minor person at Stanford, and his advice was not easily dismissed. He had been awarded the Nobel prize just a few years before in 1958 at the age of 33, and Stanford attracted him that same year to found and chair the Genetics department. He was known among the faculty as an extraordinarily energetic researcher with wide intellectual interests. Given his intellectual heft, his words of praise for another scholar like Minsky would have held significant weight with the university administration.25
The praise for Minsky was not uniformly positive. Royden told Terman that he had received letters from faculty that were pessimistic about Minsky and his research.26 By June of 1963, Royden warned him that the nomination was “viewed with considerable skepticism” by senior administrators, since many of the letters argued that Minsky was smart but had not done much in the field.27 By the end of the year, though, Royden would change his opinion of Minsky. That additional support was well-regarded by Terman, who wrote back to him that “your argument that you are sticking your neck out on this one with your mathematics colleagues makes a really strong point.”28
The philosopher of science Patrick Suppes, who had been replaced as associate dean by Halsey Royden in 1962 and had returned to the Philosophy department, also wrote in favor of the appointment. He acknowledged that Minsky lacked the kind of publication record that would typically be expected for an incoming faculty member, but argued that the dean’s office should support the candidacy since Minsky is “one of the few people of faculty status anywhere in the United States who is able to think creatively and originally about the non-routine problems of large computer systems.”29 Thus, Suppes argued that the definition of legitimacy should be expanded beyond just the typical evaluation of publications to include the potential of a candidate. Thinking strategically along the same lines as Lederberg, Suppes believed that Stanford could not become a national leader in the field relying exclusively on McCarthy, and thus an additional appointment was necessary.30
Suppes also told Royden to follow the divergence between mathematics and computer science as well as the nature of academic legitimacy. He wrote, “it is also important to realize that mathematicians [...] do not operate at the machine level.” Suppes contrasted the research of mathematicians, which “evaluates in standard fashion in the usual journal articles” with the research in computer systems that would be pursued by Minsky. Suppes wrote forcefully that “it is a hard but unpleasant fact as far as I’m concerned that the kind of original and creative thinking required to do imaginative things in computer systems is not the sort of thing that easily leads to publication, but that certainly is of an intellectual order comparable to research in many academic fields.”31 Thus, we see a professor from the Philosophy department (and with strong interests in science) with administrative experience providing a strong argument for the legitimacy of artificial intelligence as a field of study.
Not surprisingly, Forsythe was very high in his praise of Minsky, but placed the appointment in the strategic framework of the division’s growth. He wrote that Computer Science at Stanford had started around the research of numerical analysis, but that the “the next area to be cultivated was Artificial Intelligence,” which he defined as using computers to solve issues of pattern recognition, symbol manipulation and problem solving.32 Research in the field “is a long-range endeavor, whose big pay-offs are distant but of fantastic importance,” and thus Stanford should quickly take a leadership role. In terms of the department itself, Minsky would assist in bringing even more graduate students into the artificial intelligence area, and in Forsythe’s judgment, Stanford would pull ahead of the current leading centers of Carnegie Tech and MIT with the appointment.33
Forsythe, however, did not completely deemphasize Minsky’s mathematical connections. Minsky had his education in mathematics and was on the MIT mathematics faculty before moving into computation. Thus, Forsythe was sure that Minsky would “maintain high mathematical standards in Ph.D. theses written in the area of Artificial Intelligence.”34 Perhaps most interestingly, Forsythe argued that Minsky could help to heal the division between the numerical and non-numerical approaches to computer science that were beginning to split the division from the Mathematics department. Minsky would “foster an active collaboration between two groups in the Computer Science Division [the numerical analysts and the AI researchers] for the more intimate collaboration between computers and mathematicians in the solution of problems in analysis.”35
It was in mathematics, though, that the controversy over artificial intelligence would prove to be most acrimonious. Gilbarg recommended against the appointment of Minsky, arguing that he lacked a record of publications and was not necessarily a leader in the field. However, Gilbarg’s objections were directed less at Minsky and more at artificial intelligence as a whole. Gilbarg argued that “perhaps no other scientific area represented in Humanities and Sciences is so full of talk of future possibilities and yet so lacking in actual accomplishments.”36 In Gilbarg’s judgment, Stanford’s appointment of Minsky would be highly risky because “Stanford would be gambling on the future of Artificial Intelligence as an academic discipline.” He was deeply concerned that Stanford would appoint two fruitless faculty members, in a division that had only a handful of members.
