Simon, Herbert Alexander

views updated Jun 08 2018

SIMON, HERBERT ALEXANDER

(b. Milwaukee, Wisconsin, 15 June 1916; d. Pittsburgh, Pennsylvania, 9 February 2001), administration, artificial intelligence, cognitive psychology, economics.

Simon’s lifelong passion was the study of decision-making and problem-solving. He examined these processes rigorously to advance the social sciences. Computer technology enabled him to investigate human cognition by simulating it. He was a pioneer in the field of artificial intelligence. His creative work in several disciplines led to many prestigious awards, including the 1978 Nobel Prize in economics.

Simon’s goal as a student was to become a mathematical social scientist. He earned a BA (1936) and a PhD (1943) in political science at the University of Chicago. His dissertation examined administrative decision-making was later published in book form, Administrative Behavior([1947] 1997). He wrote his dissertation while directing a research group at the University of California, Berkeley. After completing his dissertation, he joined the faculty at the Illinois Institute of Technology. During his appointment he also worked with the Cowles Commission of Research Economics at the University of Chicago. In 1949 he moved to Carnegie Mellon University where he was appointed the Richard King Mellon University professor of computer science and psychology. He was a prolific scholar there for more than fifty years.

Simon’s father, an electrical engineer, came to the United States from Germany in 1903. His mother, whose maiden name was Merkel, was a third generation American. Her ancestors immigrated from Prague and Köln. During his childhood Simon become fond of books, music, and the outdoors. From his uncle, Harold Merkel, an economist, he learned about the social sciences. Simon married Dorothea Pye in 1937. They had three children, Katherine, Barbara, and Peter.

Decision-Making Mid-twentieth accounts of decision-making relied heavily on idealizations about a decision-maker’s informational and cognitive resources. Standard idealizations gave agents unlimited cognitive capacity and ample data about their decision problems. Simon relaxed these idealizations to make progress toward a realistic theory of decision-making. His theory accommodated a decision-maker’s limited ability to analyze options. Taking rationality as a capacity for reasoning, Simon recognized that people have only bounded rationality. His theory also accommodated a decision-maker’s limited information about a decision problem. A person often does not know all the options available or have enough data for a careful analysis of options. Because gathering information is costly and because the time for resolving a decision problem is limited, becoming fully informed is impractical. Simon recommended not seeking an optimal decision but instead seeking a satisfactory decision. He called the

recommended decision procedure satisficing to contrast it with optimizing.

An agent with bounded rationality does not have all logical and mathematical truths at his or her fingertips to assist analysis of a decision problem. The agent’s inferential skills are imperfect, and a lack of analytical skill makes selecting an optimal option an unrealistic goal, as there are too many options to analyze and compare. Because requirements of rationality adjust to circumstances, a person may nevertheless decide rationally, despite these handicaps. Cognitive limits lower rationality’s requirements. Optimization is a goal of rationality, but a person with good excuses for not attaining that goal may still decide rationally.

Problems requiring a decision do not come with a tidy list of options and a precise assessment of options’ prospects. In a typical decision-requiring problem an agent has more options than he or she can grasp. For instance, the number of strategies for playing a chess game is enormous. A player cannot comprehend and review all strategies before making an opening move. Time and resources do not permit thorough analysis and comparison of strategies. Instead of following a decision procedure that yields an optimal decision, Simon held that a limited agent should adopt the first satisfactory option discovered. That is, the agent should satisfice. Someone selling a house may reasonably accept the first satisfactory offer. Drumming up an optimal offer would take a prohibitively large amount of time and other resources. What counts as a satisfactory decision depends on an agent’s aspiration level, that is, the agent’s realistic expectation. That level may change as an agent acquires information and assesses the results of past decisions.

Sometimes theorists distinguish between optimizing and maximizing utility. Optimization evaluates options with respect to full information and, according to some theorists, with respect to the agent’s objective interests. Utility is a measure of desirability, and utility maximization evaluates options with respect to information in hand and with respect to the agent’s subjective goals. To emphasize utility maximization’s reliance on probabilities of options’ outcomes, instead of certainty of their outcomes, theorists also call it expected utility maximization. According to a common principle, an option’s utility equals its expected utility—a probability-weighted average of the utilities of its possible outcomes. Utility maximization takes account of a decision-maker’s limited information about options’ consequences. It does not require an optimal decision but instead a decision expected to be optimal. In many cases such a decision is rational. A person may rationally make a decision after reasonable efforts to gather information even if he or she still lacks full information. A decision made without full information may nonetheless be fully rational. For example, a businessman makes a rational decision about traveling to an appointment if he takes a train scheduled to bring him to the meeting place on time, even if an unexpected delay on the rails causes him to miss his appointment.

