Computing and the National Science Foundation, 1950-2016. William Aspray

of infrastructure. McCarthy started working at BBN with JCR Licklider and others at around that time, and it is said that McCarthy influenced Licklider’s thinking about time-sharing. Licklider later went to ARPA, where he funded Project MAC at MIT, based on CTSS, and many other important initiatives.

      NSF established the Office of Computing Activities (OCA) in July 1967 to provide federal leadership in the use of computers for research and education. Later, the directive was added as a statutory requirement to the NSF charter. Faced with ever-increasing demand for computing facilities from all sectors of academe, OCA established regional centers. In fiscal years 1968 and 1969, the Foundation explored various computer-based cooperative arrangements. Typically, each regional activity was centered on a major university, which provided computer services and technical assistance to help a cluster of nearby institutions introduce computing. Altogether, 15 regional centers were established, including 12 major universities, 116 participating colleges, 11 junior colleges, and 27 secondary schools located in 21 states. By the early 1970s, 30 regional computing networks were connecting approximately 300 institutions at all levels of education and including minority institutions.

      As the number of college and university computing centers grew, NSF also began to recognize the need for programmers and technicians to staff these centers. In its 1957 Annual Report, it noted:

      The rapid development of computing machines and their usefulness in a wide variety of research investigations have created a demand for persons trained in the use and operation of computers. Although such training may be considered a proper responsibility of colleges and universities, there is a severe shortage of teachers competent to give instruction. The Foundation has provided support for a program of training for experienced mathematicians on the faculties of colleges and universities to prepare them to develop courses of instruction in the use and operation of modern computing machines.

In 1954, Wayne University had held a Conference on Training Personnel for the Computing Machine Field⁶⁹ with a focus on educating mathematicians and on scientific rather than business applications of computing. Participants in the 1954 NSF-funded meeting identified a large but unspecified demand for people highly skilled in computation; however, the attendees were unsure whether the primary use of computers was for scientific calculations or business calculations. Educating the needed workforce led to the conclusion that there were “not enough mathematicians.”⁷⁰ Leon W. Cohen, the program director for Mathematical Sciences, made the first public announcement of NSF’s support for computing infrastructure at this meeting.⁷¹

By 1957, NSF was providing support for training experienced mathematicians on the faculties of colleges and universities to prepare them to develop courses of instruction in the use and operation of modern computing machines.⁷² This activity formed the basis for creating academic computer science programs. Training programs continued with the Office of Computing Activities created in 1967.

1.3Computers in Education

While the Rosser Report said very little about the use of computing for education, the issue did not go away. The President’s Science Advisory Committee (PSAC) commissioned another study of computers in higher education in 1967, chaired by physicist John Pierce of Bell Labs. Following extensive hearings, the committee concluded that “an undergraduate college education without adequate computing was as deficient as an undergraduate education would be without an adequate library . . . [and that] there was value in using computers for precollege education.”⁷³ The Pierce Report’s focus on education supported NSF’s expanded involvement.

      Andrew Molnar, a leader in the computing education field, asserted that:

The most significant event [related to computers in education] occurred when President Lyndon Johnson . . . directed the National Science Foundation to work with the U.S. Office of Education to establish an experimental program to develop the potential of computers in education. In response to the directive, NSF created the Office of Computing Activities (OCA) in July of 1967 to provide Federal leadership in the use of computers for research and education.⁷⁴

      When OCA was created, Molnar moved over to the NSF from the Department of Education, first on detail and later as a program director, to work on the computers in education programs.

NSF has a long history of involvement in early efforts to use computers for education. It funded three pioneers⁷⁵ in educational technology projects: The Children’s Television Workshop,⁷⁶ the computer-based learning system PLATO, and the curriculum sharing network CONDUIT.

      PLATO, the first large-scale, computer-based education system, was developed at the University of Illinois at Urbana-Champaign under the guidance of Donald Bitzer beginning in 1959. With NSF support, Bitzer showed that computers could serve thousands of students, at many different geographic locations, with hundreds of courses, at a reasonable cost. Most of the financial support for PLATO initially came from NSF. Control Data Corp. (CDC) was eventually licensed by the University of Illinois to produce and market the PLATO system.

      One unique feature of the PLATO system was a plasma display that provided high quality, low-cost graphics. The PLATO authoring language helped educators create thousands of instructional programs. Bitzer eventually moved PLATO to a Control Data 6000-class machine that served several thousand student stations and provided hundreds of lessons simultaneously. When distributed by Control Data Corporation, PLATO primarily was used for in-service training in industry, but it continued in use in many universities and secondary schools through the 1980s.

James Johnson at Iowa, Gerald Weeg at Iowa, Thomas Kurtz at Dartmouth, and Jim Parker at North Carolina Educational Computing Service, together with representatives from Texas and Oregon State, formed CONDUIT, a consortium of five regional networks involving approximately 100 colleges and universities for sharing computer-based curricula in seven fields of science.⁷⁷ In 1971, when CONDUIT was conceived, the major barrier to instructional computing was a lack of quality learning materials and computer software. CONDUIT faced significant challenges in validating shared curricula,⁷⁸ but the concept of regional networks would return as a critical part of the NSFNET project.

In addition to computer-aided instruction (CAI) systems such as PLATO and CONDUIT, NSF had an uneven but long history with some of the leaders in the cognitive and learning sciences. As Molnar stated,⁷⁹ “no other name is more closely connected to computer-assisted instruction (CAI) than that of Patrick Suppes.” As Director of the Stanford Institute for Mathematical Studies in the Social Sciences, Suppes began a program of research and development in computer assisted instruction in 1963. He and Richard C. Atkinson, who later would become NSF Director, developed sophisticated mathematical models of student learning to help design instructional materials and strategies.⁸⁰ Suppes noted that John McCarthy of Stanford’s computer science department (having moved from MIT) played an important role in the design and operation of the institute’s computer facilities. Suppes wanted to demonstrate that computers could have an immediate impact on education, even using existing equipment. He and Atkinson began initially with 12 six-year-old children who came to their lab daily and spent 30 minutes at the computer. From 1966 to 1968, Suppes used an IBM 1500 and an audiotape device for CAI. Students responded to questions displayed on a CRT via light pen and keyboard. Suppes later developed a wide variety