Aided and unaided decision making with imprecise probabilities (original) (raw)
Related papers
Essays on Decision Making under Uncertainty
2015
Professor Moshe Buchinsky, Co-chair Professor William R. Zame, Co-chair This dissertation consists of three chapters about decision making under uncertainty. Chapter 1: "Testing between Models of Smoking Risk Perceptions" Research in social and health psychology reports that smokers systematically underestimate the personal smoking risk. I build a model that captures potential determinants of smoking risk perceptions to investigate how smoking may cause an underestimation of the risk. The model is based on the premise that smokers have an incentive to be optimistic: because quitting may be hard, they find it reassuring to think that smoking is not so risky. Drawing upon the theoretical framework, I suggest two empirical tests of the model-one using survey data and another based on a laboratory experiment. iii Chapter 2: "Does Uncertainty Cause Inertia in Decision Making? An Experimental Study of the Role of Regret Aversion and Indecisiveness" Previous research has shown that in many situations there is clear inertia in individual decision making-that is, a tendency for decision makers to choose a status quo option. The status quo option may be the result of a previous choice, or may simply be the option designated as the "default." While inertia may simply reflect the fact that individuals view the status quo option as optimal, there are other factors that may explain this observed behavior. I conduct a laboratory experiment to thoroughly investigate two potential determinants of inertia in uncertain environments: (i) regret aversion and (ii) indecisiveness. A decision maker may experience regret when the outcome of a choice compares unfavorably to the outcome that would have occurred had she made a different choice. Alternatively, a decision maker may be indecisive among the options if she does not know the probability distributions over the relevant outcomes. I use a between-subjects design, with varying conditions, to identify the effects of regret aversion and indecisiveness on choice behavior. In each condition, participants choose between two simple real gambles, one of which is assigned to be the status quo. I find that inertia is quite large and that both mechanisms are equally important. Chapter 3: "Risk, Ambiguity, and Diversification" Attitudes toward risk influence the decision to diversify among uncertain options. Yet, because in most situations the probability distributions over outcomes are unknown, attitudes toward ambiguity may also play an important role. In a simple laboratory experiment, I investigate the effect of ambiguity on the decision to diversify. Participants have the opportunity to diversify between gambles; in one condition, all gambles are risky, whereas in the other all gambles are iv ambiguous. I find that diversification is more prevalent and more persistent under ambiguity than under risk. Moreover, excess diversification under ambiguity is driven by participants who stick with a status quo gamble when diversification is not feasible. This behavioral pattern cannot be accommodated by major theories of choice under ambiguity. v The dissertation of Santiago Ignacio Sautua is approved.
The Relative Importance of Probabilities, Outcomes, and Vagueness in Hazard Risk Decisions
Organizational Behavior and Human Decision Processes, 1996
composition (termed ambiguous 1). This raises a funda-This study investigated the joint effects of vagueness mental problem for all variants of SEU since, acon both the probability and the outcome dimensions cording to this model, precision of the probability diof hazard risk, using both pairwise choice and rating mension of risk should not affect the DM's preferences response tasks. Seventy-two subjects evaluated risk and choices. stimuli (concerning either environmental or health Numerous empirical studies on risky decision makhazards), which varied in the level of probability of ing using variations of Ellsberg's original problems (e.g. loss, the amount of loss, and the precision with which Curley & Yates, 1985; Curley, Yates, & Abrams, 1986; each dimension was specified. We found strong evi-Einhorn & Hogarth, 1985) have demonstrated that dence of a consistent individual attitude toward vagueness on both risk dimensions and an almost equal DMs are generally averse to taking risks with imprenumber of vagueness-seeking and precision-seeking cisely specified (vague) probabilities and are often even subjects. Choices were consistent with a decision rule willing to pay a premium to avoid vagueness. Most of based on dimension preference, but attitude toward these problems manipulated the level of vagueness by vagueness significantly predicted choice in cases specifying ranges of probabilities centered on the prewhere a dimension preference rule was not approcise value. Ellsberg's original paradox was demonpriate. An asymmetric preference reversal pattern was strated through inconsistencies in choices between also demonstrated, with many subjects choosing an opgambles and this has been the method used in most tion in a pair that they rated as more risky. ᭧ 1996 subsequent experiments (e.g. Einhorn & Hogarth, Academic Press 1985; MacCrimmon & Larsson, 1979). However, the basic pattern was also obtained with pricing and bidding procedures (e.g. Becker & Brownson, 1964; Yates & In a seminal paper Ellsberg (1961; see also Fellner, Zukowski, 1976), elicitation of certainty equivalent val-1961) used the now-famous two-color urn problem to ues (e.g. Cohen, Jaffray & Said, 1985), and with ratings demonstrate a systematic violation of the Subjective of strength of preference (e.g. Curley & Yates, 1985). Expected Utility (SEU) model: given a choice between Vagueness aversion has been documented in medical, betting on the selection of one of two colors from an insurance, and other decision contexts as well. (For urn containing 100 balls, most people prefer to bet on more details, see a recent comprehensive review by an urn with a known 50-50 composition to one with Camerer & Weber, 1992.) an unknown composition, even if they are indifferent Somewhat surprisingly, individual attitudes toward between betting on either color in either urn. Assumrisk and toward vagueness are not closely associated ing the Decision Maker (DM) has a consistent utility (Cohen, Jaffray, & Said, 1985; Curley, Yates, & Abrams, function, such a preference pattern implies that the 1986; Hogarth & Einhorn, 1990; Schoemaker, 1989, subjective probabilities of the two colors are greater 1991), suggesting that attitude toward vagueness is an in the 50-50 urn than in the urn with the unknown important, distinct and independent, factor in decision behavior. We thank two anonymous reviewers and our colleagues Drs. Claudia Gonzalez-Vallejo and Thomas S. Wallsten for useful comments 1 Although the term ambiguity is widely used in the judgment and on an earlier version of this manuscript. This work was supported in part by a grant from the Research Board of the University of decision literature to refer to imprecisely specified probabilities, it is correctly defined as descriptive of situations which allow for more Illinois at Urbana-Champaign.
