Herbert Alexander Simon (1916-2001)

No other single person has won the Nobel Memorial Prize for Economics (1978), the Turing Award of the Association for Computing Machinery (Newell and Simon 1972), the Orsa/Tims John von Neumann Theory Prize (1988), the Distinguished Scientific Contribution Award of the American Psychological Association (1969) and the National Medal of Science (1986).

Perhaps this renaissance man’s outlook on the scientific spirit he nurtured can be summarized in his characterization of ‘the central task of natural science’ to be (Simon 1996: 1-2):

[T]o make the wonderful commonplace: to show that complexity, correctly viewed, is only a mask for simplicity; to find pattern hidden in apparent chaos.... This is the task of natural science: to show that the wonderful is not incomprehensible, to show how it can be comprehended - but not to destroy wonder. For when we have explained the wonderful, unmasked the hidden pattern, a new wonder arises at how complexity was woven out of simplicity. The aesthetics of natural science and mathematics is at one with the aesthetics of music and painting - both inhere in the discovery of a partially concealed pattern.

In an academic life spanning more than six decades, Simon managed, almost singlehandedly, to create the wholly new disciplines of behavioural economics and the cognitive sciences and nurture through to growth and prosperity one of the great academic institutions, the Graduate School of Industrial Administration (GSIA) at the Carnegie Institute of Technology (now the Carnegie Mellon University - CMU) in Pittsburgh, where these disciplines were fostered with immense dedication. The eminent proof theorist, metamathematician and philosopher, Wilfried Sieg, recalled during a memorial event that Simon was also instrumental in the creation of a department of philosophy at CMU and went on to observe:

How intimately theoretical issues and practical affairs were intertwined for Herb! Having discussed some difficult administrative problems with him, he remarked without further explanation: Proceed, as if you were proving a theorem! I followed his advice and, lo and behold, it worked: proving a theorem requires, after all, to look at the problem from a variety of perspectives and to expend lots of patience.

(Sieg 2001, emphasis added)

Of course, Simon would not have suggested this metaphor to a Bourbakist or an Hilbertian Formalist, but only to a proof-theorist like Sieg, who would have thought of a proof as a procedure, in the same sense in which Simon extolled the virtues of procedural decision-making in all aspects of administrative and economic life. Incidentally, Sieg’s colleague in the philosophy department at CMU, Teddy Seidenfeld, is the current holder of the Herbert Simon Professorship of Philosophy and Statistics.

Herbert Simon (born 15 June 1916, Milwaukee, Wisconsin, died 9 February 2001, Pittsburgh, Pennsylvania) was the second son of Arthur Carl Simon, a German emigre electrical engineer, inventor and patent lawyer, and Edna Marguerite Merkel, pianist and a second-generation descendant of immigrants from Prague and Cologne. He was wholly Jewish on his father’s side; partly Lutheran on his mother’s. His brother Clarence was five years older. He was introduced to Dorothea Isabel Pye by William W. Cooper (of “Charnes-Cooper” fame and coincidentally also the man who helped establish the School of Urban and Public Affairs at CMU, serving also as its Dean from 1969 to 1975) and they married on Christmas day in 1937. Their daughter Kathie was born in 1942, followed by son Peter in 1944 and another daughter Barbara in 1946.

From the public elementary and high schools in Milwaukee, he won a full scholarship ($300 per year) to the University of Chicago, taking the examination in physics, mathematics and English. His maternal uncle, Harold Merkel, “an ardent formal debater [whom] I followed in that activity too”, who had died young, at the age of 30, in 1922, was an early intellectual influence. Uncle Harold had graduated with distinction in law from the University of Wisconsin, having also studied economics under the legendary John R. Commons and leaving behind copies of The Federalist Papers and William James’s Psychology in the family library, both of which were devoured by the young Herbert, leaving indelible impressions on the future civil libertarian, behavioural economist, computer scientist and cognitive psychologist.

His first publication was a letter to the Editor of the Milwaukee Journal, defending atheism.

To buttress his debating skills he began to read widely in the social sciences. Two books in particular were decisively influential: Richard T. Ely’s Outlines of Economics (1893) and Henry George’s Progress and Poverty (1879). By the time he was ready to embark upon a university career, he had developed a clear sense of the general direction he intended to take in his studies. He would devote himself to becoming a “mathematical social scientist”. We cannot imagine anyone else encapsulating and wearing this mantle with more grace and justification.

He obtained his BA in political science from Chicago in 1936 and a PhD in 1943 and decided to major in political science because his first choice of major, economics, required him to take an obligatory course in accounting, which he detested.

The undergraduate-term paper, written for graduation, led to a research assistantship at the Milwaukee City Government in the field of Municipal Administration, which in turn led to a Directorship at the Bureau of Public Measurement in the University of California at Berkeley, from 1939 to 1942. For Milwaukee he undertook a study of how the municipal employees made budget decisions, for example, when deciding between planting trees and hiring a recreation director. From this work grew his PhD thesis that, subsequently, became one of the fountainheads for the whole field of organization theory: Administrative Behavior, published first in 1947 and still in print.

