ask a simple question - carnegie mellon...

32
Forthcoming, Policy Sciences Ask a Simple Question: A Retrospective on Herbert Alexander Simon 1 Patrick D. Larkey Carnegie Mellon University Herbert Simon, the Richard King Mellon University Professor of Computer Science and Psychology at Carnegie Mellon University and the most decorated behavioral and social scientist of the last millennium, died on February 9, 2001, aged 84. Simon’s lifelong obsession was to understand how humans make decisions and solve problems. The obsession was triggered by a policy question posed to him at the age of 19 about allocation behavior in a municipal government. This obsession stimulated contributions in several disciplines and the creation of new fields of inquiry. The relevance of Simon’s wide-ranging contributions in psychology, economics, computer science, political science, organization theory, philosophy, logic, and public administration to the policy sciences may not be immediately obvious to policy scientists. It should be. Very early one Saturday morning in the late 1970s during a “Christmas holiday,” Herbert Alexander Simon was part of a small group of faculty from Carnegie Mellon University waiting to confer with an influential local Congressman – Doug Walgren, then a member of the House of Representatives’ committee responsible for the National Science Foundation budget. Walgren had called the meeting for advice on how he might counter threats to cut allocations to scientific research in the federal budget; the behavioral and social science budgets were especially at risk. The Congressman was delayed by severe winter weather gripping Pittsburgh. The choice looked dismal: waiting for an indefinite period in the chairless, deserted, poorly lit, cold hallway in Pittsburgh’s Federal Building to see if the Congressman would show or venturing back out into a dark, bitterly cold downtown for a trek home on icy roads. 1 Susan Ambrose, Thomas Anton, William Ascher, Garry Brewer, Steven Brown, Jonathan Caulkins, W.W. Cooper, Robert Coulam, Otto Davis, Robyn Dawes, Erik Devereux, George Downs, Phoebe Ellsworth, Dick Hayes, Yuji Ijiri, Joseph Kadane, David Klahr, David Krackhardt, Ramaya Krishnan, Janet Larkey, Andrew Marshall, Marino Parascenzo, Mike Prietula, Steve Roehrig, Anita Sands, Zur Shapira, Peter Simon, Richard Smith, Shyam Sunder, Joel Tarr, Fred Thompson, and Kenneth Warner provided useful reactions to and comments on an early draft of this paper. They made it better. Any remaining failures are mine.

Upload: others

Post on 20-Sep-2020

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

Forthcoming, Policy Sciences

Ask a Simple Question:

A Retrospective on Herbert Alexander Simon1

Patrick D. Larkey

Carnegie Mellon University

Herbert Simon, the Richard King Mellon University Professor of Computer Science and Psychology at Carnegie Mellon University and the most decorated behavioral and social scientist of the last millennium, died on February 9, 2001, aged 84. Simon’s lifelong obsession was to understand how humans make decisions and solve problems. The obsession was triggered by a policy question posed to him at the age of 19 about allocation behavior in a municipal government. This obsession stimulated contributions in several disciplines and the creation of new fields of inquiry. The relevance of Simon’s wide-ranging contributions in psychology, economics, computer science, political science, organization theory, philosophy, logic, and public administration to the policy sciences may not be immediately obvious to policy scientists. It should be.

Very early one Saturday morning in the late 1970s during a “Christmas holiday,”

Herbert Alexander Simon was part of a small group of faculty from Carnegie Mellon University waiting to confer with an influential local Congressman – Doug Walgren, then a member of the House of Representatives’ committee responsible for the National Science Foundation budget. Walgren had called the meeting for advice on how he might counter threats to cut allocations to scientific research in the federal budget; the behavioral and social science budgets were especially at risk.

The Congressman was delayed by severe winter weather gripping Pittsburgh. The

choice looked dismal: waiting for an indefinite period in the chairless, deserted, poorly lit, cold hallway in Pittsburgh’s Federal Building to see if the Congressman would show or venturing back out into a dark, bitterly cold downtown for a trek home on icy roads.

1 Susan Ambrose, Thomas Anton, William Ascher, Garry Brewer, Steven Brown, Jonathan Caulkins, W.W. Cooper, Robert Coulam, Otto Davis, Robyn Dawes, Erik Devereux, George Downs, Phoebe Ellsworth, Dick Hayes, Yuji Ijiri, Joseph Kadane, David Klahr, David Krackhardt, Ramaya Krishnan, Janet Larkey, Andrew Marshall, Marino Parascenzo, Mike Prietula, Steve Roehrig, Anita Sands, Zur Shapira, Peter Simon, Richard Smith, Shyam Sunder, Joel Tarr, Fred Thompson, and Kenneth Warner provided useful reactions to and comments on an early draft of this paper. They made it better. Any remaining failures are mine.

Page 2: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

2

Not so dismal. Herb convened an impromptu seminar. A few faculty. A little time. What else?

Herb related his recent experience as a member of a National Research Council study

panel on meteorology, especially what he had learned about hurricane control. The panel had concluded from the findings of a few fairly rough experiments that the United States had the ability to reduce the intensity of hurricanes significantly – something on the order of a 20 to 30 percent reduction as measured by wind speeds – by seeding the storms heavily with silver nitrate or other substances that precipitated water from storm clouds. Herb summarized the recent loss experience from hurricanes in property damage, lives, and economic dislocation; the losses were very substantial. He also summarized the state of meteorological modeling of storms as fairly primitive with a lot of error in predicting the intensity and tracks of major storms.

Seeding may lessen the intensity of a storm but it is also apt to change the likely path

of the storm in unpredictable ways. The best possible outcome would be seeding a severe storm headed for a landfall along a track through densely populated, heavily capitalized areas and weakening that storm to a track such that it would never make landfall at all but move back out to sea; a storm that was likely to cause the loss of hundreds of lives, thousands of injuries, and billions of dollars in property damage and disruption of commerce becomes, at worst, a hazard for whatever shipping there is along its track. The worst possible outcome would be seeding a severe storm headed seaward or toward a landfall on a relatively undeveloped, unpopulated area of the coast and, arguably, redirecting it as a somewhat less severe but still deadly storm to a heavily developed and populated area of the coast; a storm on course to discomfit, if not harm, a scattered rural population, while perhaps recharging aquifers and relieving droughts along the way, would become a storm that causes major losses in life and property.

On balance, however, the United States could probably reduce aggregate losses from

storms through an aggressive and intelligent seeding program that emphasized likely best cases and avoided potential worst cases. Such a program would also provide an empirical basis for aggressively improving the meteorological models of storm intensity and track which would, in turn, permit an even more beneficial seeding program which would, in turn...

Herb then posed a puzzle: The relevant federal officials subsequently decided

against seeding hurricanes in spite of our ability to, on average, reduce damages for society as a whole. Why?

Our guesses were along predictable lines – environmental contamination from the

seeding substances, risk to the flight crews flying into the hurricanes to do the seeding, certain costs versus uncertain and probably inestimable benefits and other objections now lost to memory; there were surely several more because Herb never gave the answers to one of his puzzles easily or quickly if he had a good hypothesis,. All of our guesses were wrong.

Page 3: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

3

The decision not to seed, according to Herb, was driven by liability issues. Once the federal government intervened by seeding any storm, it might become a “federal storm” -- the federal government might become legally responsible for any and all storm damage even in instances where seeding reduced damages in an area from what damages would have been absent seeding. Courts are unpredictable in their treatment of counterfactuals and scientific uncertainties in such cases. The government’s lawyers strongly advised against the federal government going into the “storm business.”

With Herb’s revelation, the focus of this impromptu seminar shifted to even more

challenging questions: Was this a good decision? Are there ever any circumstances in which the federal government is justified in refusing an action with significantly positive expected collective benefits and uncertain redistributive effects? How might a government be designed and operated in concert with private insurance providers so as to appropriately manage risk? How could we better reconcile the different and often conflicting professional perspectives of scientists, analysts, politicians and lawyers? The Congressman arrived all too soon.

This was quintessential Herbert Simon. Herb was never bored or boring. By

himself, there were always things to observe and think about. With others, there were always things to discuss.

A dinner party with Herb might turn into a discussion of why the photons emitted

from a flashlight have focus and extent and don’t just scatter randomly or drip down from the front of the device; or of whether academic careers are affected by the spelling of one’s last name with “a’s” and “b’s” advantaged by alphabetic conventions in listing authors and the tendency of evaluation committees to give disproportionate credit to “first authors.”

The ostensible business in a half-hour meeting with Herb on some important

administrative or personnel matter might take two to five minutes with the rest of the time devoted to a challenging discussion on U.S. science policy or the incentive effects of patent law or the difficulties in measuring productivity or why two and only two ancient civilizations evolved base-60 number systems while most evolved base-20 or base-10. The ostensible business was settled quickly and effectively so that the meeting could move on to the truly important business of a university: thinking and talking about interesting things.

In the spring of 1978, the late Lee W. Gregg, then Head of the Department of

Psychology at Carnegie Mellon University, walked into the main hallway housing the department and observed Herb with a distinguished Professor of Economics from the University of Chicago pinned against a wall and saying loudly, “You don’t really believe that, do you?” The visitor was there for his child’s graduation from Carnegie Mellon University and had dropped by for a courtesy call on Herb that had, somewhat inevitably, evolved into a vigorous discussion of economic theory. Lee, who had missed the substance of the conversation, noted that the visitor looked much more interested in finding an avenue of escape than in winning the argument, whatever it was. Lee

Page 4: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

4

expressed surprise: “I didn’t know that Herb still cared so intensely about the nonsense that economists believe,” and then chided himself, “but I should have known.” Herb was 61 at the time.