More broadly, Gilbarg’s primary concern was the direction of the division as a whole. By appointing Minsky, the division’s focus would move away from numerical approaches to computer science and more toward non-numerical research. Considering that Forsythe was originally appointed to add numerical strength to the department, it was perhaps inevitable that there would be tension over the division’s development. Gilbarg thought that the new direction would undermine the attempts to build up numerical analysis, and thus would weaken Stanford’s first strong research area without developing a suitable replacement. Gilbarg concluded, “I do question whether such appointments will in the long run make for a high quality department of Computer Science, and whether they are appropriate for the School of Humanities and Sciences.”37
The collected comments of Lederberg, Royden, Forsythe and Gilbarg were provided to Robert Sears, a professor from the Psychology department who had responsibility for the appointment as dean of H&S. He was strongly enthusiastic of the appointment, believing that Stanford would not only build up artificial intelligence, but could develop a specialization in the computational aspects of biology and medicine. He sent Minsky’s forms to the school’s Appointments and Promotions committee, where the committee evaluated the conflict between Mathematics and Computer Science. The committee weighed the proposed appointment carefully, and concluded that it met the standards for a faculty position within the school.38 The committee overruled Gilbarg’s concerns by noting that he, unlike Forsythe and the other letter writers, was not in a position to judge the quality of much of his work.39
Sears himself added to the committee’s dismissal of Gilbarg’s objections, writing that “it is understandable” that he would want a numerical analyst who would benefit the department. However, Sears believed it best to allow the faculty of the division the autonomy to make their own decisions regarding hiring, providing them more authority than was previously granted.40 More importantly, Sears addressed the criticism that adding another faculty position in artificial intelligence was dangerous, given the relatively recent formation of the field Sears wrote that “I am fully aware that the whole enterprise of Computer Science represents a gamble on the part of the University.”41 However, the difference between Sears and Gilbarg was that the dean was willing to foresee where artificial intelligence could lead: “In my opinion, the University must be prepared to make this kind of gamble every so often.”
Sears was aware that the gains in the field might not be realized for many years, but that it was important to lay the groundwork immediately. He wrote:
I am convinced, however, that the imaginative application of brilliant intellection to the more effective use of computers will ultimately produce great dividends for many branches of science, and I feel a strong conviction that Universities have the responsibility to exploit every possible opportunity to stimulate and develop new fields of knowledge. Stanford made the decision four years ago to make one of its speculative enterprises the field of Computing Science. I think it is important that we back up this decision with the appointment of vigorous creative young faculty who can convert this speculation into a solid, producing, blue chip enterprise.42
By the end of 1963, artificial intelligence had established a significant base of academic legitimacy within the administration, if not entirely among the faculty. It was this organizational flexibility that facilitated the rapid development of computer science at Stanford.
