INSIDE MAN

The Discipline of Modeling Human Ways of BeingBy Mihnea C. Moldoveanu

Stanford University Press

Copyright © 2011 Board of Trustees of the Leland stanford Junior University
All right reserved.
ISBN: 978-0-8047-7304-1

Contents

Figures and Tables.............................................................xiPrelude and an Outline.........................................................xv1 Introduction: What Are Models For?...........................................12 Decisions and Choices, Immediate and Planned.................................213 Beliefs and Believing........................................................874 Calculating, Reasoning, and "Doing Things with the Mind".....................1335 Learning and Learning to Learn...............................................2136 Communicating: Turning Oral Noises into Speech Acts..........................233Coda A "Cruel science": What ask Is Up To......................................251Bibliography...................................................................257Index..........................................................................263

Chapter One

INTRODUCTION What are Models For?

Wherein the aims and goals of ascriptive science—the discipline of making valid ascriptions of mental states to humans on the basis of observing their behavior—are introduced and illuminated against the current intellectual landscape of normative, descriptive, and prescriptive approaches to the human sciences and against the background of strands of hermeneutics, the philosophy of language, and the theory of rational choice inherited from a century and a half of writing and thinking about the subject.

WHAT FOLLOWS is an attempt to build a rigorous science of human action from a repertoire of moves and conceptual structures provided by decision and rational choice theory, classical epistemology, artificial intelligence, the philosophy of language, and the experimental methods and results of cognitive and social psychology. The resulting science expressly and explicitly distances the modeler of human behavior and thought from the subject of his or her models. In building this science I make deliberate and frequent use of mathematical and logical models of thinking, believing, emoting, and behaving, and I use them to the explicit end of helping modelers achieve distance from the behavior being modeled with the aim of increasing the precision with which one can represent what and how humans do, think, and feel.

By abducting the essence of the models away from the often lukewarm and fuzzy innards of common English language usage, I hope to accomplish two objectives. The first is to increase the precision with which we can formulate propositions about thinking and behavior and design tests of those propositions. Mathematical representations and first-order logic help greatly with the project of turning quality into quantity or scale, which is important for ascertaining progress in a field of inquiry—even if not always for progress itself: "Better" is made more precise when it is interpreted as "more accurate" or "more valid." If we want "better" models and can agree on measures of validity and accuracy, then we will be able to know and tell which way we are going when we model. It is one thing to say that John is "poor at reining in his appetite for whipped cream" but quite another to characterize the rate at which he trades off a certain amount of guaranteed whipped cream consumption for other entities he values—including the value he derives from the sustained validity of his self-concept as a being capable of self-control—as a function of various visceral states (such as satiation, hypoglycemia, level of sexual arousal) and of various prototypical social situations in which he finds himself (work-related meeting with bosses, work-related meeting with subordinates, stroll with friends). By showing the payoff that decision science brings to action science in terms of precision, I hope to exculpate the former from the often valid accusation that it is a "toy science"—a banal endeavor that is good enough at making all-things-considered ensemble predictions about the behavior of consumers but that cannot and should not be deployed when things begin to matter: in this particular case, in the high-stakes decision scenario, in the one-off interaction that "does one or foredoes one quite." I want, then, to create a decision-science-for-when-it-matters.

The second objective is to capitalize on the distancing effect that is produced when we talk about people as agents or decision agents (or TOTRePs, "trade-off-talking rational economic persons" [Kreps, 1988]) and attempt to measure various quantities that are relevant to our models of these agents—just as when, in a study of "animal learning," we would measure the proclivity of a rat in a maze to exhibit a modification of its behavior in response to a repeated set of pain/reward-mediated stimuli. Indeed, the distancing effects that characterize the modeling approach of traditional rational choice theory and the experimental approach of pre– and post–Cognitive Revolution psychology are, I predict, among the most valuable contributions these fields will be deemed by future historians of ideas to have made to the understanding of human behavior. Universal models—such as those provided by rational choice and decision theory—will be used to create an emotional distoscope, which functions (conversely to what one would expect of an emotional microscope) to produce emotional distance between the modeler and the "modelee"—a move that is particularly helpful when those we wish to model are either ourselves or other "emotionally close" individuals. Thereby, "action science" will become more science-like even as it remains focused on action.

Achieving these goals hinges delicately on what I mean by "models" and what I intend to do with them; delicately because, if misunderstood, the new action science I am aiming for quickly becomes another "discipline"—which I would consider an unfortunate outcome—rather than "a way of living" for those interested in the competent prediction and intelligent production of behavior. so, on to models, then, and their uses.

1. WHAT MODELS ARE FOR: REPRESENTATIONAL AND PERFORMATIVE DIMENSIONS

We have inherited the following picture of models in science: They are representations of "reality," of behavior or thought, that can be used to take us from a set of observable or known quantities or variables (past choices, past measured features of thinking) to a set of predictions of future—or otherwise unobserved, and thus unknown—quantities or variables (future behavior, hidden and private features of thinking). Models embed within them algorithms and formulas for predicting the evolution of observables. Thus, a simple answer can be given to the question of "why model?": to make inferences about what we do not know on the basis of what we do know. If I observe Mathilda choose white bread from a bread-stand that offers both white and wheat bread, then I can model Mathilda as an individual who instantiates—through her behavior—a set of preferences (of white over wheat bread, in this example) and use it to infer that she will choose white bread over wheat bread the next time she has a choice between these two options. The representation of Mathilda's behavior as the instantiation of a set of preferences (which...