This was quintessential Herbert Simon. The same man who could charm anyone

from any walk of life at a neighborhood party by listening attentively to their interests and by sharing amusing anecdotes about Wisconsin in the 1920s, Chicago in the 1930s, or California in the 1940s could turn with incredible intellectual ferocity on those ostensibly capable of defending themselves who were taking an intellectual position that needed a little more thought. Simon’s monologues that began, “Look, friend...,” were always cogent and signaled that the gloves were off, that it was time to get to the core issues, and that the recipient of the monologue was about to learn the flaw in his reasoning, whatever that happened to be. Herb always cared intensely about intellectual issues.

In the mid-1980s, an Assistant Professor at Carnegie Mellon University finished a

lecture on some aspect of microeconomics -- price theory or consumer theory. Simon, set to preside over the next scheduled class in that hall, had apparently been listening to the end of the lecture. As the young faculty member departed the lecture hall, he passed Simon who handed him a note with a friendly nod and no comment and strode to the lectern. Outside the lecture hall, the young faculty member opened the note. It said, “Young minds are too precious to waste on that drivel.”

The young faculty member never did seek Simon out to find out exactly what he

meant by the note. Was Simon referring to the faculty mind, the students’ minds, or both? Why did Simon consider the standard microeconomic fare “drivel”? Alas, another opportunity to learn from Herb missed.

This was quintessential Herbert Simon. He was always reading and listening.

Faculty at Carnegie Mellon never knew when they might receive a note from Herb. Often the note came in the form of a scribbled, signed comment on the upper margin of the front page of one of his reprints published a decade or three ago remarking on his having enjoyed something you had published or asking a penetrating question about a position you had taken. Usually these notes referenced an article that you had not sent him but that he had seen in his perpetual scanning of the social and behavioral sciences. Usually his enclosed article was related to your work, something that you wish you had written or, at least, referenced as relevant. Apparently Simon did this with faculty around the world in several disciplines.

Immediately after winning the Nobel Prize for Economics in the Fall of 1978, Herb’s

schedule got a little more complicated than usual. He had previously agreed to give a lecture at a large midwestern university during the next week. In the schedule crunch, Simon arranged for Dr. W.W. Cooper, a distinguished colleague and friend for some forty years, to take his place. Bill Cooper was met at the airport by his host, the Chairman of the Department of Economics, who immediately launched a diatribe on Simon winning the Nobel Prize. He complained bitterly about Simon winning the prize

Page 5: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

5

asserting that Simon was not a real economist and that there were much more deserving “real economists” – he could name at least six -- available to the Nobel Committee. The diatribe continued throughout the long drive from the airport to the university and into the chairman’s office.

Cooper shifted the conversation by asking the chairman, obviously a real, card-

carrying economist and proud of it, what his current research was. The chairman spoke enthusiastically about applying control theoretic models to some problem in macroeconomic theory. Cooper then asked the chairman what procedure he was using to solve the complex dynamic programming models. The chairman told him. Cooper then suggested that the chairman look up an article published in the 1950s, where he would discover that Simon was the first to prove the optimality of the solution procedure the chairman was using in his research (Simon, 1956). That was, of course, the end of the diatribe on Simon not being a “real economist.”

This was quintessential Herbert Simon. Simon did so much over such a long period

of time that even direct beneficiaries of his work such as this chairman of a major economics department, a productive research economist, failed to understand and appreciate the breadth and depth of Simon’s contributions. For most economists whose careers began after the mid-1960s, Simon was a psychologist and computer scientist who had made something of a name for himself in the days of yore criticizing assumptions routinely made in exercising and extending the neoclassical model.

For these young economists, many now of retirement age, Simon was a prominent

academic personage who had somehow failed to appreciate the enormous analytic power and elegance of the mathematical representations of neoclassical microeconomics. For reasons incomprehensible to them, Simon persisted in questioning the widely-accepted foundational assumptions of rational behavior for individuals and organizations, particularly the postulates that individuals maximize utility and corporations maximize profit. Simon kept talking about “bounded rationality” and “satisficing”, concepts that did not obviously lend themselves to the elegant mathematics made possible with the optimization postulates.

Perhaps some of the younger economists mistakenly thought that Simon was not

comfortable with their increasingly advanced mathematics. Why else would a seemingly otherwise brilliant, rational man forsake the power and elegance of the models of mainstream economics for the grubby world of experiments and computational models in the backwater discipline of psychology and the nascent discipline of computer science? Why would Simon persist in advocating computational models representing processes of decisionmaking when one could ignore the tedious, tangled, obscure processes of actual behavior, postulate an outcome of those processes, whatever they were, and get on with real mathematical science? For these economists, perhaps, Simon’s move was further evidence of the validity of natural selection, those relentless processes that winnow out all but the rational and fit making mainstream economic theory true: those who can’t cut it in the rigorous mathematical disciplines move into the less rigorous, softer disciplines. Sciences of the Artificial (Simon, 1969) was not part of economics curricula.

Page 6: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

6

This would be an unfortunate and somewhat ironic reputational fate for a brilliant,

driven man who spent the better part of seventy incredibly productive years – Simon didn’t get seriously started until his teen years although by his own account he was actively studying and thinking about things before then – attempting to increase the rigor of the social and behavioral sciences. Some of the very best economists such as Jascha Marschak, Kenneth Arrow, Paul Samuelson, Tjalling Koopmans, Milton Friedman, George Stigler, Oskar Morgenstern, and Franco Modigliani may have disagreed with many of Simon’s theoretical predilections but never had any doubts about his rigor.

Simon’s “rigor” was in the correct application of any language, device, or

methodological procedure that would, for the purpose of building and testing explicit models of processes, advance his understanding of how human beings decided and solved problems. This is different from the “rigor” in reasoning from an axiomatic foundation. This is different from the “rigor” in conventional experimentation to test an hypothesis. This is different from the “rigor” in conventional statistics and econometrics to derive and test static models against invariably weak null hypotheses before moving on to do the same for the next, often disjoint, hypothesis.2

Simon’s rigor is the rigor of understanding how something works to an extent that

you can build and test a model by putting it to a task in a specified domain. For Simon, the model and what the model could do were always much more important than the orderliness of the processes by which you derived the model.

Simon’s relentless focus on results in the form of working models of behavioral

processes invited unconventional approaches wherever conventional approaches did not contribute. If conventional mathematics, formal logic, and natural language are not adequate for expressing an explicit model of behavior on a symbol processing task, simulate. If conventional empirical methodologies for experimentation and data analysis are not adequate to learn how to build the model, invent new techniques for understanding, such as protocol analysis. If conventional notions about sampling are irrelevant when one or a few subjects are all that is required to understand behavior on a specific task, ignore the conventions.

Once there is a model of an agent capable in a specific domain, the extent to which a

broader sample of human beings behave like the model and the extent to which the same agent is capable in different domains are empirical questions. Understanding how at least one agent actually behaves in at least one specific domain is the necessary first step.

The economists’ rigor was too often the logically correct elaboration of a theoretical

paradigm that Simon believed to be elegant but descriptively wrong and prescriptively useless. Why assert that all agents optimize in consequence or expectation in all domains

2 For a compact and insightful treatment of some of the more fundamental issues such as inductive versus deductive reasoning, measurement, language, and testing in theory construction and the relationships among models, theories, descriptions, and explanations behind the casual discussion here, see Coombs (1983). The issues are neither easy nor settled.

Page 7: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

7

and attempt to defend that obviously false assertion when you can study how agents actually behave? Why prescribe optimization as a decision process unless you can deliver an algorithm that actually does the optimization?

For someone who spent a career defying the conventional wisdom on theory and

practice in several fields, most notably Economics and Psychology, Simon did pretty well by all conceivable measures of success for an academic career. His substantive contributions were prodigious. For those contributions, he won many awards including the A.M. Turing Award (1975 with Allen Newell) in computer science; the Alfred Nobel Memorial Prize in Economic Sciences (1978); the National Medal of Science (1986); the American Psychological Association Award for Outstanding Lifetime Contributions to Psychology (1993); the Award for Research Excellence from the International Joint Conferences on Artificial Intelligence (1995); the Dwight Waldo Award from the American Society of Public Administration (1995); the Harold Pender Award from the Moore School of Electrical Engineering at the University of Pennsylvania; and awards from the Association for Computing Machinery, the American Political Science Association, the Academy of Management, the Operations Research Society and the Institute of Management Science.

Simon belonged to all of the main honorific U.S. academic organizations, the

Chinese Academy of Sciences and at least two unusual organizations for someone with a doctorate in political science, the Automation Hall of Fame at the Chicago Museum of Science and Technology and the Institute of Electrical and Electronic Engineers (honorary member). He was also the first social scientist appointed to the President’s Science Advisory Committee.