With the nearly unanimous support of the administration, an offer was made to Minsky, much to the chagrin of Gilbarg.43 The consequences of the decision for the division of Computer Science were quick. Just a few weeks later, Gilbarg asked Sears in a brief memo to be removed from approving any new appointments within the Computer Science division, arguing that it would be “superfluous” given their growing size. He wrote briefly that the connections between the two fields were “at best tenuous” and that their research agendas were “diverging.” Thus, it would be best for Computer Science to receive its own department removed from Mathematics.44
After some conversation, Sears approved the general sense of Gilbarg’s goals, and informed Terman on how to implement it. He wrote that Computer Science will have two directions ahead for it, “one will be the quite elaborate ‘how-to-do-it’ teaching program” and the other will be “the non-mathematical research and graduate training.”45 Sears also acknowledged that the research in computer science, especially in artificial intelligence, was moving away from its origins within numerical analysis, and that “it is more in the nature of technology.” Sears recommended the elimination of the Mathematics chair’s review of new faculty positions, feeling that it would encourage the Computer Science faculty and “free them from what I sense they now feel as a kind of ‘Big Brother’ control.”46 Terman concurred with Sears, and felt that Computer Science was nearing department-level status within the university — perhaps in a year’s time.47
Gilbarg may have been right on one level. Back in mid-1963, when he recommended against the appointment of Minsky, Gilbarg argued that another appointment outside of numerical analysis would likely increase the difficulty of making an appointment in that field. He was referencing the case of Seymour Parter, a mathematician and numerical analyst with strong interests in the growing field of computation. In his two previous positions at Indiana University and at Cornell, Parter had served in a joint appointment between mathematics and the campus computing center.48 Considering the strength of Stanford in this field, he was a natural addition to the faculty, but Parter already had a full tenure offer to return to Indiana, putting pressure on Forsythe. He wrote to Gilbarg and Sears that people like Parter “come dear” and that the department should make an equivalent offer.49
The issues surrounding the appointment of Minsky would spill into the debate over Parter. The Mathematics department refused to offer him a joint appointment between Computer Science and Mathematics. Since the faculty of the division were members of the department, the difference was not one of budget but likely one of politics. Giving Parter a joint position would be acknowledging the authority of the Computer Science division to make its own appointments, and Gilbarg and many of the Mathematics faculty were likely opposed to setting such a precedent. Forsythe noted that Parter was “extremely sensitive” to the relationship between the two fields, and he would eventually turn down Stanford’s offer to join the faculty.50
Computer Science and H&S
Building Connections and the Computation Center
As the Computer Science division grew out of the Mathematics department, it increasingly had to engage with the rest of the School of Humanities and Sciences to secure its faculty slots and to support its budget recommendations. Developing connections between Computer Science and other departments thus became crucial in the drive to develop the division into a full-fledged department. These connections were built most easily at the Computation Center, which was led by Forsythe from 1962-1965 and provided centralized computing resources for researchers across the university. However, there was a growing fear that Computer Science would be subsumed by other departments and relegated to an exclusively service role within the university.
As director of the center, Forsythe was a major proponent of computing’s power to influence other disciplines, and he encouraged the use of the center’s computers by maintaining an open access policy.51 By autumn of 1963, the center was being used for sponsored research from several different departments and schools, including Physics, Electrical Engineering, Aerospace Engineering and the Graduate School of Business. The unsponsored usage, though, is more interesting. Several social science departments, including Economics, Sociology, Political Science and Communications were using the computers as unsponsored users, as was Petroleum Engineering in the School of Earth Systems.52 Thus, Forsythe’s open access policy was having a significant impact on the spread of computation to disciplines outside of computer science and its immediate counterparts, since many of these departments had no sources of revenue to cover computing costs.
Beyond facilitating computing, Forsythe, as division head, actively sought to build connections with other departments. One example comes from late 1963, in which Forsythe worked with Marcia Ascher, a mathematician with interests in archaeology, on developing a computation class for archaeologists. Forsythe wanted the class to be accessible, even to those who did not feel comfortable in mathematics, writing that “Engineers, natural scientists, mathematicians, and anyone else who knows what a function is are to be barred at the door – so we can have a more cultural environment!”53 This expansive use of computers and the need for computer science was noted by Forsythe in a report written shortly after the creation of the department in 1965: “It is now clear that students of social science must also acquire a familiarity with computing methods, And the serious student of humanities will soon find computers indispensable, if he is to carry out research on any substantial volume of data.”54 The same report also mentions that the department had arranged for one week sessions on computing taught directly to other faculty.