Simon’s main intellectual impulses during the Chicago years came from Henry Schultz in mathematical economics and econometrics, who was also a mentor, from Rudolf Carnap, in the philosophy of science, from Nicholas Rashevsky, in mathematical biophysics and from Harold Lasswell and Charles Merriam, in political science.

Simon observed, when writing an appreciation of Allen Newell, that the four great questions of human intellectual endeavour are those on the nature of matter, the origins of the universe, the nature of life and the working of the mind.

There is little doubt that he himself devoted the whole of his professional life to various aspects of the problem of the working of the mind. How does the mind perceive the external world? How does perception link up with memory? How does memory act as a reservoir of information and knowledge in interacting with the processes that are activated in human decision-making, in individual and social settings? In short, human problem solving (the title of his 1972 book with Newell): in the face of internal constraints emanating from the working of the mind; and constraints, imposed on its workings, by the external, perceived, world.

Much is made these days of “ecological rationality” and “ecological cognitive computing”, by which is meant that any internal constraints on the working of the mind should be taken in conjunction with the “external constraints of the environment” in which the mind is situated for its interaction with the external world. Simon’s definition of bounded rationality - the term, although not the concept, was introduced in the Introduction to part IV of Simon (1957: 198), while the analytical content, together with its conceptual underpinnings, were fully developed in Simon (1955, 1956) - from the outset, was to encapsulate both of these aspects of the workings of the mind, in its rational, decision-making, incarnation.

Nothing, and no one, in the burgeoning field of modern behavioural economics (see Velupillai and Kao 2014 for the definitions of modern and classical behavioural economics) seem to have ever underpinned any of their theories on a model of computation. Thus they are unable to comprehend the nature of the decision problem framework, intrinsically framed with an underlying algorithmic basis, within which Simon first advanced, and then developed, boundedly rational and satisficing decisions.

George Polya’s influential little book How to Solve It? (Polya 1945) introduced generations of students to heuristics - the art of guided search.

Simon and Newell (1983) felt that the Polya framework provided a starting point for investigating, experimentally, the creative aspects of the workings of the mind in two formally and rigorously definable areas - human problem solving and in the art of discovery. From lessons that could be learnt in understanding the formal aspects of human problem solving they felt they could move on to more ambitious tasks: to an understanding of human thinking, in general. From there it would be a natural step to a formal understanding of the underpinnings of human decision-making in general.

The art of discovery, based on heuristics as guided search, led Simon to develop Models of Discovery (1977), resurrecting the Peirce-Hanson (Hanson 1958) emphasis on retroduction (or abduction; thus, circumventing the tiresome dichotomy between induction and deduction) and, simultaneously, taking a well-aimed attack on the nihilism in Karl Popper’s stance that there was no scientific basis for a “logic of scientific discovery” (see, in particular, Simon, 1977: ch. 5.4).

Simon used “experimentally” in the sense of exploring by computer simulation, guided by programmed heuristics. These were, for Simon, algorithms that were not necessarily constrained by the Church-Turing thesis of computability theory. In this sense his lifelong adherence to heuristic methods in human problem-solving, models of discovery, design of organizations and evolutionary dynamics, had more in common with the notion of algorithms as proofs in constructive mathematics. A lack of appreciation of this subtle difference may have been the reason for even the great Hao Wang to be critical of the way Newell, Shaw and Simon automated proofs in Principia Mathematica, in contrast to the way he had done the same (Wang 1970: 227; emphasis added): “There is no need to kill a chicken with a butcher’s knife. Yet the net impression is that Newell-Shaw-Simon failed even to kill the chicken with their butcher’s knife To argue the superiority of

'heuristic' over algorithmic methods by choosing a particularly inefficient algorithm seems hardly just.” What is a heuristic, if not an algorithm?

Simon was a member of the Cowles Foundation for Economic Research in its early Chicago days, before its move to Yale in New Haven; he was also a member of the Rand Corporation in its glory days, the early 1950s.

The former nurtured, in Simon’s own words, the econometric “mafia”; the latter fostered the mathematical economics “mafia”. He remained a gadfly inside these citadels of orthodoxy while enjoying the respect, perhaps even the envy, of his distinguished and eminent peers.

His contributions to formal and traditional economic theory - both to micro and macro variants - and to econometric theory were fundamental and path-breaking. At a very early stage in the mathematization of economics he deduced, in joint work with the mathematician David Hawkins (Hawkins and Simon 1949), conditions for stability, which came to be known as the Hawkins-Simon conditions in the folklore of the subject, for linear multi-sectoral models of the economy. This led to an amusing episode with the House Un-American Committee hearings, during the “McCarthy era”, because Hawkins - whom Simon had never met and with whom he had written the famous paper entirely by correspondence - was a paid-up member of the Communist Party.