At least 24 colleges and universities presented Simon with an honorary doctorate. Simon chose not to climb the prestige ladder in terms of university affiliation. For

the better part of 50 years Simon could have had an appointment on very attractive terms on virtually any faculty of any university in the world. His faculty affiliations were the Illinois Institute of Technology for a few years in the mid to late 1940s and the Carnegie Institute of Technology, which became Carnegie Mellon University, for the remaining 51 years of his life. While Simon’s salary was never one of the highest five for the 501c3 filings required of nonprofit organizations, he was, unusually, a Life Trustee of Carnegie Mellon University for decades and had a building on campus named for him (and one of his principal collaborators, Allen Newell) several years prior to his death.

Simon published a lot. In a era when academic lives are frequently summarized by

naming the decedent’s title, disciplinary affiliation, university affiliation, and a count of books and articles, Simon’s obituaries centered on the substance of his contributions and the fields he helped create. His disciplinary affiliation was “polymath” or, simply, “scholar.” Simon’s oeuvre included about 30 books (with dozens of editions and translations), numerous book-length research reports, and approximately one-thousand journal articles, book chapters, forewords, comments and book reviews. The ideas were always more important than the counts.

Page 8: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

8

If you really want to understand Herbert Simon, you must read Herbert Simon. Start

with his autobiography, Models of My Life (1991), and move on to Administrative Behavior (1947), Sciences of the Artificial (1969), Models of Man (1957), Organizations (with J.G. March, 1958), Human Problem Solving (with A. Newell, 1972) or any of the several volumes of Simon’s collected works as your interests lead you.

Simon began his career heavily immersed in the pragmatic problems of municipal,

state, and federal governments and, somewhat later, business organizations and universities. Simon’s direct contributions to policy and management include his early work on municipal government, the Economic Cooperation Administration (the Marshall Plan), state relief administration, peaceful uses of atomic energy, empirical work on business decisionmaking, and later reports co-authored for committees of the National Academy of Science or the National Research Council.

One striking feature of Simon’s early work, the work grounded in real, complex

government and business organizations, is the extent to which every project became an occasion to identify and expound upon fundamental issues. Work on performance measurement in local governments became an occasion to discover principles of organizing and managing. Work on how business organizes accounting functions became an occasion to discover principles of organizational structure and the role of information in decisionmaking and organizations.

Also striking is the extent to which Simon was never satisfied with the theoretical

and empirical foundations for whatever problem he was working on at the moment. These dissatisfactions with the foundations in the behavioral and social sciences drove successive rounds of research to improve the foundations. Simon ended his career heavily immersed in how subjects solved much simpler problems in laboratory settings. While he maintained an active interest in both policy issues and foundational issues, his effort on policy issues waned while his effort on foundational issues waxed over his lifetime.

This trek from messy municipal management problems to chess, logic, and

cryptoarithmetic problems was not blind reductionism. It was, as he describes it (Simon, 1991; p. 113), a “biased, random walk” through a “Darwinian maze.” His instincts for researchable, important problems in traversing his maze were obviously very good. The eighteenth century English author, Horace Walpole, once noted that “(t)he whole secret of life is to be interested in one thing profoundly and a thousand things well.” Herbert Simon had this secret early in life.

Every traversable maze has a starting point. In his autobiography, Models of My

Life (1991, p. 370), Simon describes how he acquired a research problem that sustained him intellectually for life. The empirical question posed to him was: How are funds divided between playground maintenance -- planting trees, cutting grass, etc. -- and playground activity leadership -- planning and running programs -- in Milwaukee?

Page 9: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

9

The year was 1935. Simon was 19 years old, an undergraduate at the University of Chicago studying economics because the mathematical rigor of the field appealed to him but majoring in political science because he did not want to take an accounting course required of economics majors.

Pondering this somewhat mundane question about allocation behavior, Simon

concluded that the ready hypothesis from his then-limited study of economics, “Divide the funds so that the next dollar spent for maintenance will produce the same return as the next dollar spent for leaders’ salaries”, was obviously deficient in two respects. First, no one seemed to be thinking about the decision in this way; the hypothesis failed descriptively. Second, he could not see how to weigh the value of one against the value of the other; the hypothesis failed prescriptively.

For Simon, the specific problem about behavior on allocations to playgrounds in

Milwaukee became a much more general problem: “How do human beings reason when the conditions for rationality postulated by neoclassical economics are not met?” In much of his subsequent research, Simon labeled the behavior of interest as decisionmaking, choice, or problem-solving interchangeably as synonyms for “reasoning.” From the discovery of this problem in 1935 until his death in 2001, most of what Simon published can be understood as contributing more or less directly to answering this question.

Obviously, Simon took the seemingly simple question about allocations to

playgrounds a little more seriously than thousands of others who have pondered and continue to ponder variants of the same question.

“Practical man3,” the manager or analyst, is not very curious about what it would

take to answer this question rigorously. For practical man the only serious questions are those where the questioner has standing and must be answered or where the answer may have other proximate values. If a critic alleges misfeasance, malfeasance, or nonfeasance in the allocations, the competent practical man who is a principal in playground allocations will be primed to defend the decision. The ability to justify decisions in process regardless of consequences is important in much decisionmaking; it is critical for individuals in organizational and other political contexts. Verifiable demonstrations of having maximized expected utility are not the order of the day; alas, the ability to point to someone or something else as responsible for unfortunate consequences, e.g., bad weather or vandals made the allocations to playground maintenance inadequate, may be much more useful in actual decisionmaking.

On Simon’s question about allocations to playgrounds, practical man satisfices,

producing an answer good enough for the circumstances. Next question please. The foundational issues in Simon’s question remain unseen and untouched.

3 “Man” as used here is gender-neutral.

Page 10: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

10

The more natural, consequential question for practical man is, “Can the allocations be improved?” He may be able to answer this question and contribute improved allocations without ever confronting the difficulties in understanding the complex behavioral processes that generated the original, inferior allocation.

If practical man can look at the allocations and make a persuasive case that shifting

playground money from one sub-account to the other would produce superior service outcomes, improvement may be at hand. Or maybe not if advocates of greater public safety take the occasion to argue that citizens would be better served by moving an equivalent amount of money to reduce burglary rates rather than simply moving it around in the intrinsically less important playgrounds area. A real donnybrook might ensue if the advocates of several other service areas then take the occasion to argue the merits of increased expenditures in their areas. There are always potholes to fill, children to educate, and restaurants to inspect with accompanying good arguments for doing more of each and for doing each better.

“Academic man” with a pat but untested, and perhaps untestable, theory that answers

Simon’s descriptive question a priori is not very curious about how well the theory fits the facts; he shifts his attention in an unending search for confirming instances to more hospitable phenomena. Foundational issues are not on the table; they were settled a priori. Once you know that everyone is maximizing their subjective expected utility because competitive pressures require them to do so, what else is there to know?

Academic men without such pat theories are not very curious about Simon’s problem

for several reasons. A research focus on allocations to playgrounds in one local government is in many respects a recipe for an unsuccessful academic career. All of the proximate incentives in universities and academic disciplines discourage academics from becoming too serious about such topics.

Substantively, the problem falls between the disciplinary cracks. It involves money

and competition for scarce resources, the rightful province of economics; it involves elected officials and their politics, the rightful province of political science; it involves individual and group behaviors, the rightful province of psychology; it involves bureaucratic behavior, the rightful province of sociology, anthropology or public administration depending on the level of abstraction; it involves a slew of issues in the design and delivery of services, the rightful province of management scientists, planners and domain specialists, e.g., educators and traffic engineers; it involves interpreting complex financial records, the rightful province of accountants.

The multidisciplinary character of Simon’s problem is a significant deterrent to

academics aspiring to disciplinary careers. The deterrent is especially effective for the most capable and ambitious young economists who might be tempted to stray from the paradigm, to become “atheoretical” and/or “descriptive.”

Simon’s problem presents no great opportunity for academic “rain-makers.” There

is not much external funding to study the problem because understanding precisely how

Page 11: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

11

local governments allocate resources lacks both intellectual panache and obvious “relevance” for those controlling the purse-strings of the research funding institutions, public and private.

Simon’s problem offers no great opportunity for high-flying methodologists and

theorists. The closer you look, the messier Simon’s problem gets with a very complex agent deciding a potentially complex issue in a complex context. Attacking the problem effectively requires learning a lot about both the substantive domain and the motivations and capabilities of the people involved. The problem requires the researcher to directly observe behavior in the field and to understand what the numbers actually mean. Field work of the required sort is time consuming, messy conceptually, and unglamorous. Besides, nothing complicates abstract reasoning more than knowing too much about a domain and trying to incorporate what you know into your theories and empirical analyses.

Explanations, however carefully crafted and tested, of how one local government

decided one allocation to a particular functional area are not good grist for the publication mill, at least the prestigious publication mills within particular disciplines.

“Policy man” is not very curious about anything concerning playgrounds in

Milwaukee because the topic is such small potatoes in the larger scheme of things. Why waste time on such a trivial policy matter when you can grapple with one or more of the grand issues?

Some grand issues such as eradicating world hunger, promoting world peace,

substance abuse, eugenics, voter rights, free trade vs. protectionism, or improving the efficiency of government are eternal. Other grand issues are more epochal. In 1935, amid the Great Depression, the epochal issues surely included business cycles, social security, farm policy, unemployment, hunger, fascism, communism, militarism, and imperialism. In 2002 the list of epochal grand issues includes terrorism, global warming, racism, air and water pollution, crime and violence, healthcare, and alternative energy sources.