For agents with limited information, utility maximization is attainable. For agents with additional limitations, are other types of maximization in reach? One interpretation of satisficing takes it as utility maximization under constraints. For example, a chess player with a limited amount of time for a move must make a decision before he or she can thoroughly assess all possible moves. The decision the player makes may maximize utility given the cost of delay, although it does not maximize utility in the absence of time constraints. Given more time, the player may have made a decision with better prospects. Perhaps a decision that satisfices is also a decision that maximizes utility under constraints concerning time and the like. The maximization may occur within the set of options the agent actually considers instead of within the set of all options, considered or not. Several theorists have explored this topic. Some, such as Sydney Winter (1964), conclude that satisficing is not equivalent to utility maximization under constraints.

A comparison of satisficing and maximizing utility requires distinguishing two types of decision principles. One type formulates a procedure that an agent may follow to reach a decision. Another type presents a standard for evaluating a decision. An observer may apply the standard of evaluation after the agent reaches a decision. Simon advanced satisficing as a decision-making procedure. Maximizing utility may also be taken as a procedure [in] for making decisions. Taken as a procedure, it has comparative steps that satisficing lacks and so is distinct from satisficing.

Simon distinguished procedural and substantive rationality. A decision meets procedural standards of rationality if the method of making the decision was rational and so, for example, employed sufficient deliberation. A decision meets substantive standards of rationality if its content fits the agent’s circumstances and so, for example, selects an act reasonable to perform in the agent’s situation. The procedure that generated the decision is irrelevant.

A rational decision procedure may yield a decision that is not substantively rational. The decision reached may be defective because of its content. Satisficing may be a rational decision procedure although it may yield a decision that falls short of a standard of substantive rationality. The first satisfactory option discovered may not maximize utility, for example. An agent may have good reasons to follow a shortcut procedure such as satisficing despite the risk of reaching a decision with a substantive defect. For example, a driver may have to make a snap decision about taking a freeway exit despite the risk that the decision does not maximize utility.

Conversely, an irrational decision procedure may yield a decision that is rational because of its content. Possibly a decision reached in irrational haste is by good fortune the same as the decision careful deliberation would have generated. Perhaps a student picks a career without deliberation but chooses the same career he or she would have chosen after thoughtful reflection and information gathering. Then the student violates a standard of procedural rationality but nonetheless meets a standard of substantive rationality.

Satisficing may be taken as a substantive standard of rationality. A decision meets that standard if it is satisfactory, regardless of the procedure that led to the decision. Utility maximization under constraints, taken as a substantive standard of evaluation, may be equivalent in some cases to satisficing, also taken as a substantive standard of rationality. The aspiration level a decision must reach to be satisfactory may adjust so that only options maximizing utility given the constraints count as satisfactory.

Although treating satisficing and utility maximizing as standards of substantive rationality brings them closer together, their applications still have different informational requirements. Discovering whether an option maximizes utility requires an account of the probabilities and utilities of options’ possible consequences. Discovering whether an option satisfices requires only a classification of options. Suppose that an agent does not make quantitative probability and utility assignments to options’ possible consequences, but still classifies options as satisfactory or unsatisfactory. Then the substantive standard of satisficing, but not the substantive standard of utility maximizing, applies to the agent’s decision.

Simon’s term bounded rationality is the rubric for many current research programs in the decision sciences. Theorists consider how cognitively limited agents may reasonably cope with decision problems. A good example is Ariel Rubinstein’s book, Modeling Bounded Rationality(1998).

Economics . Simon specialized in decision-making within administrative organizations. He recognized that reasonable executives of corporations may fail to maximize profits because they do not access all information, not even all available information, and so misjudge the effects, especially the long-term effects, of their decisions. His models of administrative decision-making gained credibility by acknowledging an administrator’s limited time for deliberation and limited capacity to discover options and to acquire information about their consequences. Simon looked for efficient, time-preserving methods of achieving acceptable economic objectives while at the same time, reducing risks. He also recognized that factors independent of an organization’s goals contribute to decision-making within the organization. Whether an administrator makes a decision that advances the organization’s goals often depends on whether he or she identifies with the organization. Promoting that identification makes an important contribution to successful decisions within an organization.

Herbert Simon’s ideas are also influential in behavioral economics, which examines methods people use to make economic decisions. Their methods may result in systematic errors. They may, given a certain triggering event, apply a heuristic outside its successful range of application. That is, they may follow a shortcut procedure for making a decision in a context where the shortcut is unreliable. For example, a person may follow an expert’s advice on a topic outside the expert’s area of specialization. Although Simon’s decision principles are normative, their attempt to set realistic standards draws attention to actual decision processes, which in some cases yield decisions falling short of the appropriate norm.

