Overview

This collection of (all but two) previously published papers is a sequel to H.R. Arkes and K.R. Hammond's Judgment and Decision Making: An Interdisciplinary Reader (Cambridge, 1986). Similar to the latter, which surveyed the field of JDM in the decade prior to its publication, the present volume is a collection of papers published between 1986 and 1995. The variety of subjects covered reflects the editors' opinion that "JDM researchers have expanded their interests to include nearly everything that one might place under the label of higher-order thought processes, thus blurring the line between JDM research and cognitive psychology (Chapter 1, p. 34). The volume is divided into four parts. Part I, "Introduction", consists of a single chapter (one of the two original contributions to the volume), a concise and clear historical review by Goldstein and Hogarth the literature on preferential choice and judgment under uncertainty, from its roots in Von Morgenstern & Morgenstern and Savage, to the present virtual merging into cognitive psychology. Part II, "Currents", consists of 13 papers. Two papers deal with "anomalies of judgement and choice" (Shafir, Simonson & Tveresky on reason-based choice and Gegerenzer, Hoffrage & Kleinbolting on probabilistic mental models); two deal with "decision processes and their adaptiveness" (Hammond, Hamm, Grassia, & Pearson on the cognitive continuum and quasi-rational decision processes, and Payne, Bettman & Johnson on adaptive decision making); two chapters deal with "acquisition and use of knowledge" (Klayman & Ha on the confirmation bias in hypotheses testing and Hogarth, Gibbs, McKenzie & Marquis on learning from feedback); two deal with "causation, mental simulation, and counterfactual reasoning" (Cheng & Novick on the probabilistic contrast model of causal induction and Kahneman & Varey on close counterfactuals); one paper each deal with "training and expertise" (Camerer & Johnson on forecasting by experts); the "temporal context" (Lowenstein & Thaler on inter-temporal choice) and "groups situations" (Dawes, Van Kragt & Orbell on group identity driven cooperation). Finally, two papers deal with "subjective experience, metacognition and insight" (Kahneman & Snell on predicting future tastes and Tversky & Griffin on judgments of well being).

Part III, "Connections", consists of 4 papers. One paper deals with "Memory" (Hastie & Park on the moderating effects of task type on the relationship between memory and judgment), two deal with "explanations and arguments" (Pennington & Hastie on explanation-based decision making and Hogarth & Kunreuther on argumentation in decision-making under ignorance); and one paper deals with "affect, attitudes and motivation" (Ajzen on the effects of positive attitudes on decision making). Finally, Part IV , "Controversies" consists of 6 papers. One paper deals with "paramorphic models vs. process models" (a follow-up by Doherty & Brehemer on Hoffman's classic paper on paramorphic representation); Two papers deal with "problem representation, domain knowledge, and content specificity" (Wagenaar, Keren & Lichtenstein on the effects of deep vs. surface structures of decision problems on risk seeking and risk aversion, and Goldstein & Weber on how domain specificity affects decision making); one paper deals with "rule-governed vs. rule-described behavior (Smith, Langston & Nisbett on the use of abstract rules); lastly, two chapters dealing with "radical departures and new beginnings" (Tetlock on decision makers as politicians and Lopes on risk).

Looking over the above list I am impressed by the variety of subjects covered in the volume. I am not sure, however, whether to attribute this variety (as Goldstein & Hogarth suggest) to the expanded interests of JDM researchers or to a recognition on their part that work once considered outside the fold of JDM contributes to our understanding of how decisions are made. Either way, this volume supports Hastie's prediction that "expected utility theory�¬will fade away gradually as more and more psychologically valid and computationally tractable revisions of the basic formulation overlay the original" (Hastie, 1991, p. 138).

Recurring themes

Although my graduate training included a chapter on decision making in the early seventies, I never matured into a bona-fide member of the JDM/BDT community. In fact, I later gravitated in heretical directions which have finally gained their own triple-lettered collective label, naturalistic decision making (NDM). I took on to review Goldstein & Hogarth's volume, in order to pay a visit, so to speak, to an old neighborhood. To my pleasant surprise, JDM no longer seems to be as alien as it seemed to me not too long ago. This impression is based on the fact that some 15 chapters speak to one or more of four themes which are perfectly compatible with NDM:

Dissatisfaction with the legacy of the Rational Choice model

NDM was largely motivated by dissatisfaction with the Rational Choice model's legacy to JDM. A principal feature of this legacy is the treatment of gambling decisions as prototypical of virtually all decisions which reduces "all of the considerations involved in a decision �¬to two types of variables, (a) degree of value and (b) degree of belief" (Goldstein & Weber, p. 566). One result of this reductionism is the prevalent practice of studying decision making in sparse laboratory environments using simple monetary gambles that equate uncertainty with probability estimates. As Lopes (quoted in Goldstein & Weber, p. 567) suggests, simple monetary gambles "remain as important to researchers on preferential judgment and choice as fruit flies to genetics". Several authors criticize the legacy of Rational in the present volume. Lopes, Shafir et al, Hogarth & Kunreuther, Smith et al. and Tetlock question the appropriateness of the gambling metaphor. Shafir et al, and Kahneman & Snell cast doubt on the existence of utility functions, showing that decision makers do not order options according to value and have trouble predicting their future tastes, respectively. Hogarth & Kunreuther show that uncertainty is rarely conceptualized as risk (as in the rational choice model), and even then is not dealt with by some form of cost/benefit analysis compatibly with that model. The remaining three themes identifiable in the volume articulate this basic dissatisfaction with the Rational Choice model.

The importance of contingencies and contexts.

Three assumptions underlie the reductionism noted in the preceding section: (1) Descriptive imperialism (assuming that all decisions are made by choosing among alternatives), (2) prescriptive imperialism (assuming that applying the rational choice model is always the preferred method for making "good" decisions). (3) as a corollary, the assumption of invariance posits that descriptive and prescriptive imperialism are valid in all contexts and contingencies.

The assumptions of descriptive imperialism and invariance were challenged by Einhorn (quoted in Payne et al., p. 181) who concluded that "no single model such as additive utility was likely to be an adequate general representation of evaluative decision making [and] proposed that conditions should be specifies under which various models apply as representations of human decision making." Three alternatives to choice are offered in the present volume: Decision making as a process of argumentation (Hogarth & Kunreuther; Shafir et al.), decision making as a process of rule following (Smith et al.) and decision making as story construction (Pennington & Hastie). Underlying all three is a conception of decision making as a form of reasoning which "seems closer to the way we normally think and talk about choices" (Shafir et al., p. 71). Bearing Shafir et al.'s claim out, the alternative conceptions can account for various findings enumerated by the above researchers which are unaccountable by choice models.

The assumption of invariance is challenged by findings regarding the effect of task (and problem) structure on the decision process: Shafir et al. note that "there is a growing body of evidence that people's preferences depend on the context of choice, defined by the set of options under consideration. Hastie & Park resolve contradictory findings regarding memory effects on judgment (e.g., the availability heuristic), by noting the distinction between on-line vs. memory-based task structures. Finally, both Hammond et al., and Payne et al. note that decision makers adjust the cognitive effort which they expend to task requirements. Payne et al and Hammond et al. challenge, in addition, the assumption of prescriptive imperialism. The former (p. 194) found that "subjects showed an ability to take advantage of changes in the structure of the available alternatives so as to reduce processing load while maintaining accuracy" (prescriptive imperialism ignores both task structure and processing load). The latter showed that using analytical processes (entailed by the application of prescriptive choice models) produces the best as well as the worst performance, and hence is inferior to quasi-rational decision processes which involve using both analytical and intuitive processes compatibly with task demands.

The semantics of decision making are as (if not more) important as its syntax

An important aspect of the reductionism that characterizes JDM theory and research is a singular concentration on the syntax or "deep structure" of decision making, treating all substantive aspects of decision problems as superfluous "surface" details. As Goldstein & Weber (p. 566) note, "This state of affairs contrasts sharply with several other areas of psychology, where the semantic content of the stimuli, that is, what the task is "about" has been found to influence behavior, and where energetic research has been undertaken to identify functionally distinct domains of content." The controversy regarding semantics vs. syntax is not unique to the study of decision making. Searle (1995) have recently framed his Chinese Room Argument against strong AI in a way that is applicable in a straightforward fashion to JDM's computational image of decision makers (replace 'programs' by 'formal models' and 'minds' by 'human decision making processes'):

Programs are syntactical.

Minds have a semantic.

Syntax is not the same as, nor by itself sufficient for, semantics.