There is something automatically honorable, even heroic, about confronting grand

issues even if the issues as framed popularly are not researchable and confronting them consists only of talking about them. Knowing the lingo and being “committed” to an issue that is widely perceived as important confers legitimacy; it reassures some audiences, especially analytically-challenged audiences, that policy man is a good person working for the betterment of humankind with good instincts about what problems are important.

Policy man can be morally indignant about the villains purportedly causing the

problem and deeply compassionate about the plights of the purported victims. Policy man can be overtly ideological; the conservatives’ villains are the liberals’ victims and vice versa. Success for policy man is more often defined in terms of op ed pieces, stints on news and talk shows, charismatic lectures on some rubber chicken circuit, and/or

Page 12: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

12

popular trade books rather than in terms of his substantive contributions to resolving issues.

Policy men are also not very curious about the scientific foundations that would be

required to address grand issues effectively. Serious science about complex issues tends to challenge cherished beliefs, cloud ideological positions and impede simplistic communications, the pith of most public dialogue over grand issues. Many grand issues become morality plays with no speaking parts for serious scientists. Scientists are merely one among many interest groups.

“Curious man,” and Herbert Simon was the archetypical curious man, pushes matters

further. Curiosity leads to questions about questions about questions... Curiosity is the implacable enemy of extant beliefs. Curiosity, in close cahoots with elbow-grease and serendipity, drives intellectual progress: from Aristotelian physics to the search for the elusive Higgs Boson; from phenotypical taxonomies of flora and fauna to molecular biology; from phlogiston to analytic chemistry; from counting to calculus and supercomputing; from alchemy to modern materials science...

While essential to intellectual progress, curiosity is, curiously, dysfunctional for

many social roles. Most social organizations from churches to clubs to formal business and government organizations to academic disciplines to political parties rely to a greater or lesser extent on shared beliefs about what is valuable in consequence and process. Curiosity is seditious for shared beliefs. Curiosity is a main target of socialization processes from early childhood; we learn early and often that not all questions are appreciated and conducive to sustained, positive social relationships.

The curious theologian may find a different faith or no faith. The curious manager

may be unable to motivate employees with simplistic articulations of what the group is striving to accomplish. The curious, gifted athlete may run afoul of the authoritarian, less-than-gifted coach. The curious theoretician may find conversations difficult with other theoreticians who are heavily invested in the correctness of a particular theory. The curious empiricist may find conversations difficult with other empiricists who are heavily invested in the correctness of particular methodologies. The curious politician may garner little support from colleagues and constituents athirst for affirmation of their own beliefs.

The curious employee in any collective calling, including academics, may have their

career hampered by their reputation as “argumentative,” “disputatious,” “arrogant,” “confrontational,” “difficult,” or, the curse of all reputational curses for corporate types, “not a team player.” These epithets, whispered in the confidential conversations of reputation mongers with more gravitas than evidence, accumulate as “reputational baggage” that hinders the average scholar, even the better than average scholar.

Page 13: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

13

Simon was somehow immune to most of the potentially perverse reputational effects of his life-long intense curiosity.4 Perhaps it was because he was so substantively good so early and for so long in so many different fields. Perhaps it was because he did not merely visit the several disciplines and criticize their practices but stayed long enough to make enduring, constructive contributions. Perhaps it was because of his significant institutional contributions locally, nationally, and internationally. Perhaps it was because of his knack for keeping ideas and personalities apart thus enabling him to sustain respectful, even warm, personal relationships with the best of his intellectual sparring partners.

Perhaps it was because of Simon’s strong belief that “You can’t beat something with

nothing.” Understanding how real human beings can and do behave in real task environments was much more important for Simon than railing against the numerous incorrect and inappropriate uses of rational models to describe or prescribe decisionmaking behavior. Bounded rationality was less an attack on boundless rationality than an attempt to replace something so obviously flawed with something better.

It has been 67 years since Simon found his lodestone question in Milwaukee. Given

all that Simon accomplished singly and collaboratively and all that has transpired in the several potentially relevant disciplines over six and a half decades of effort by thousands of scholars, analysts, and practitioners, how prepared are we to answer Simon’s original question, the one concerning the allocation decision to different playground accounts? Why X$ to programmed activities and Y$ to maintenance?

Simon responded to the question by writing his first scholarly paper on the

“Administration of Public Recreational Facilities in Milwaukee” (Simon, 1935). While never published, the gist of his argument with respect to the allocation question from this paper found its way into Administrative Behavior a decade later (Simon, 1947, pp. 211-212).

For the question at the ripe old age of 19, Simon answered with a verbal model based

on organizational identifications. The Extension Department of the School Board had apparently been given responsibility, without adequate new resources, for maintaining many new facilities built by the Playground Division of the Department of Public Works. The Playground Division believed that landscaping (maintenance) was important to the success of playgrounds. The Extension Department, after years of operating programs on poor facilities, believed that the quality of the programs was key to the quality of the service, not facilities, and, therefore, was reluctant to shift money from programs to their new maintenance responsibility.

By Simon’s own later standards of rigor, the explanation was ad hoc and

unsatisfactory. He offered no rigorous test of competing explanations or detailed empirical support for the single explanation. His model asserted the importance of

4 But not all of the effects. Chapter 17 in Simon’s autobiography (Simon, 1991) is entitled “On Being Argumentative.”

Page 14: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

14

different beliefs and values associated with functional perspectives within the organization and of the locus of decision control. The Extension Department controlled the allocation decision and spent relatively less on maintenance and relatively more on programs than would have been the case if the Department of Public Works had controlled the allocations. Simon’s 1935 model does not specify exactly how the conflicting beliefs and values are resolved in the allocations. Where are the higher-level managers, elected officials, voters, and playground users in Simon’s account? What was the psychological basis of “identifications?” What are the procedural mechanisms through which a conclusion is reached?

While we can, in 2002, draw on the rich organizational decisionmaking literature

developed in the intervening years to tell a somewhat more elaborate story about what probably happened with the decision in Milwaukee, we are still not in a very good position to posit and rigorously test alternative models of allocation decisions, certainly not along the lines of Simon’s own later work on decisionmaking and problem solving. The foundations that would enable explicit models of the agent, the government, deciding the allocation in a specified task environment have still not been laid.

The gulf between our rigorous understanding of how human subjects solve the

Missionaries and Cannibals problem in the laboratory and how human beings embedded in complex organizations solve allocation problems is vast. This gulf should not, however, be taken as evidence that Simon’s work on decisionmaking and problem-solving in the laboratory is inadequate or irrelevant to the allocation problem any more than the gulf between Johannes Kepler’s work on orbits and modern cosmogony or space travel is evidence of the inadequacy or irrelevance of Kepler’s work.

Also, the vastness of the gulf between our understanding of behavior in the

laboratory and behavior in Milwaukee should not discourage. It is not clear that understanding behavior is intrinsically harder than the ab initio problems in more mature disciplines. On the next clear night, look skyward and think about what it would take if left to your own devices to formulate a persuasive theory of the origin of the universe or to formulate a feasible plan for space travel. If you’re not that ambitious, propose a feasible test of Kepler’s assertion that “The ratio of the cube of a planet's distance from the sun and the square of the planet's orbital period is a constant and is the same for all planets” (Kepler, 1619).

If you are more biologically than astronomically inclined, walk into a wooded area

and think about how to use the information you can gather from your immediate surroundings to think productively about how to use organic molecules to build a computer or to control human disease. Less ambitious? Catalogue the plant and animal life you observe and defend your taxonomy as useful.

There have been few analyses since 1935 that attempt to observe and understand

how general purpose local governments, some 35,000 of them in the United States, allocate resources, a mere $952,330,313,000 in 1998-99 revenue collections according to the Census Bureau’s data (www.census.gov/govs/estimate/9900us). The amount that

Page 15: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

15

local governments control is actually much larger because they spend significant amounts of state government revenue collections, $1,152,869,754,000 in 1998-99 according to the Census Bureau, on behalf of their states.

In the last 20 years there has not been a single published study that rigorously posits

and tests alternative models of how local governments make resource allocation decisions, computational models at any level of abstraction in any non-verbal language capable of making the decisions. Evidence to the contrary is welcome.

Most academic attention on resource allocation decisionmaking has centered on the

U.S. Federal Government, where there’s some “real money” involved and issues are grander. This focus of attention has produced much more in the way of verbal descriptions and piecemeal empirical studies than working models of all or part of the federal budget process.

The economists’ answer to Simon’s descriptive question has not substantially

changed in the 67 years since Simon noticed its inadequacies. Boundless rationality remains as highly implausible an account of what real people in real organizations solving real problems do in 2002 as it was in 1935. The prescriptive advice derived from models of boundless rationality remains as impractical in 2002 as it was in 1935.

For every analysis that attempts to observe and understand how local governments

make allocation decisions, there are many analyses that assume a priori that the governments are unitary actors passively perceiving and translating citizen (or voter) preferences into allocations and services that satisfy those preferences. None of these “theories” specifies any explicit mechanisms used by governments to allocate resources or to provide services or any explicit mechanisms used by citizens and voters to evaluate public services or to articulate their preferences.