Drawing on his prodigious mathematical skills, Simon also made major contributions to mathematical economics, especially general equilibrium theory and econometrics. With David Hawkins, he proved the Hawkins-Simon Theorem. It states conditions for the existence of positive solution vectors for matrices representing the input and the output of an economic system.

Artificial Intelligence To study problem solving, Simon turned to computer simulations of human cognition. His path-breaking work stimulated research in the field of artificial intelligence. With Allen Newell, he produced in 1956 a machine capable of proving theorems of formal logic. The following year, he and Newell invented a general problem-solving machine. His program BACON simulates the process of scientific discovery. It proposes a law governing a phenomenon, compares its proposal with reality, and makes adjustments. His book Scientific Discovery (1987) describes the program’s operation. In Simon’s eyes, computers running problem-solving programs are thinking machines. His book with Newell, Human Problem Solving (1972), is a classic in the literature on artificial intelligence. The Association of Computing Machines awarded Simon the Turing Medal in 1975.

The breadth and depth of Simon’s research is astonishing. He won top honors in a variety of disciplines. Besides awards already mentioned, he was a member of the National Academy of Sciences, received the National Medal of Science (1986), and won the American Psychological Association’s Award for Outstanding Lifetime Contributions to Psychology (1993). He was a brilliant twentieth-century scientist.

BIBLIOGRAPHY

WORKS BY SIMON

Models of Man: Social and Rational; Mathematical Essays on Rational Human Behavior in a Social Setting. New York: Wiley, 1957. Essays presenting mathematical models of human behavior.

The Sciences of the Artificial, 3rd ed. Cambridge, Massachusetts: MIT Press, 1997.

With Allen Newell. Human Problem Solving. Written with Allen Newell. Englewood Cliffs, NJ: Prentice-Hall, 1972.

Models of Discovery: And Other Topics in the Methods of Science. Boston: D. Reidel Publishing Company, 1977.

Models of Thought. 2 vols. New Haven, CT: Yale University Press, 1979. Essays on psychology, human information-processing, and problem-solving.

Models of Bounded Rationality. Volumes 1 and 2. Cambridge, MA: MIT Press, 1982. Essays on decision-making.

With Pat Langley, Gary Bradshaw, and Jan Zytkow. Scientific Discovery: Computational Explorations of the Creative Process. Cambridge, MA: MIT Press, 1987.

Administrative Behavior: A Study of Decision-Making in Administrative Organizations, 4th ed. New York: The Free Press, 1997.

Models of Bounded Rationality, volume 3. Cambridge, MA: MIT Press, 1997.

The Carnegie Mellon University Herbert A. Simon Collection has the complete corpus of Simon’s work.

OTHER SOURCES

Augier, Mie, and James March, eds. Models of Man: Essays in Memory of Herbert A. Simon. Cambridge, MA: MIT Press, 2004.

Byron, Michael, ed. Satisficing and Maximizing. Cambridge, MA: Cambridge University Press, 2004. A collection of essays reviewing Simon’s ideas about satisficing.

Courtois, Pierre Jacques. Decomposability: Queuing and Computer Systems Applications. New York: Academic Press, 1977. Continues the work of Simon and Albert Ando on decomposable computer systems.

McCorduck, Pamela. Machines Who Think. San Francisco: W. H. Freeman, 1979. Presents Simon’s contributions to artificial intelligence.

Rubinstein, Ariel. Modeling Bounded Rationality. Cambridge, MA: MIT Press, 1998.

Weirich, Paul. Realistic Decision Theory: Rules for Nonideal Agents in Nonideal Circumstances. New York: Oxford University Press, 2004. Pursues Simon’s program of making decision principles realistic.

Winter, Sydney. “Economic ‘Natural Selection’ and the Theory of the Firm.” Yale Economic Essays 4 (1964): 225–272. Compares satisficing and optimizing.

Paul Weirich

Herbert Alexander Simon

views updated Jun 08 2018

Herbert Alexander Simon

The study of decision-making behavior, especially in large organizations, led Herbert Simon (born 1916) to develop new theories in economics, psychology, business administration, and other fields. He was awarded the Nobel Prize in economics in 1978. He was also the first social scientist elected to the National Academy of Sciences.