The connections between Computer Science and other departments also went the other way. In the early years, the Computer Science division could hardly teach a complete curriculum due to the small number of faculty in the program. Thus, a large part of the curriculum was taught by other departments, particularly the Mathematics department. At an educational panel discussion in 1965, Forsythe noted the importance of interdisciplinary activity when he noted, “Important Idea: CS is interdisciplinary. It’s essential that our students learn supporting disciplines. I feel that control of curriculum is essential. That doesn’t mean we should actually teach the entire curriculum.”55 This integration of other departments into the division’s curriculum provided immediate legitimacy — no department will question the quality of its own classes.
Finally, the division developed its legitimacy through the creation of joint appointments with other academic units on campus. Stanford’s provost, Frederick Terman, encouraged these joint appointments, including a critical one with the Stanford Linear Accelerator Center that would provide a billet for William F. Miller in the division.56 Terman told Forsythe that these joint faculty members must serve the real needs of Computer Science, and he reminded him that Computer Science cannot be strong in all fields — a reference to his “steeples of excellence” approach to building academic departments. The need to maintain a strong leash on joint appointments became increasingly important toward the early 1970s when physicists increasingly left their field and sought work in computer science, many of whom were under-qualified for research.57
These academic connections from the Computation Center brought new ideas to the Computer Science faculty, but it led to on-going concerns about the development of computer science as a discipline. This concern was never far from the mind of Forsythe, who perceived that computer science had to create a coherent intellectual area with the requisite institutions to build legitimacy inside the academy. Forsythe pushed officers at the National Science Foundation to fund and develop a publication similar to Mathematical Reviews that would provide comprehensive coverage of new developments of computer science.58 Later, he used the growing number of journals as indication of the rise of Computer Science as a discipline when requesting authorization to begin a PhD program.59 The issue of legitimacy continued even after the creation of the department, and Forsythe argued that the main intellectual problem for Computer Science was expanding out while not becoming so diffuse that it dissolved into the individual departments.60
The Case of William F. Miller
The divide over the direction of Computer Science after 1963 and its role in the university came up in the tenure decision of William F. Miller, one of the original members of the Computer Science faculty who would later serve as Stanford’s provost. Miller was a physicist, receiving his PhD in the field from Purdue in 1956 before joining the Argonne National Laboratory, working on the development of the computer. Given Miller’s background in physics, he was nominated in 1964 to fill a joint faculty position between the Stanford Linear Accelerator Center (SLAC) and Computer Science. Such a joint position logically fit the interdisciplinary Miller, and also allowed the division to expand its faculty for just half of a salary.
His appointment was recommended by the division, and by the administration in the School of Humanities and Sciences. Despite this support, the case was not received well by the school’s advisory Committee on Appointments and Promotions (A&P). There, the faculty voted unanimously against Miller’s appointment. Given its advisory role, the committee’s decision was not binding, and the School overruled and appointed Miller full professor of Computer Science later that year.61
Chemist Paul Flory was a member of the A&P Committee at the time of the decision. He came to Stanford in 1961, already among the most eminent chemists in the United States. Previously on the faculty of Cornell University, he studied the physical chemistry of macromolecules, developing a theory for analyzing chain molecules quantitatively that would eventually lead to a Nobel Prize in Chemistry in 1974. Flory’s stature was recognized at Stanford, where the administration placed him quickly on the A&P Committee.62
Flory was deeply concerned about the dean’s decision to overrule the committee’s decision on Miller. In a three-page letter, Flory explored not only the Miller case but also the wider issues of the role of professional and applied research at a place like Stanford. Regarding Miler, Flory wrote that he voted against him because of a lack of scholarly achievement. Commenting on speculation regarding the vote, he said that “The assertion that the Committee underestimated the significance of his contributions to computer science because of the unorthodox media of communication (ditto reports, etc.) in this field lacks credibility.”