The best defense of these theoretical representations of how governments allocate

resources and provide services is in the Tiebout Hypothesis (Tiebout, 1956) which, in theory, puts performance pressure on governments analogous to the competitive pressure on firms in markets. If citizens are dissatisfied with any particular government’s policies on taxation and service provision they will move to another jurisdiction with policies more to their liking; if the bundle of taxes, playgrounds, education, police, fire, etc. in Milwaukee does not measure up, citizens will move to Dubuque, Dallas, Fargo, New York, or Phoenix where the bundle is a better match for their preferences. Local governments, in effect, compete for the affections of current and potential residents.

The answer to Simon’s question from this literature is that whatever the split is in the

allocations between playground maintenance and programs, it represents an adaptive response of the government to the on-going pressure from citizens who articulate their preferences for the split by moving to and from Milwaukee. If this explanation were even approximately true, this quasi-market will eventually produce an optimum in the variety of bundles offered by local governments and the distribution of citizens across those jurisdictions. There is no need for a scientific foundation in understanding how

Page 16: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

16

governments and citizens actually make decisions. It doesn’t matter. The market will take care of it. Simon’s question is irrelevant.

Hundreds of articles, many of them very sophisticated and highly technical in their

theoretical specifications and empirical tests, have been written by economists and political scientists over the years about the Tiebout Hypothesis. The entire literature, a self-perpetuating gedanken experiment, contributes next to nothing to our understanding of how governments actually allocate resources and provide services or how citizens actually make location decisions.

The other line of argument in justificatory support of using such normative rational

models as descriptive models is Friedman’s “as if” defense (Friedman, 1953): Decisionmakers most often cannot possibly do the detailed computations required of them in the model but behave “as if” they do perchance because of selective pressures in competitive markets. Friedman argues that direct tests of the axioms underpinning the theory cannot invalidate it; the only thing that matters are the predictions from the theory. In the ensuing 50 years there have been few ex ante predictions from the theory to test. The “predictions” claimed are, for the most part, ex post explanations. These explanations are subjected to weak confirmatory statistical tests. The theories are such that it is hard to see how they could even be used to make ex ante predictions.

A slightly weaker line of justification is that the theory provides a “good first

approximation” of behavior. The theory does have something of a foundation in common sense. Individuals will, given some choice on the matter, ceteris paribus, prefer more of an obviously good thing to less of an obviously good thing. Business firms will, given some choice on the matter, ceteris paribus, prefer more profit to less profit.

General purpose government organizations pose significant problems for this line of

justification. It is hard to identify any coherent outcome metric equivalent to profit or the individual’s “good thing” on behalf of a local government as chooser. It is also hard to envision a local government as a unitary actor -- disputes are legion on everything from curb cuts to the form of government. With business firms there is the somewhat sustainable illusion that there is a final authority in a chief executive officer and congruent goals among members of the organization. Even casual observation gives lie to such illusions for local governments.

The common sense foundations for rational models provide no useful guidance for

crafting an answer to Simon’s question. There is no obvious consensual agreement on what the split between playground maintenance and programs should be even among the experts and users. Perhaps we should provide great facilities without any supervised activity in order to foster greater creativity and self reliance in children? Perhaps we should intensify supervised play to inculcate the values of working with others cooperatively within a structured system? Or perhaps we ought to scrap all playground activity in favor of a few more police with the likely result of preventing a rape or a murder?

Page 17: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

17

It is rarely clear in allocation decisionmaking what the ultimate consequences are of one action versus another action. The deciders know in rough terms what an allocation buys in terms of inputs, e.g., the number of recreation directors or sworn police officers. The deciders know in much rougher terms the service activity potential in the inputs purchased, e.g., the amount and type of supervised play on the playgrounds and the number of police patrols per shift. The deciders have little or no clue what the valued consequences of the activities will be. What will more supervised play contribute to the social adjustment of children, their lifelong “wellness” through exercise, and rates of juvenile delinquency? What impact will increased police patrols have on crime rates for each category of crime?

Godot will arrive before the “second approximation” arrives from relentless use of

the “first approximation.” Simon noticed in 1935 that the “first approximation” was not a veridical, useful

account of how local government officials were deciding allocations. They must be doing something else. It is interesting and important to discover what that “something else” is.

While Simon took the obvious discordance between theory and practice in local

resource allocation as a call for better theories of actual behavior, most others noting the discordance over the years have taken it as conclusive evidence that something is wrong with existing practices and as a call to change those practices to more closely resemble the theory; their mantra is “if the obviously correct theory does not fit the facts, change the facts to fit the theory.” They take the theory as a useful prescriptive guide for reforming allocation decisionmaking.

This perspective has spawned many prescriptive schemes from the Good

Government initiatives out of the populist movement of a hundred years ago to the elaborate Planning, Programming, Budgeting Systems (PPBS) of the 1960s and 1970s to the current “accountability budgeting,” “performance budgeting” and “outcome budgeting” of the new millennium.

These rationalizing schemes have always been and will always be computationally

infeasible, indefensible in terms of their effects on outcomes, and unsustainable administratively and politically. They are an administrative equivalent to medieval alchemy in which the acolytes attempt to transform the base metal, extant allocation decision processes, into the target metal, a rational decision process, without understanding either base or target to know what transformations are required and what is feasible and what is not.

Optimization is attractive advice. After all, you can’t do better than that. The

difficulty is always, “How?” With all of the effort that has gone into developing and applying optimization techniques over the years, the lack of progress with respect to real problems is rather astounding. Very few real problems in business or government have succumbed to optimization.

Page 18: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

18

There have been a few notable successes. The petroleum industry’s decisions on

product mixes were never the same after the publication of “Blending Aviation Gasolines: A Study in Programming Interdependent Activities in an Integrated Oil Company” (Charnes, Cooper, and Mellon, 1952). Given a demonstrably better decision procedure providing an optimum solution to a financially important decision problem in a competitive market, the better decision procedure, mathematical programming, was widely adopted for an entire class of tractable problems.

Even though the origins of the analytic techniques that Charnes et al successfully

applied were in George Dantzig’s efforts to devise a better method of allocating resources in the United States during the Second World War, there are few if any comparably clear and important success stories for prescriptive decision techniques in government. The dream is the same abstractly. Much effort has been put into realizing the dream. The results have been very different pragmatically.

The reasons for the meager impact of optimization techniques on governmental

affairs and operations are complex. A main reason is that the complexity of the decision problems in real government organizations makes optimization impossible; the irreducible characteristics of the problems grossly violate the assumptions required by the various optimization techniques.

Another reason, not always acknowledged, is that governmental problems amenable

to optimization sometimes have great difficulty attracting the political attention and funding required to optimize. For example, operations researchers and engineers have the ability to optimize traffic flows through every intersection in the United States. We know how to do the analysis and design. We have all of the sensing and signaling equipment required. Yet, most Americans sit through long, unnecessary delays at one or several inefficient intersections in their daily commute. Millions of hours are lost annually to this source of inefficiency. Vehicles idling unnecessarily at intersections contribute substantially to air pollution problems. The costs in terms of wear and tear on vehicles needlessly running are real but difficult to estimate precisely. The psychic costs are real but unmeasurable. The activity of optimizing intersections would generate jobs for analysts, traffic engineers, electronics technicians and installers, and equipment manufacturers.

The public costs in optimizing intersections are non-trivial and perhaps the local

money could be better spent on playgrounds. But the federal and state governments could with some initiative greatly reduce these costs by using their market power through cooperative purchasing in bulk to force down the prices of analysis and equipment.

Why aren’t governments optimizing intersections? The question is at least as rich as

Simon’s question about playgrounds. The answer might have value both in understanding human behavior and in immediate application of that knowledge.

Page 19: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

19

There are comparable questions for private sector organizations who, as Simon reminded us again and again for more than 50 years, are not maximizing profits in any meaningful sense. Firms most often don’t know how to define profit much less maximize it. Firms survive and even thrive in competitive markets with substantial inefficiencies—employees overpaid relative to their contributions and solvable production problems unsolved; only relative efficiency matters and even that may not matter much in benign economic conditions and lax regulatory environments. What exactly was Enron maximizing? The fact that so many economists find the assumption of profit maximization convenient mathematically does not constitute empirical support for the existence of profit maximization as typical firm behavior.

Is the lack of scientific foundations for understanding allocation behavior by local

governments an anomaly? No. The scientific foundations for understanding the behavior of all actors central to

policy processes -- individuals, groups and organizations -- are somewhere between underdeveloped and nonexistent. We do not understand the behavior of criminals, public assistance recipients, polluters, taxpayers, terrorists, politicians, investors, savers, business firms, and bureaucracies in any satisfactory scientific sense. If this conclusion strikes you as unduly harsh, name one policy domain in which we have an explicit model of the germane behavioral processes and a significant track record of accurately predicting, ex ante, responses to policy changes.

Absent valid scientific foundations, the inevitable fallback is some variant of the

rational model, usually omniscient optimizing with postulated success in outcomes. The postulated computational mechanics through which actors achieve their maxima and minima are those of decision theory wherein decisionmakers choose actions on the basis of expected values. Decisionmakers optimize by maximizing or minimizing expected value with estimates of probability, outcomes, and values conditioned on actions taken.