During the Great Depression, Simon had seen a chart on the walls of his father’s study, tracking the dismal progress of a faltering American economy. This chart was constructed on the basis of a model of the macroeconomy and its flows built on the principles of servomechanism theory, using hydrodynamic analogies devised by an imaginative engineer, with a doctorate in sociology, A.O. Dahlberg. Inspired by the memories of the Dahlberg chart in his father’s study, he began to look at the economy from the point of view of the theory of servomechanisms and feedback control. This line of research led him to his celebrated results on certainty equivalence in the devising of optimal policy in decisions on production scheduling in firms. He did not pursue the servomechanism metaphors for too long, because he felt, by then, that analogue simulations were a distinct second best to the digital possibilities he was pioneering.

The origins of the inspiration that led to the influential work with his eminent Japanese student Yuji Ijiri on the size distributions of the growth and decay of business firms and organizations are narrated with humour and candour in his charming autobiography, Models of My Life (Simon 1991). It is also a tale of academic bloody-mindedness, recounted without rancour, and revisited with nostalgia and regrets on the fallibility of memory.

In 1946 Richard Goodwin (1947) had begun interpreting economic agents, markets and the economic system as (nonlinear) oscillators, analysing markets as coupled oscillators with hierarchies of coupling strengths. All of them were linked by economy-wide, common, expenditure impulses (and averaged expectations). In a series of papers, extending over half a century, Simon exploited this simple idea in many fertile ways: to study causality in economic models; to formalize causality and link it with identifiability in econometric models; to theorize about aggregation in economic models; and to formalize the idea of the hierarchy of complexity utilizing near decomposability in hierarchical organizations (Simon 1996: 173-216); to study counterfactuals in scientific theorizing, and on using the idea of near-decomposability to study evolutionary aspects of mind, thought, organizations and nature.

The idea of “near decomposability”, extracted from Goodwin’s notion of unilateral coupling in markets with production lags, was made mathematically rigorous by the idea of approximately decomposable (or indecomposable) matrices. It was in Goodwin’s own pioneering work on multi-sectoral dynamic models (in the late 1940s and early 1950s) that both the link between indecomposability and the Perron-Frobenius theorem(s) and, indeed, the use and introduction of the Perron-Frobenius theorem(s) first appeared. The Goodwin-Simon analytical nexus straddled both inter-industrial and macroeconomic dynamics, in that Simon’s work on applying servomechanism theory to the modelling of macroeconomic systems was also inspired by the former’s classics on aggregate nonlinear macrodynamics (see Goodwin 1951: Simon 1952).

In Models of My Life, he takes this particular example of the inspiration he got from Goodwin’s attempt to represent markets interacting with delayed responses as hierarchically coupled oscillators to wonder about the kinds of representations scientists use in thinking about research problems: where do the metaphors for scientific representations come from? How are they represented and retained in the human mind? How are they recalled - when, and why at that particular juncture? What are the triggering mechanisms and the catalysts? He had answers, tentative, testable and, as always, interesting and provocative.

In his Raffaele Mattioli Lectures of 1993 (Simon 1997), he argued for a study of economies in terms of organizations as the basic unit rather than the traditional device of markets. His case was based on solid empirical and theoretical results. He felt - justified by empirical and experimental data and results - the reliance on markets led to the unnecessary and false claims for their optimality properties (true only in one of many possible mathematical worlds and false in all others) as well as the propagation of the false dichotomy between the virtues of decentralization and the vices of centralization, without forgetting the merits of the former and the disadvantages of the latter.

Finally, it is rarely recognized that Simon, in his imaginative experimental approach to human problem-solving was one of the undisputed pioneers of so-called agent-based modelling. He always made clear, for example, in studying and automating chess, that understanding the difference between the rules determining the exact dynamics of an individual chessman, say the rook, and its particular position in the configuration of a game, was formally - that is, algorithmically - indeterminate (without going as far towards claiming algorithmic undecidability for this “aggregation” problem).

He was optimistic about the future of economics and even more so of computer science and the interaction between the two and psychology. In a letter written after reading Velupillai (2000), he wrote that “the battle has been won, at least the first part, although it will take a couple of academic generations to clear the field and get some sensible textbooks written and the next generations trained”.

The precept that may have guided his astonishingly fertile scientific life may well have been, in his own words (Simon, 1996: 28; original emphasis): “What a person cannot do he or she will not do, no matter how strong the urge to do it.”

At the time of his death he was the Richard King Mellon Professor of Computer Science and Psychology at Carnegie-Mellon University, a post he had held since 1966.

K. Vela Velupillai and Ying-Fang Kao

Herbert Alexander Simon (1916-2001)

More on the topic Herbert Alexander Simon (1916-2001):