Given the state of the computer science field at the time, these kinds of reports, despite Flory’s stature, may very well have been crucial scholarship, and we see the issue of legitimacy enter into the discussion. In addition, Flory was concerned that a candidate for the joint appointment would need “superhuman capacities” since the job would entail so many different types of activities, and thus he was skeptical if any candidate existed who could fill the position.63
These concerns regarding scholarship were certainly not unusual in tenure discussions. However, Flory’s arguments on the role of the professions deserves strong analysis. He began his letter by noting that he did not oppose Miller on the grounds that areas of applied science were moving too close to the School of Humanities and Sciences, even though “caution in this regard is imperative.” While he argued that “scholarship should certainly take precedence over shades of distinction between the professional and the central disciplines,” he continued, “the distinctions must nonetheless be regarded as significant in the academic scene.”64
After the discussion of the Miller case, Flory unleashes his main argument against the direction that Stanford was taking: “In some way, appointments in ‘growing edge enterprises’ (my underline) are to be fostered with emphasis on areas of practical concern, because these latter are said to be the well springs of new disciplines. Accepted doctrine backed by a good deal of experience replete with familiar examples teaches the obverse, namely, that areas of practical import spring from advances in the disciplines.”65 He noted that his views are traditional and not keeping with the spirit of the times at Stanford. Flory then crescendoed into his paramount argument:
No university can hope to mirror all new and promising areas of technology with their manifold proliferations in the present age. In fact, it must constantly guard against the ever present temptation to try to do so in an age of specialization. There are also the closely related pressures to develop enterprises, and these can be lethal to a great university. It is no secret that many of the faculty are gravely concerned over recent tendencies in this direction. It would be a matter of great regret if the School of Humanities and Sciences were to abandon its position as the bulwark of the disciplines in order to take unto itself technologies and professions at what may momentarily appear to be “cutting edges” of “new frontiers.”66
Flory’s argument was the most articulated response to the rise of Computer Science at Stanford, and showed both the importance of legitimacy and the politics of knowledge in the development of computer science.
However, what makes his letter particularly notable is that Flory was not opposed to industrial research. On the contrary, he had conducted it himself. His first job after receiving his doctorate was at DuPont, and he later worked at the Standard Oil Development Company and the Goodyear Tire and Rubber Company during World War II. While Flory was most notable for his theoretical contributions in chemistry, he was not the quintessential academic scientist who avoided practical work. Computer science thus faced a more difficult battle for legitimacy than perhaps initially perceived. The faculty of the division did not just have to convince ivory tower academics, but also academics who had spent significant time in industry.
Despite Flory’s criticisms, Computer Science continued to grow rapidly throughout 1964, and the university administration authorized it as a department on January 1, 1965. The discipline was very different from what Forsythe saw when he arrived in 1957. Numerical analysis was the exclusive province of the discipline then, but by 1965, the department included faculty engaged in areas as wide as physics and artificial intelligence. This development particularly worried Forsythe, who asked in mid-1965, “Has our creation of a new Ph.D. degree in computer science actually worsened the situation for the would-be numerical analyst?”67
Forsythe himself was fairly bitter about the entire situation ever since the conflict over the Minsky appointment. He had wanted a joint appointment with Mathematics when the Computer Science department was formed, but the request was denied. In addition to the lost benefits of communication, Forsythe was frustrated about the work he had put into Mathematics since arriving at Stanford in 1957, especially guiding six or seven PhD dissertations in Mathematics.68
The politics of knowledge at the heart of the debates over Minsky and Miller provide a rich window to develop a framework to analyze the formation of disciplines. Academic legitimacy may often need to be redefined in new fields, and this can lead to particularly difficult challenges in developing support among the members of other disciplines, which are already established and viewed as legitimate. In the Stanford case of Computer Science, the discipline benefitted from strong university administration support that helped to create the conditions that increased its legitimacy within the university, and it is here that we turn to next.