The techniques for computing expected values have improved dramatically since the

mid-17th century. These improvements matter a lot for prescriptive applications of statistical decision theory by extending the domain of potential applications through ever more sophisticated computational procedures. But for descriptive applications of decision theory, applications in which individuals or organizations are assumed to behave in accordance with the theoretical prescriptions, the effect of the improvements is to make the theory ever less like any actual behavior in terms of process and outcome.

There are several thousand papers over at least 50 years showing that subjects in

more or less carefully controlled experiments have difficulty behaving in accordance with even the simplest versions of the rational model on relatively simple tasks. There are many other papers based upon field observations of individuals and organizations behaving that find the observed behavior discordant with rational models. The scientific foundations for this ubiquitous fallback model of human behavior are, by the standards of any natural science, simply not there. The theory does not fit the observable facts.

Page 20: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

20

Aristotle wrote in Metaphysica (1-5) that “(t)he so-called Pythagoreans, who were the first to take up mathematics, not only advanced this subject, but saturated with it, they fancied that the principles of mathematics were the principles of all things.” Simon recounts the moment in 1953 in a conversation with Tjalling Koopmans when he first clearly understood that his conception of the role of mathematics in developing theories of actual behavior was distinctly different than the economists’ conception of the role of mathematics in theory development (Simon, 1991, 106-7). The economists were Pythagorean and considered logical proof support for a theory even if the axioms upon which the proof depended were empirically false.

There is something of a behavioralist revolution happening in economics with a

fairly large number of very capable researchers determined to bridge the gap between the real world and the theoretical world. This reconciliation may be neither possible nor desirable. Continued efforts to understand behavior as departures from some variant of the rational model may actually impede progress on understanding how individuals, groups and organizations actually behave.

Simon’s original formulation was “How do human beings reason when the

conditions for rationality postulated by neoclassical economics are not met?” The formulation of the behavioral economists, much inspired by the seminal work of Ward Edwards, Amos Tversky and Daniel Kahneman, is something like “Under what conditions do human beings depart from rationality as postulated by neoclassical economics and are there regularities in how they depart? The mainstream economists’ formulation is “Human beings are rational and optimize regardless of the conditions for reasoning.”

These formulations are very different. Simon’s formulation invites a model of

human beings reasoning -- the agent -- under specific conditions for reasoning -- the task environment or problem space. The behavioral economics’ formulation invites a search for modifications in the rational model that will describe and predict behavior under specified conditions. The mainstream economics formulation invites endless elaboration of the rational model and imperialism in applications.

Notice that only one of the formulations lacks a question mark. The first

requirement for a fecund scientific research program is some acknowledgement, at least nodding acknowledgement, that there may be things that you don’t know about how the world works that warrant additional research.

A promising and relatively new line of work on individual decisionmaking is by a

psychologist, Gerd Gigerenzer, and his colleagues from many disciplines at the Center for Adaptive Behavior and Cognition, Max Planck Institute for Human Development. This is a research program that builds directly on Simon’s notions of bounded rationality (Gigerenzer and Selton, 2001; Gigerenzer, 2000). The research has already provided a number of provocative examples of decision circumstances in which simple heuristic decision procedures derived from studies of actual behavior outperform procedures derived from normative theories by some descriptive and predictive criteria. These

Page 21: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

21

examples are not just laboratory experiments but include fieldwork on such important topics as medical decisionmaking. To the extent that this line of research maintains a focus on understanding actual behavior and withstands the siren call of heuristic decision procedures as normative decision procedures, it can become an important extension of Simon’s work.

The problems that Simon found in 1935 with the microeconomic explanation for

allocation behavior stimulated his work on “bounded rationality” and organizational decisionmaking which, in turn, stimulated his search for a more stable set of theoretical primitives for a theory of human behavior in the development of cognitive science. It was all about understanding constrained, intelligent, goal-seeking behavior.

Simon essentially abandoned his early attempts to modify theories of rational

behavior as descriptions of actual behavior in favor of attempting to devise an alternative theory of how people actually behave. Some of the questions that kept him engaged for half a century included: How do individuals derive simplified, workable representations of problems in semantically rich domains? What makes some problems harder than others? What is the difference between ill-structured and well-structured problems? How do people set goals in specific domains? How do individuals acquire, store and process information in the course of solving problems?

Simon and his many collaborators learned much about what a rigorous, empirically-

grounded theory of human behavior will look like. Successively better approximations were offered of humans as information-processing systems busily deciding when to decide, acquiring information along afferent channels about their environment, building simplified representations of their environments, setting goals, generating pragmatic alternatives for accomplishing those goals, and deciding when to stop and move on to the next occasion for a decision. However, the gulfs between behavior in the laboratory and behavior in the field and between individual behavior and organizational behavior remain vast.

Simon’s original question about playground allocations remains unanswered and, for

the moment, unanswerable. It remains worth answering. Given that Simon concluded by the early-1950s that attempting to understand actual

behavior through the lens of rational models was an inefficient path to understanding actual behavior, it would be procedurally rational for those who would close the gulf between theoretical behavior and actual behavior to understand clearly why Simon took the path that he did. For perspective, sheer persistence and adroitness, Simon had few, if any, peers.

Why are there no useful scientific foundations for understanding allocation behavior

by local governments?

Page 22: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

22

There are at least two primary reasons. First, Simon’s question was and remains difficult conceptually and methodologically. Second, there are disincentives for those capable of progressing on the conceptual and methodological issues to do so.

Simon’s question was descriptive: How did Milwaukee decide on the relative

allocations to maintenance and program leadership for playgrounds in 1935? How was it decided that maintenance would get X$ and program leadership would get Y$? The question asks for an explanation of how a particular decision was actually made by a specific agent in a specific context.

The question is somewhat ambiguous. If by “allocations” the questioner meant

budgeted amounts, then it is a question about how a local government planned expenditure ceilings over a several month period in 1934 prior to the start of the 1935 fiscal year. If by “allocations” the questioner meant actual expenditures, it is a question about the planning for 1935, the execution of the plans during 1935, and the machinations of the financial managers and auditors at closing and after closing in 1936 to “make the numbers come out right.”

These are different questions. The second question is somewhat harder than the first

because it involves modeling thousands of decisions at the level of what the government did when an unplanned break in a water main made half of the surface of one playground suddenly unusable. Was the surface restored? How quickly was it restored? What was done with the permanent employees running programs at that playground? What were the savings from not running programs for the period that the playground was unusable and where did those savings go? Such decisions are the warp and woof of life in implementing those seemingly authoritative expenditure ceilings that emerge from annual resource allocation decisions.

The first question is hard enough. The complexity of the decision agent and the

potential complexity of the task environment pose significant challenges to the researcher who would answer the question.

The “agent” is a bureaucracy consisting of several levels of formal organizational

components and a number of important legal rules ranging from the requirement that the annual budget be balanced, i.e., authorized expenditures set equal to estimated revenues, to civil service or union contract protections for employees. The scientific foundations for modeling explicitly a complicated bureaucracy are not well developed; simply describing a decision in which dozens of people with different roles, different information, different causal beliefs, and different values participate in a process spanning months is a fairly daunting prospect.

Resource allocation decision processes are highly interdependent both substantively

and temporally; they are not, in the terminology of Ando, Fisher, and Simon (1963), decomposable in their inner environment or from their outer environment.

Page 23: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

23

In the inner environment substantively, the allocations to playgrounds cannot be studied in isolation from allocations to other purposes and from revenue considerations. Those responsible for playgrounds report to those responsible for educational and park facilities and programs who report to those responsible for educational and park and recreational functions who report to those responsible for fiscal and service functions who report to an appointed or elected chief executive who reports to or works with a legislative body chosen by citizens who use or don’t use the facilities and programs available on playgrounds.

There is strong temporal interdependence. While there is an annual event called the

budget decision, this is mainly an arbitrary structure imposed on continuous processes; many of the same people are doing many of the same things with respect to many of the same problems year after year. History matters. Past behavior constrains current behavior.

There is a public physical infrastructure, above ground and underground, that

changes gradually, mostly for the worse, and which must be maintained. The resource allocation problem with respect to this infrastructure at any point in time reflects the way in which the resource allocation problems in prior years were solved. If the local government deferred maintenance on physical infrastructure in several prior years, this may be the year in which resources must be put to maintenance and reconstruction if the infrastructure is to continue to deliver sewer, water or transportation services.

For Simon’s question on playground allocations, the cost of maintaining playgrounds

at any particular level is a function of many prior decisions on construction and maintenance quality; if previous decisionmakers built shoddy playgrounds with poor drainage, improper turf and asphalt, and bargain-basement equipment, current decisionmakers will have to spend relatively more on maintenance to sustain quality levels.

If decisionmaking on prior year budgets used indebtedness to cover operating

deficits, this may be the year in which taxes must rise or services cut to balance the budget. New York City, for example, faced a $7 billion shortfall (the difference between projected revenues and the projected costs of maintaining current levels of service) at the outset of the current budget year against the backdrop of the losses associated with the World Trade Center catastrophe, of a national recession of unknown duration and local impact, of over $41 billion of bonded indebtedness and deferred maintenance from the actions of “fiscal conservatives” who increased spending, cut taxes, and borrowed heavily, of a state government facing a large projected deficit, and of a federal administration that is reliably sympathetic in public statements and less reliably forthcoming with the hard cash.

