Allen Newell | Biography

Allen Newell, an expert on how people think and a developer of complex information processing programs, was a pioneer in the field of artificial intelligence. From his development in the 1950s of Logic Theorist, one of the initial forays into artificial intelligence, to his presentation of the sophisticated problem-solving software system know as "SOAR" in the 1980s, Newell worked to link computer science and advances in understanding human cognition.

Newell was born in San Francisco on March 19, 1927, the son of Robert R. and Jeannette (LeValley) Newell. Robert Newell, a professor of radiology at Stanford Medical School, had a strong influence on his son. "[My father] was in many respects a complete man," Newell told Pamela McCorduck in an interview reported in Machines Who Think. "We used to go up and spend our summers on the High Sierra. He'd built a log cabin up in the mountains in the 1920s. And my father knew all about how to do things out in the woods--he could fish, pan for gold, the whole bit. At the same time, he was the complete intellectual.... My father knew literature, all the classics, and he also knew a lot of physics." Newell told McCorduck, however, that his own desire for scientific achievement had led him to focus his interests much more narrowly than had his father.

Newell served for two years on active duty in the Naval Reserve during World War II. In 1947, he married Noel Marie McKenna; they would have one son, Paul Allen Newell. After obtaining his B.S. in physics from Stanford University in 1949, Newell spent a year at Princeton University doing post-graduate work in mathematics, then went to work in 1950 as a research scientist for the RAND (Research and Development) Corporation in Santa Monica, California.

While at RAND, Newell worked with the Air Force to simulate an early warning monitoring station with radar screens and a crew. His need to simulate the crew's reactions led to his interest in determining how people think. Working together throughout the 1950s and into the 1960s, Newell and his colleagues Herbert A. Simon and Clifford Shaw were able to identify general reasoning techniques by observing the problem-solving behavior of human subjects. One of the best known of these techniques is means-ends analysis, a process that analyzes the gap between a current situation and a desired end and searches for the means to close that gap.

In order to make use of computers in studying problem-solving behavior, Newell, Simon, and Shaw observed individuals as they worked through well-structured problems of logic. Subjects verbalized their reasoning as they worked through the problems. The three scientists were then able to code this reasoning in the form of a computer program. To make the program work, the scientists used a language called Information Processing Language (IPL) that they had developed previously for a computerized chess game. Their program, known as Logic Theorist, was not subject-matter specific; rather, it focused on the problem-solving process. Newell, Simon, and Shaw followed Logic Theorist with the development of General Problem Solver, a program that used means-end analysis to solve problems. Like Logic Theorist, General Problem Solver used the IPL language they had developed earlier.

During the summer of 1956, Newell and Simon were among a group of about a dozen scientists that gathered at Dartmouth College. The scientists came from a wide variety of fields, including mathematics, psychology, neurology, and electrical engineering. Though their backgrounds differed, they all had one thing in common: all were using computers in their research in an effort to simulate some aspect of human intelligence. With their Logic Theorist program, however, Newell and Simon were the only participants who could offer a working program in what would come to be known as "artificial intelligence." The Dartmouth Conference is generally viewed as the formal beginning of the field of artificial intelligence.

In 1957, Newell earned his Ph.D. in industrial administration from Carnegie Institute of Technology in Pittsburgh, Pennsylvania. In 1961 he left his position at RAND to join the faculty of Carnegie-Mellon University (formerly the Carnegie Institute of Technology), where he helped develop the School of Computer Science.

During the 1980s, Newell, along with his former students John Laird and Paul Rosenbloom, developed a more sophisticated software system that solved problems in a manner similar to the human mind. This system, called SOAR (State, Operator, and Result), was a general problem-solving program that learned from experience in that it was able to remember how it solved problems and to make use of that knowledge in subsequent problem-solving. SOAR, like humans, used working memory and long-term memory to solve problems. If SOAR was working toward a desired goal, it used working memory to keep track of the current situation, or "state," in the problem-solving process compared with the desired goal or "result." In order to make the decisions necessary to achieve a goal, people use information they have accumulated through experience. People use long-term memory to access information; SOAR also used long-term memory, programmed as a series of IF/THEN statements.

While the use of IF/THEN statements in a computer program wasn't a new idea, the way in which SOAR processed those statements was new. In the past, only one IF/THEN statement could control a computer program at any given time. If conflicting statements could apply to a problem, the problem-solving process would break down. SOAR, on the other hand, was designed to look at all of the programmed IF/THEN statements at once. After looking at all of the statements, SOAR would weigh them as suggestions, then decide which move, or "operator," would best advance it towards the desired result. If there were no IF/THEN statements stored in memory that applied specifically to the problem at hand, SOAR would use any available information that seemed potentially useful to try to resolve the problem. Whenever it solved one of these unexpected problems, SOAR would remember how it solved the problem, adding this information to its long-term memory. Like the human mind, SOAR was thus able both to generate original responses to new problems and to "learn" from its experiences.

In the late 1980s, Newell began an active campaign to promote the use of SOAR as the basis for a new effort to develop a unified theory of cognition. Whereas current research in artificial intelligence tended to focus on narrow and isolated aspects of cognition, Newell hoped SOAR would help cognitive psychologists devise broad theories of human cognition and advance towards an integrated understanding of all aspects of human thought.

Newell received the National Medal of Science from President George Bush just a month before his death from cancer on July 19, 1992. His work had already brought him a number of other honors, including the Harry Goode Memorial Award, which he received from the American Federation of Information Processing Societies in 1971, and the A. M. Turing Award, presented jointly to Newell and Simon by the Association for Computing Machinery in 1975. Newell was founding president of the American Association for Artificial Intelligence and also served as head of the Cognitive Science Society. Along with his colleague Herbert Simon and computer scientists Marvin Minsky and John McCarthy, he is considered one of the fathers of artificial intelligence.