Continue to Chapter 3, "The Computer Science Department and Entrepreneurial Culture" or return to Chapter 1, "Introduction"
- In this study, “field” and “area” will be considered synonyms of academic discipline.↩︎
- A readily readable account of this history is provided by E. O. Wilson, Consilience, Vintage Books, 1998.↩︎
- The term computer is used quite loosely to encompass the varied instrumentation available throughout this era.↩︎
- It is important to recognize here the role that theories of technological determinism play in the development of computer science. The increasing computational power played a necessary role in the development of the discipline, but it was not, I contend, a sufficient condition for the rise of the field↩︎
- Lloyd N. Trefethen,“The Definition of Numerical Analysis,” SIAM News, Nov. 1992.↩︎
- Iterative methods generally do not lead to a precise answer, but they approach the answer one small step at a time, and thus arrive arbitrarily close to the precise answer.↩︎
- James Varah, “The Influence of George Forsythe and His Students” part of The History of Scientific Computing, Ed. Stephen G. Nash, ACM Press, 1990.↩︎
- Donald E. Knuth, “George Forsythe and the Development of Computer Science,” Communications of the ACM, Vol. 15, No. 8 (Aug. 1972), pg. 721-727.↩︎
- Donald E. Knuth, “George Forsythe and the Development of Computer Science,” Communications of the ACM, Vol. 15, No. 8 (Aug. 1972), pg. 721-727.↩︎
- The idea of an autonomous division within Mathematics was a bureaucratic construction. The basic design was to place Forsythe as head of the division with his own budget, but with the chair of Mathematics continuing to hold final administrative authority. The change allowed Stanford to argue that it had an autonomous Computer Science unit.↩︎
- Gilbarg to Rhinelander, “Division of Computer Science,” 9 Jan. 1961, H&S Files, SC36/89-114/8/“CS: 62-63.”↩︎
- The academic backgrounds of these deans is important to this story, and receives fuller treatment in Chapter 3. Suppes to Gilbarg, “Division of Computer Science in the Department of Mathematics,” 1 Feb. 1961, H&S Files, SC36/89-114/8/“CS: 62-63.”↩︎
- Dartmouth was a major site of Computer Science in this era. John Kemeny, who studied under Alonso Church at Princeton and was an assistant to Albert Einstein, developed the BASIC computer language and would later become president of Dartmouth College.↩︎
- Royden would become Associate Dean later in 1962, and would be dean of H&S from 1973 to 1981. His membership on the A&P committee is cited from Sears to Gilbarg, “Computer Science Division Appointments,” 6 Apr. 1962, H&S Files, SC36/89-114/8/“CS: 62-63.”↩︎
- Royden to Sears, “Possible appointments in the Computer Sciences Division,” 6 Apr. 1962, H&S Files, SC36/89-114/8/“CS: 62-63.”↩︎
- Sears to Gilbarg, “Computer Science Division Appointments,” 6 Apr. 1962, H&S Files, SC36/89-114/8/“CS: 62-63.”↩︎
- Letter from Forsythe to Royden, 17 Dec. 1963, H&S Files, SC36/8/“CS: 64-65.”↩︎
- Lederberg to Sears and Robert Alway, “Appointment of Dr. Marvin L. Minsky,” 24 May 1963, H&S Files, SC36/89-114/8/“CS: 63-64.”↩︎
- “The Joshua Lederberg Papers,” Profiles in Science, National Library of Medicine, http://profiles.nlm.nih.gov/BB/↩︎
- Unfortunately, Royden does not describe who wrote these letters.↩︎
- Royden to Terman, “Marvin Minsky,” 6 Jun. 1963, H&S Files, SC36/89-114/8/“CS: 63-64.”↩︎