Local governments are subject to intrusions from state, federal, and other local

governments that change their resource allocation problem. States may impose unfunded mandates for local services or further constrain local sources of revenue. The federal government may increase the costs of pension or health care benefits, important

Page 24: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

24

components of local budgets. Local governments for adjacent or overlapping jurisdictions may reconfigure roads such that the traffic flows create severe problems for the local government.

Local government’s resource allocation problem is sensitive to all manner of

exogenous shocks. A recession may both lower tax revenues and increase demands for service. Harsh weather may reduce fee revenues, disrupt planned playground activities and/or drive up costs in other parts of the budget, e.g., greater than expected snow removal costs putting pressure on the overall budget felt in all other accounts.

Even if the would-be modeler assumes stability in the external environment, the

internal environment for deciding anything about playgrounds is still potentially very complex, especially if you frame the allocation problem as microeconomic theory does. In 1935, Simon noted the infeasibility of the economic approach for local officials: “I did not see how it could be done. How were the values of better activity leadership to be weighed against the values of more attractive and better-maintained neighborhood playgrounds?” (Simon, 1991, P. 370) This perceived difficulty is only part of the analytic problem. Before confronting the difficulties of integrating incommensurate values, there must first be estimates of what is to be valued.

Say Milwaukee is contemplating the addition of one playground with a specific

location, rules for access, and configuration for supervised activities. This is one possible component of Simon’s full allocation problem. What information exists or could be created to inform this decision?

The costs are pretty straightforward unless it is discovered after the acquisition that

the planned site has unacceptable levels of toxic material in the soil or the prime contractor for developing the site declares bankruptcy and moves to South America or a tragic accident at an existing playground dramatically increases accident insurance premiums for each and every park or creating the playground requires removing a stand of first-growth hardwoods on the proposed site that is reportedly habitat for an endangered species of squirrel or rates of vandalism are increasing or...

Estimating the probable benefits of the new playground is harder. All potential

benefits from this playground arise in the first instance from the numbers and types of people who will use the playground. The first analytic chore in estimating benefits is to predict marginal changes in use. How many users will use the new playground with the planned location and activities? How many of the users of the new playground are net new users as opposed to people who would have otherwise used another playground? How many people who are currently using the proposed playground site for recreational and other purposes will be displaced by the creation of the playground? What are the characteristics of the predicted new users in terms of such factors as age and socioeconomic status?

It is not at all clear how these predictions could be reliably done by local officials or

by a team of world-class scholars with a multi-year, multi-million dollar project. There is

Page 25: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

25

no history on a new playground to serve as the basis of prediction. There are probably no directly comparable playgrounds in this city or in another city whose experience is directly useful to the predictions. There are no extant planning models to predict reliably how potential users might respond to the new playground.

The fact that the data and models for predicting playground participation rates are in

shambles is not evidence of local officials’ nonfeasance. These data collection and modeling problems, necessary and feasible in neoclassical fantasies, are impossible in practice. How would you have local officials gather data on participation rates for a single, open access playground? Who counts as a “user” for this facility? The retiree or out-of-town visitor out for a stroll? The mother watching a child playing in a sandbox? The teenagers who spend most of the daylight hours on weekends shooting hoops? For those you would count, how do you propose to determine their age, sex, race, and socio-economic status?

Rather than confront this obviously difficult predictive problem, an economist might

simply survey some significant sample of conceivable users to estimate their likelihood of actually using the new facility if it is created and to estimate their “willingness to pay” for what is proposed to be a free facility. The sampling, instrumentation, and interpretation issues with such a procedure are horrendous. The resulting estimates are neither robust nor predictively useful. But the economist can talk authoritatively if not verifiably to like-minded colleagues and lay audiences about “consumer surplus” and perhaps even “net present value” if the estimates are part of a larger study.

Now for a really hard part. Assume that you have somehow succeeded in predicting

the levels and composition of participation at your new playground. What are the effects of the predicted increases in participation? Are there effects in the form of lesser congestion and higher quality experiences for users of existing playgrounds from the departures to the new playground? What are the marginal differences in the quality of experiences? What is the effect on the life of one child of either gender participating in organized baseball at your new playground? What would that child have done with the time in place of playing ball? Pickup games in their backyard? Quality time with their parents? Computer games? Studying mathematics? Writing poetry? Praying? Is the value of a 10-year old from a poor family per hour playing baseball or just hanging out with their friends at the new playground greater than the per hour value for a 15-year old from a wealthy family? You don’t have to dig very deeply into the analytics of evaluating the proposed new park before difficulties overwhelm.

Now for the really, really hard part. Assume that you have somehow succeeded in

predicting the valued consequences, costs and benefits, of the contemplated new playground. Unfortunately, you’re not done if you aspire to be homo oeconomicus. You’ve only analyzed a single alternative in this frame when there are an infinite number of other ways of using the same resources for playgrounds and other purposes. How do you know that there is not a higher yield if the equivalent resources are used for a playground at another site or for enhanced programming on existing playgrounds or for some entirely different function of the government such as public safety?

Page 26: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

26

The full potential complexity of the task environment for Simon’s playground

allocation decision is revealed in microeconomics’ frame for the decision. The frame requires predicting ex ante a host of conceivable effects and their value at the time the effects are experienced. It is not at all surprising, as Simon noted in 1935, that real decisionmakers develop much simpler representations of the allocation task environment, representations that are not even remotely related to the economists’ representations. These simpler representations ignore much of the potential complexity and enable decisionmakers to actually make allocation decisions. We do not know exactly how they do this. We still need to know.

Perhaps your local government is deciding “as if” they are doing all of these benefit-

cost computations; that is, they somehow arrive at the optimal allocations by some mystical process that avoids collecting all of the information, understanding the world in its full causal glory, and applying valuation and choice calculi. Perhaps not. Unfortunately, we do not really know because the foundational science has not been done.

In some respects, allocation decisions are potentially easier to study than many other

occasions that might be carved from the thicket of behavior and labeled decisions. The main decisions are annual and there are rich archival records on what is proposed and what is decided throughout the decision process. Many local officials are willing to talk about the difficulty of the allocation problems they face and how they approach those problems.

The archival record on allocation outcomes will be fairly good. Budgets and

supporting financial documents are matters of public record and usually exist for many prior years in some public library in most jurisdictions.

Yet, the foundational science on Simon’s question is difficult empirically; the

question poses significant challenges for the canonical empirical methods from the academic disciplines.

The archival record is fairly useless with respect to why decisions were made as they

were. The legalisms about authority and process will have much to say about the form of decision and nothing to say about its substance, e.g., the informational premises and criteria for the decision. Even using the archival records on budget outcomes is not straightforward. The terminology and charts of accounts change periodically requiring adjustments to create a temporally consistent record of decisions. Researchers should bring a good text on Fund Accounting with them to confront these records.

Experimentation is not a strong option. It is hard to imagine useful laboratory

experiments; those with adequate controls would have little external validity and those with external validity would be too messy to understand. Field experiments are probably not feasible; budget decisions are too important to their organizations.

Page 27: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

27

Participant-observation, while probably necessary to understand the behavior, is an inherently expensive and conceptually sloppy. There are serious problems in deciding what to observe on decisions so distributed spatially and temporally. There are all of the usual problems in taking decisionmakers’ accounts of what they are doing at face value. These self-reports on behavior are often strategic, e.g., defensive accounts crafted to deflect potential criticism. Also, there is little reason to believe that most local officials understand their behavior in depth and can describe it in terms that directly enable the construction of a model of their behavior. Finally, there is always the danger that the participant-observer will “go native” and lose the ability to abstract from the nominal level of discourse.

Surveys are of little or no value on Simon’s question because they would be

sampling on a process that is not understood. Who do you ask? What do you ask? To what extent can you ask about the decisions directly? How do you aggregate responses from heterogeneous respondents? How do you use the responses to build a model?

Interviews are necessary but have some of the same problems as surveys. The

interviewees will give differing accounts of how the decision was actually made. Most of them will have played only a limited role in the decision and their account will strongly reflect their role. Some interviewees will have axes to grind in their account of how the decision was made.

There are real problems in deciding whom to interview; there were probably dozens

to hundreds of people with some role, however slight, in the decision ranging from facilities personnel providing information about grounds and operations to citizens proposing changes in playground maintenance or programs. When all of the interviews are complete, there is the problem of integrating the various accounts? If the chief executive of the government, an elected member of the legislative body, the head of the operating department, and the supervisors of maintenance and program leadership give discordant accounts, how should these be reconciled into a coherent explanation of how the decision was made?

The second important reason for the lack of scientific foundations to address

Simon’s question is the presence of strong disincentives for those capable of progressing on the conceptual and methodological issues to do so. The discussion above of why academic men are not curious about Simon’s question recounted many of the superficial disincentives associated with academic disciplines and universities.

These superficial disincentives to doing fundamental science are, by and large,

incentives to do other things -- to do that which can attract external funding; to do that which ingratiates you to those senior disciplinary colleagues who will decide your fate at the time of a tenure decision; to do that which yields publication volume and citation counts for a world inept at quantity/quality distinctions and riddled with ambiguous, arbitrary performance standards; to be “relevant” in your applied work; to avoid research topics that might invite ridicule from non-researchers, perhaps one of the dreaded

Page 28: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

28

“fleecing awards” from journalists with impeccable grooming, flawless recitation from a teleprompter, a great sense of how to pander to a mass audience, and no analytical ability.