QED

Four papers in the present volume pertain to the issue of semantics vs. syntax. Goldstein & Weber and Lopes provide excellent discussions of the poverty of describing decision making purely in terms of its syntax; Wagenaar et al. and Dawes et al., provide empirical evidence for the importance of the semantic content of decision problems. "Starting from �¬83% preference for the certain outcome [Wagenaar et al.] have, after a number of changes in the problem's surface structure, finally come to 92% preference for the uncertain outcome�¬[illustrating] that the jump from surface structure to deep structure might be tricky." (pp. 562-564). Dawes et al. showed that a variable that is extraneous to game theory, group identity (playing with total strangers vs. playing with acknowledged same group members) radically affects cooperation levels in prisoner dilemma situations. These series of carefully conducted laboratory experiments brought to my mind the following episode which convinced me, long ago, that concentrating on the syntax of decision making , and construing it as choosing among alternatives rather than as semsemaking-driven action, is akin to missing the wide side of the barn:

Writing about his experiences in the Spanish civil War, George O rwell recounts an episode in which he was lying in the trenches waiting for an opportunity to shoot at the enemy. At last a dis turbance took place and a man "jumped out of the trench and ran along the parapet in full view. He was half-dressed and was hol ding up his trousers with both hands as he ran. I refrained fro m shooting at him�¬I did not shoot partly because of the detail a bout the trousers. I had come here to shhot at 'Fascists'; but a man holding up his trousers is'nt a 'Fascist', he is visibly a fellow-creature, similar to yourself, and you don't feel like s hooting at him.

(Looking back on the Spanish Civil War. In A Collection of Essa ys, Garden City, NJ: 1957, p. 199).

The decline of heuristics and biases

Kahneman and Tversky's research program on heuristics and biases had a tremendous impact on JDM. In the second edition of his popular text, Hogarth (1987) enumerated 30 biases or sources of biases that characterize human decision making and information processing. In stark contrast no paper in the present volume is written in this stance (except for Camerer & Johnson's paper on the poor performance of experts on forecasting tasks). Furthermore, consistent with the preceding three themes, four papers challenge this stance directly: Hammond et al. show (as noted above) that the alleged superiority of formal models disappears when intuitive judgment is not pitted directly against them but rather against these models as used by human agents. Gigerenzer et al. "avoided classifying judgment as either rational or biased, but instead focused on the underlying cognitive processes and how these explain extant data�¬One cannot speak of a general overconfidence bias anymore�¬[C]hanges in the task�¬and in the relationship between task and environment�¬can make the two stable effects reported in the literature �¬ overconfidence and the hard-easy effect �¬ emerge, disappear, and invert at will (pp. 135-137). Klayman & Ha showed that 'confirmation bias' is not a basic characteristic of human inferential processes. Rather, it is produced by inappropriate application of a more general "positive test strategy [that] is actually a good all-purpose heuristic across a range of hypothesis testing situations�¬[partly because] in probabilistic environments, it is not even necessarily the case that falsification provides more information than verification. What is best depends on the characteristics of the specific task at hand (pp. 234-235). Finally, Cheng & Novick present a normatively warranted descriptive model of causal induction and show that previously demonstrated systematic deviations from presumably "correct" causal inference could be due to discrepancies between what subjects and researchers define as the focal set of events.

Conclusion

One should not understand from my review that Goldstein & Hogarth produced a volume on NDM. Three notable features that still distinguish JDM from NDM are the latter's conception of decision making as sequential option evaluation driven by situation assessment (as opposed to concurrent choice), the study of decision making in naturalistic settings (or high fidelity simulations), and the value of applied research for theory development. Notwithstanding these differences, assuming that Goldstein & Hogarth selected a representative sample of JDM papers, and that my interpretation of these papers does not misrepresent their content or spirit, the time seems ripe for a fruitful dialogue between the two disciplines. To this end researchers within NDM should stop worrying about the deficiencies of JDM �¬ there is little the former can tell the latter which they have not discovered on their own, and researchers of JDM should test ideas and findings coming from NDM instead of rejecting them outright on the grounds of "weak" methodology. After all, multi-method research is valued in both communities. Why not divide the work then?

References

Hogarth, R. M. (1987). The psychology of judgement and choice. San Francisco, Ca: Jossey Bass. (2nd ed.).

Searle, J.R. (1995). The mystery of consciousness. The New York Review of Books, November 2, 60-66.