- Terman to Royden, “Marvin Minsky,” 3 Dec. 1963, H&S Files, SC36/8/“CS: 64-65.”↩︎
- Suppes to Royden, “Appointment of Marvin Minsky,” 7 Oct. 1963, H&S Files, SC36/89-114/8/“CS: 63-64.”↩︎
- Forsythe to Sears, 11 Oct. 1963, H&S Files, SC36/89-114/8/“CS: 63-64.”↩︎
- Gilbarg to Sears, “Minsky appointment,” 10 Oct. 1963, H&S Files, SC36/89-114/8/“CS: 63-64.”↩︎
- Sears to Vennard, 19 Dec. 1963, Terman Papers, SC160/3/12/1.↩︎
- Sears to Terman, “Recommended Appointment of Marvin Minsky as Professor of Computer Sciences,” 11 Nov. 1963, H&S Files, SC36/8/“CS: 64-65.”↩︎
- Minsky would later decline the offer, for reasons unclear, and remained at MIT for the entirety of his career.↩︎
- Gilbarg to Sears, “Computer Science Division,” 2 Dec. 1963, Terman Papers, SC160/3/12/1.↩︎
- Sears to Terman, “Mathematics and Computer Science Division,” 5 Dec. 1963, Terman Papers, SC160/3/12/1.↩︎
- Terman to Sears, “Mathematics and Computer Science Division,” 6 Dec. 1963, Terman Papers, SC160/3/12/1.↩︎
- “Biography of Seymour Parter,” Feb. 1963, H&S Files, SC36/89-114/8/“CS: 62-63.”↩︎
- Forsythe to Gilbarg and Sears, “Seymour Parter,” 12 Mar. 1963, H&S Files, SC36/89-114/8/“CS: 63-64.”↩︎
- Forsythe to Bowker, Gilbarg, Royden and Terman, “Seymour Parter declines,” 12 July 1963, Terman Papers, SC160/3/12/1.↩︎
- Essentially, no one was turned away from using the center, regardless of whether the work was sponsored or unsponsored.↩︎
- Computation Center to Hubert Heffner, “Incremental costs attributable to student computing,” 5 Mar. 1964, Terman Papers SC160/3/12/2.↩︎
- Forsythe to Marcia Ascher, 18 Nov. 1963, Forsythe Papers, SC98/2/5.↩︎
- Forsythe, “Stanford University’s Program in Computer Science,” Technical Report CS26, 25 Jun. 1965, ftp://reports.stanford.edu/pub/cstr/reports/cs/tr/65/26/CS-TR-65-26.pdf↩︎
- Forsythe notes, “Education Panel,” 4 Sept. 1965, Forsythe Papers, SC98/2/50.↩︎
- Forsythe to SLAC File, 15 Jan. 1964, Terman Papers, SC160/3/12/2; Terman’s support is noted in Forsythe, “Final Conversation with Bowker,” 25 Sept. 1963, Forsythe Papers, SC98/2/17.↩︎
- For a history of the physics community, see Daniel Kevles, “The Physicists,” Harvard University Press, 1995.↩︎
- Forsythe to Leland Haworth, 25 Nov. 1964, Forsythe Papers, SC98/15/2.↩︎
- Forsythe to Whitaker, “Ph.D. in Computer Science,” 29 July 1964, Forsythe Papers, SC98/15/4.↩︎
- Forsythe, Abstract of Presentation to AAAS Berkeley, Dec. 1965, Forsythe Papers, SC98/14/24.↩︎
- Paul Flory to Dean Robert Sears, “Committee on Appointments and Promotions,” 29 Jun. 1964, Terman Papers, SC160/3/12/2.↩︎
- “Paul J. Flory - Autobiography,” Nobelprize.org, 22 Jan. 2011, http://nobelprize.org/nobel_prizes/chemistry/laureates/1974/flory-autobio.html↩︎
- Of course, this linear model of the progress of theory to applied science has been expanded by a wide range of scholars. See Stokes, Donald “Pasteur’s Quadrant” for an encompassing discussion of these models↩︎
- Forsythe, “Stanford University’s Program in Computer Science,” Technical Report CS26, 25 Jun. 1965, ftp://reports.stanford.edu/pub/cstr/reports/cs/tr/65/26/CS-TR-65-26.pdf↩︎
- Forsythe to Royden File, “Conference 22 Jan. 1965,” 25 Jan. 1965, Forsythe Papers, SC98/15/1.↩︎
Continue to Chapter 3, "The Computer Science Department and Entrepreneurial Culture" or return to Chapter 1, "Introduction"