Simon’s question invites understanding for the sake of understanding. It is

impossible to predict the value of understanding thoroughly how local governments allocate resources. Perhaps the understanding could be better used to design federal and state grant programs in terms of effectiveness on their intended purposes. Perhaps the understanding would help the federal government to achieve the cooperation of state and local governments required for homeland security. Perhaps the understanding would facilitate meaningful reforms and greater efficiency in local governments’ performance of their functions. Until we understand, we simply cannot know the eventual value of understanding.

Understanding for understanding’s sake has always been hard to support but is

falling on even harder times. There is a strong and growing penchant to micromanage research. The problems are most acute in the criteria for research funding where researchers are asked to predict the unpredictable: What will we learn from your proposed research and how and why will that knowledge be valuable? Researchers, like local officials allocating resources to playgrounds, do not behave as simplistic theories assume. One common adaptation by many researchers is to request funding for research just being completed for which the results and possible applications are known; any new funding is used to extend the current research and the new results become the basis for the next funding request and so on. Novel, high-risk projects, however important their success might be, get short shrift from researchers and funders. Most projects that challenge or otherwise deviate from disciplinary paradigms fall in the novel, high-risk category.

The relevance bias against basic research, understanding for the sake of

understanding, is not unique to the social and behavioral sciences. Other scientific fields have, however, somehow achieved greater tolerance of “irrelevant” research. When the recently deceased Max Perutz, for example, began work on understanding the horse hemoglobin molecule with X-ray crystallography in the 1930s, the ultimate relevance of his work in terms of potential applications was not clear. By all accounts (there are several fine obituaries in major British and American newspapers and magazines), his chosen scientific problem was believed to be somewhere between difficult and impossible. When Perutz succeeded in understanding the molecule in the 1950s and received the Nobel Prize in Chemistry in 1962, the ultimate relevance of his work in terms of potential applications was still not clear. There have been a slew of important medical applications based on efforts of thousands of organic chemists and molecular biologists building on Perutz’s work on the proteins in hemoglobin molecules over the past 40 years. It is difficult to say how much credit Perutz should receive for any of the numerous applications of value to humankind that resulted, in part, from his basic research. Perutz laid foundations in technique and understanding that enabled subsequent foundational work and eventually applications.

Page 29: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

29

Simon’s eventual importance to the foundations of the behavioral and social sciences and, hence, to the theories and applications built on those foundations is unknowable much as Perutz’s importance is unknowable. Regardless, the amount of not-immediately-relevant science Simon accomplished in the frame of his original, general question, “How do human beings reason when the conditions for rationality postulated by neoclassical economics are not met?”, was truly remarkable given the powerful disincentives from each of the several disciplines he touched.

Simon was competing in a different game than most social scientists. It is a game

scored by rules familiar to many natural scientists, some behavioral scientists, and a few social scientists.

Science, viewed as a competition among theories, has an unmatched advantage over all other forms of intellectual competition. In the long run (no more than centuries), the winner succeeds not by superior rhetoric, not by the ability to convince or dazzle a lay audience, not by political influence, but by the support of data, facts as they are gradually and cumulatively revealed. As long as its factual veridicality is unchallenged, one can remain calm about the future of a theory. The future of bounded rationality is wholly secure. (Simon, 1991, pp. 364-65)

Simon’s game is a very difficult game to play in disciplinary cultures busily playing other games. A much easier game is join an intellectual clan not subject to the tyranny of facts, learn the clan’s language and levels of discourse, and work hard to earn the respect and admiration of clan members. Clan games are games of status rather than games of consequence. Status within the clan is determined by such factors as the degree of fealty to the clan’s shared beliefs and skill in articulating and extending the clan’s shared beliefs in increasingly arcane languages. The clan may have many brilliant members; the competition for status within the clan can be ferocious. Heretics and apostates are most often shunned rather than attacked because attacks might entail public airings of the clan’s shared beliefs. The clan and its members optimize locally and satisfice globally.

Neoclassical economic theory played an important role in setting Simon’s lifelong

research agenda:

My training in economics, evoked in the context of a budget situation, disclosed a contradiction between what theory taught me ought to be happening and what my eyes and ears showed me was actually happening. Without the training in economics, the observed behavior would have appeared entirely “natural.” Without the observations I could have continued in the happy illusion that the neoclassical theory of utility maximization explains human behavior in the domain of budgeting. And since my exposure to the economics profession was still rather minimal, I had not acquired the habit, so common in that profession, of ignoring the real world when it contradicts the theory. (Simon, 1991, p. 371)

Page 30: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

30

Simon’s response to the contradiction was very much in the spirit of the natural sciences; when the theory does not meaningfully fit the facts, fix the theory. He assumed, somewhat naively at the outset, that the response of economists to discordant facts would be like the response of physicists or chemists to discordant facts.

Theories discordant with experimental facts live on borrowed time in the natural

sciences. While the exchanges between theorists and experimentalists in physics are often intense, even nasty, and play out over decades, the facts ultimately rule in favor of theories that are not disconfirmed; the theories gradually improve in their veridicality through the complicated interplay of facts and theory. Abstract theories that do not make strong, empirically testable claims about how the world actually works are neither predictively nor prescriptively useful. Strong theoretical claims disconfirmed do not persist.

In the social sciences circa 2002, it is unfortunate that those with sufficient

theoretical and methodological training tend to have limited exposure to real decision processes and those with extensive exposure to real decision processes tend to have insufficient theoretical and methodological training. There is very little of the strong interplay between theory and facts so common and fecund in other scientific disciplines.

Theories of behavior that make strong and testable claims, right or wrong, confer

potential benefits. Correct theories afford correct explanations and predictions. Incorrect theories are much more important to understanding behavior and building better theories but only if the theories are allowed to fail and the failure triggers search for the cause of failure.

In many respects, optimization has played a role in twentieth-century behavioral

theories equivalent to the role that phlogiston played in the seventeenth and eighteenth-century theory of chemistry.5 Simon’s theory of humans as information-processing systems may eventually have the same sort of impact on theories of behavior as Lavoisier’s oxidation theory had on the theory of chemistry. Only time and a lot of proper scoring of competing theories with facts by scientific rules will tell.

Optimization as a description of decisionmaking behavior will be harder to supplant

than phlogiston as a description of chemical process because the chemists had many more occasions to predict and build things with serious consequences for being wrong. Those who use optimization descriptively need more occasions to be wrong. Ask for a few ex ante predictions. If ex ante predictions are not possible, ask for at least one specific demonstration of the feasibility of optimization in the specific circumstance purportedly described.

Optimization as a prescription for behavior is already supplanted in many application

domains by heuristic methods that provide provably better but not provably best

5 A nice capsule summary of Phlogistic Chemistry can be found at: http://www.brookscole.com/chemistry/timeline/phlogistic.html.

Page 31: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

31

solutions. The heuristics in many application domains are gradually improving. Satisficing with adaptive aspirations?

Herbert Simon’s lifelong intellectual passion was to understand how human beings,

individually and collectively, actually make decisions and solve problems. He was an anomaly in terms of how he found his problem and somewhat preternatural in his search for a solution. Yet, he never did get around to answering his original question fully:

How are funds divided between playground maintenance and playground activity

leadership in Milwaukee? Good question. Significant man.

References Ando, A., Fisher, F., & Simon, H.A. (1963). Essays on The Structure of Social Science Models. Cambridge, MA: MIT Press. Charnes, A., Cooper, W.W. and Mellon, B. (1952). Blending Aviation Gasolines: A Study in Programming Interdependent Activities in an Integrated Oil Company. Econometrica 20:135-159. Coombs, C.H. (1983). Psychology and Mathematics. Ann Arbor: University of Michigan Press. Gigerenzer, G. (2000). Adaptive Thinking: Rationality in the Real World. Oxford: Oxford University Press. Gigerenzer, G. and Selten, R., (eds.). (2001). Bounded Rationality: The Adaptive Toolbox Cambridge, MA: MIT Press. Kepler, Johannes. (1619). Harmonice Mundi March, J.G., & Simon, H.A. (1958). Organizations. New York: Wiley. Newell, A., & Simon, H.A. (1972). Human problem solving. Englewood Cliffs, NJ: Prentice-Hall. Simon, H.A. (1947). Administrative behavior: A study of decision-making processes in administrative organization. New York: Macmillan.

Page 32: Ask a Simple Question - Carnegie Mellon Universitycasos.cs.cmu.edu/publications/papers/larkey.pdf · 2008. 11. 17. · Ask a Simple Question: A Retrospective on Herbert Alexander

32

Simon, H.A. (1956). Dynamic programming under uncertainty with a quadratic criterion function. Econometrica, 24, 74-81. Simon, H.A. (1957). Models of man: Social and rational. New York: Wiley. Simon, H.A. (1969). The sciences of the artificial. Cambridge, MA: MIT Press. Simon, H.A. (1991). Models of my life. New York: Basic. Tiebout, C. (1956). “A Pure Theory of Local Expenditures.” Journal of Political Economy. 64 (4): 416-424.