Herbert Alexander Simon was born in Milwaukee, Wisconsin, on June 15, 1916. He received an A.B. from the University of Chicago in 1936 and a Ph.D. in 1943. He stayed on at Chicago for two years as a research assistant before becoming a staff member of the International City Managers Association and assistant editor of the Public Management and Municipal Year Book (1938-1939). The following year he joined the University of California as director of administrative measurement studies. After a teaching post at the Illinois Institute of Technology (1942-1949), Simon joined the teaching staff of the Carnegie-Mellon University, first as professor of administration and psychology (1949-1955) and later as professor of computer science and psychology (1956 to the mid-1980s).

In his work Simon brought greater realism to neoclassical economic models, which he found to be lacking because of their idealized vision of the "rational" consumer, businessperson, or worker. Instead of maximizing their welfare, profits, or wages on the marketplace, Simon believed that lack of information about alternatives and the impossibility of foreseeing the future makes all of these participants "satisficers." Their rational behavior is "bounded" by the cost of obtaining information and uncertainty; hence Simon proposed the concept of "bounded rationality." That is, economic agents try to do as well as possible given the constraints, but these constraints keep them from ever achieving what neo-classical economists would call a "maximum" (of profits, for example). Simon argues that individuals would be acting rationally by "satisficing," given real world circumstances.

The notion of "bounded rationality" is explained by analogy to the search for a needle in the haystack. The neoclassical approach would be to search for the needle in the stack (a maximization process). Simon's approach is to find the needle which is sharp enough to handle the contemplated sewing tasks (a "satisficing" process).

In another example, consider a chess game: every move involves potentially millions of calculations about alternative actions. Since it is impossible for players to examine all the possibilities, they learn to follow promising lines of play and to utilize "rules of thumb" in decision-making. Over time these rules of thumb change as outcomes are evaluated.

Simon's views on rationality have been expounded in numerous books and articles, including Models of Man (1956), Human Problem Solving (with Allen Newell, 1972), The Sciences of the Artificial (1969), Models of Discovery (1977), and Models of Bounded Rationality and Other Topics in Economic Theory (1982).

Simon also disputes whether economic models centered on "equilibrium" solutions are useful or accurate. The idea of equilibrium derives from the science of mechanics and was adapted to economic problems by neo-classical economists of the late 19th century. Most modern American economists until the mid-1970s also utilized this methodology. Simon, in his Richard T. Ely Lecture to the American Economic Association in 1978, argued that: "when the system is complex and its environment continually changing (that is, in the conditions under which biological and social evolution actually takes place), there is no assurance that the system's momentary position will lie anywhere near a point of equilibrium."

Simon made other significant contributions to economic analysis. The Hawkins-Simon theorem (1949) contains a powerful test for the sustainability of an economy as measured by input-output tables. In the area of production scheduling Simon co-authored the "Certainty Equivalent" theorem (1956, 1960), which provided practical help to businesses concerned with the needs for labor and inventory when demand fluctuates.

In spite of his own mathematical prowess, Simon sought to break economic methodology out of the rigorous mathematical modeling which requires strong assumptions and quantifiable data into a broader arena of qualitative analysis using interdisciplinary theories. Indeed, he believed economists have much to learn from other social sciences and in his own career he drew widely from them. Much of his writing dealt with issues in psychology as applied to organizations, or what Simon called "micro-micro-economics." To promote these views Simon, along with colleagues at Carnegie-Mellon, founded The Journal of Organizational Behavior. Simon's textbook Administrative Behavior was first published in 1947 and became a classic in the field, going through several editions.

Simon was a consultant to the International City Managers Association (1942-1949), the U.S. Bureau of the Budget (1946-1949), the U.S. Census Bureau (1947), and the Cowles Commission for Research in Economics (1947-1960); chairman of the board of directors of the Social Science Research Council (1961-1965); member of the President's Scientific Advisory Committee (1969-1971); chairman of the Committee on Air Quality Control of the National Academy of Sciences (1974); chairman of the Committee on Behavioral Sciences of the National Science Foundation; winner of the Award for Distinguished Scientific Contributions of the American Psychological Association (1969), and Distinguished Fellow of the American Economic Association (1976). He lectured extensively around the world and received nine honorary degrees.

For his many and diverse contributions Herbert Simon was awarded the Alfred Nobel Memorial Prize in Economics in 1978. Yet the label "economist" is far too narrow for this extraordinary social scientist and philosopher. While not a household name, Simon is still widely-read and has had a profound influence on the underpinnings of nearly every social science. Often referenced in both the abstract as well as the specific, some of Simon's views were discussed in 1996 by Herbert Kaufman in his acceptance of the Dwight Waldo Award of the American Society for Public Administration (ASPA), of which Simon is a previous recipient.