Psychological Bulletin, 1994
Much of decision aiding uses a divide-and-conquer strategy to help people with risky decisions. Assessing the utility of outcomes and one's degree of belief in their likelihood are assumed to be separable tasks, the results of which can then be combined to determine the preferred alternative. Evidence from different areas of psychology now provides a growing consensus that this assumption is too simplistic. Observed dependencies in the evaluation of uncertain outcomes and the likelihood of the events giving rise to them are frequent and systematic. Dependencies seem to derive from general strategic processes that take into consideration asymmetric costs of over-vs. underestimates of uncertain quantities. This asymmetric-loss-function interpretation provides a psychological explanation for observed judgments and decisions under uncertainty and links them to other judgment tasks. The decision weights estimated when applying dependent-utility models to choices are not simply reflections of perceived subjective probability but a response to several constraints, all of which modify the weight of risky or uncertain outcomes.
The Irrelevance of the Risk-Uncertainty Distinction
Science and engineering ethics, 2017
Precautionary Principles are often said to be appropriate for decision-making in contexts of uncertainty such as climate policy. Contexts of uncertainty are contrasted to contexts of risk depending on whether we have probabilities or not. Against this view, I argue that the risk-uncertainty distinction is practically irrelevant. I start by noting that the history of the distinction between risk and uncertainty is more varied than is sometimes assumed. In order to examine the distinction, I unpack the idea of having probabilities, in particular by distinguishing three interpretations of probability: objective, epistemic, and subjective probability. I then claim that if we are concerned with whether we have probabilities at all-regardless of how low their epistemic credentials are-then we almost always have probabilities for policy-making. The reason is that subjective and epistemic probability are the relevant interpretations of probability and we almost always have subjective and ep...
Choice under Strict Uncertainty: Processes and Preferences
1996
cision alternatives have several uncertain outcomes In decisions involving choice under strict uncerwhose likelihoods are unknown and cannot be inferred tainty, decision makers' choice heuristics varied from the available information. Besides being a comwidely depending on the task conditions. Two experimon type of decision, studying choice under strict unmental task variables, the format of the outcome districertainty is also important because many other types bution and the presence or absence of a time conof decisions may actually be made as if there were comstraint, were used to test the sensitivity of choice proplete ignorance about likelihoods. In many decisions, cesses and choice preferences in these decisions. With probabilities are simply ignored (Hogarth, 1987) or the no time constraint, decision makers made compensaoutcomes are given equal probability. At present many tory tradeoffs on the displayed information when the disagree about the form of the choice rule which ratiodistribution of outcomes was presented as a range, but they used a loss minimization heuristic when the dis-nal decision makers should use to maximize utility in tribution was presented as variability around the choice under strict uncertainty (Seale, Rapoport, & range midpoint. With a time constraint, decision mak-Budescu, 1995) as opposed to risky choice decisions ers used a loss minimization heuristic when exposed where most choice rules are based on probabilityto the range format presentation of outcomes, whereas weighted outcomes (Kahneman & Tversky, 1979). Bein the midpoint format, they focused on the midpoint cause of the seemingly ad hoc nature of the choice rules outcome. Choice preferences also shifted with changes in choice under strict uncertainty, the information acin time constraint. With no time constraint, decision tually used by decision makers and the way it is used makers preferred high variability alternatives when should be of interest to both normative and experimenoutcomes were presented in range format, but low tal decision researchers. In the present study we are variability alternatives in midpoint format presentation. With a time constraint these preferences re-interested in the choice heuristics used with these deciversed. Decision makers seemed to adapt their choice sions and intend to demonstrate the effects of different processing to the decision conditions rather than ustask conditions (format and time constraint) on the seing some canonical choice rule regardless of the condilection and use of choice heuristics in choice under tions. This result is consistent with the findings of prestrict uncertainty. vious studies that looked at risky choice and risk-free The way people adapt their choice heuristics to differchoice. The implication is that although people may ent decision conditions has recently become an imhave intuitive models of choice under strict uncerportant area of research because decision making heutainty such as loss minimization, the stronger factor ristics appear to be fairly sensitive to changes in the in determining how they make their decisions seems decision task setting. There is growing evidence of a to be the actual task conditions of the decision. ᭧ 1996 Academic Press, Inc. systematic foundation for these adaptations based on the effort involved with and accuracy resulting from the individual choice heuristics (Bettman, Johnson, & A common yet little studied type of decision, known Payne, 1991; Coupey, 1994; Payne, Bettman, & Johnas choice under strict uncertainty, occurs when the deson 1988, 1993). In addition, heuristics provide an explanation for the effects of various decision task conditions on choice preferences.
The descriptive and predictive adequacy of theories of decision making under uncertainty/ambiguity
Journal of risk and uncertainty, 2010
In this paper we examine the performance of theories of decision making under uncertainty/ambiguity from the perspective of their descriptive and predictive power, taking into account the relative parsimony of the various theories. To this end, we employ an innovative experimental design which enables us to reproduce ambiguity in the laboratory in a transparent and non-probabilistic way. We find that judging theories on the basis of their theoretical appeal, or on their ability to do well in testing contexts, is not the same as judging them on the basis of their explanatory and predictive power. We also find that the more elegant theoretical models do not perform as well as simple rules of thumb.