Further Reading

Further information on Herbert Simon can be found in articles by two leading economists in H. W. Spiegel and W. J. Samuels (editors), Contemporary Economists in Perspective (1984), and Mark Blaug, Great Economists Since Keynes (1985). Simon's own autobiographical work, Models of My Life (1991), received generally favorable reviews. □

Simon, Herbert A

views updated May 29 2018

Simon, Herbert A.

American Professor of Computer Science and Psychology
19162001

Herbert A. Simon combined the study of social and behavioral science with the disciplines of mathematics, physics, and economics in a career that included a longtime focus on the science of decision-making in organizations. Of particular note is his analysis of decision-making and problem-solving, but he was also interested in artificial intelligence (AI) and the use of the computer to study intelligence and cognition, both in problem-solving, such as the discovery of theorems, and in game playing, such as chess.

Simon was born in Milwaukee, Wisconsin, on June 15, 1916. His father was an electrical engineer and his mother an accomplished pianist. He enrolled at the University of Chicago in 1933 and graduated in 1936 with a degree in political science. He received his doctorate through the University of Chicago in 1943 while heading a research group at the University of California, Berkeley, between 1939 and 1942. He taught at the Illinois Institute of Technology from 1942 to 1949, and he engaged in research with colleagues at the University of Chicago and the Cowles Commission for Research in Economics. His next professional post was at the Carnegie Institute of Technology (now Carnegie Mellon University), where he helped build the Graduate School of Industrial Administration.

Simon's career in Pittsburgh as an academic, researcher, and author spanned more than fifty years. He was well respected by colleagues and students. He believed that the approach of the "hard" sciences, such as physics and mathematics, could be applied to the behavioral sciences, both in economics and political science, his first field of study, and the behavioral sciences, primarily psychology and cognitive science.

One of Simon's earliest books, published in 1947, was Administrative Behavior. The book was an expansion of his doctoral dissertation, which began his studies of rationality. Later publications include Models of Man (1957), The Sciences of the Artificial (1969), Human Problem Solving, with Allen Newell (1972), and Models of Discovery (1977), among others. In 1991 he published an autobiography, Models of My Life.

Simon firmly believed that the computer could and should aid in the study of human cognition and, ultimately, that what the computer could do in terms of cognition was "think." He considered the computer to be a laboratory for epistemology, the study of knowledge or truth, as well as a tool for investigating the human mind. In 1954 Simon began using computers to model problem-solving.

Simon developed what he termed the theory of "satisficing," that is, the making of decisions on the basis of a satisfactory rather than optimal (absolute best) solution. This is a technique familiar to anyone who has done even such a routine task as develop a schedule of college courses for a term. One must make choices that meet certain requirements for one's degree, balancing other factors such as personal preferences for times of classes, subjects one is interested in, distance to and from classes, and cost to create a satisfactory, albeit possibly imperfect, schedule.

Simon studied "bounded rationality," the theory of making rational decisions under constraints such as a lack of knowledge, computational difficulty, and personal and social circumstances. The decisions are rational, but not in the sense of an all-knowing, infallible optimizer. This leads to finding acceptable, but not necessarily optimal, solutions to problems.

From 1966 until his death on February 9, 2001, Simon was Richard King Mellon University Professor of Computer Science and Psychology. He won the Nobel Prize in Economics in 1978 for "pioneering research into the decision-making process within economic organizations." He was awarded the National Medal of Science in 1986 and the A.M. Turing Award by the Association of Computing Machinery (ACM) in 1975, with Allen Newell (19271992). He collaborated with Newell and Clifford Shaw to write a computer program, the Logic Theorist, or the Logic Theorem Machine, designed to find logical proofs. Together, the three also collaborated on a software program designed to play chess as a human, not an expert. He was involved in several computer projects to study human cognition and form models of human learning, problem solving, and "thinking" using computer programs.

see also Artificial Intelligence; Chess Playing; Decision Support Systems; Newell, Allen.

Roger R. Flynn

Bibliography

Newell, Allen, and Herbert A. Simon. Human Problem Solving. Englewood Cliffs, NJ: Prentice-Hall, 1972.

Simon, Herbert A. Models of My Life. New York: Basic Books, 1991.

Internet Resources

Simon, Herbert A. Autobiography. The Nobel E-Museum. <http://www.nobel.se/economics/laureates/1978/simon-autobio.html>

Herbert Alexander Simon

views updated May 14 2018

Herbert Alexander Simon

1916-

American computer scientist and economist who was awarded the 1978 Nobel Prize in Economics for his research into decision-making processes within economic organizations. Simon has investigated the intellectual processes behind decision-making in an effort to help construct computer programs that can replicate human thought processes. Along the way, he helped develop list processing computer languages that are commonly used among artificial intelligence researchers.