Belief and Preference in Decision Under Uncertainty

Abstract

Most decisions in life are gambles. Should I speed up or slow down as I approach the yellow traffic light ahead? Should I invest in the stock market or in treasury bills? Should I undergo surgery or radiation therapy to treat my tumor? From mundane choices rendered with scarcely a moment's reflection to urgent decisions founded on careful deliberation, we seldom know in advance and with certainty what the consequences of our choices will be. Thus, most decisions require not only an assessment of the attractiveness of potential consequences, but also some appraisal of their likelihood of occurrence.Virtually all decision theorists agree that values and beliefs jointly influence willingness to act under uncertainty. However, there is considerable disagreement about how to measure values and beliefs, and how to model their influence on decisions. Our purpose in this chapter is to bring into sharper focus the role of values and beliefs in decision under uncertainty and contrast some recent developments in the descriptive modeling of choice under uncertainty with the classical normative model. The Classical Theory and the Sure-Thing PrincipleThe primitives of most decision theories are acts, states, and consequences (Savage, 1954; for an alternative approach, see Luce, 2000). An act is an action or option that yields one * The view of decision making under uncertainty outlined in this chapter was heavily influenced by the late Amos Tversky. Of course, any errors or deficiencies of the present work are entirely the responsibility of the authors. We thank

Contrasting stochastic and support theory accounts of subadditivity

“…First, a number of studies have shown that when descriptions of events (e.g., ''precipitation next April 1'') are unpacked into a disjunction of constituents (e.g., ''rain or sleet or snow or hail''), judged probability sometimes increases, but not as dramatically as the sum of the probabilities of these constituents when they are judged separately (Rottenstreich & Tversky, 1997;Fox & See, 2003). In fact, unpacking descriptions into a disjunction of constituents and a catch-all sometimes leads to a reduction in judged probability.…”

Weighting Under Ambiguous Preferences and Imprecise Differences in a Cardinal Rank Ordering Process

“…Naturally, many concepts, such as knowledge, beliefs, and preferences, are not naturally represented with numeric precision (cf., e.g., Refs. [13][14], and the difficulty of eliciting precise decision parameters (utility values, criteria weights, and probabilities on uncertain consequences) have been widely discussed 21 . There are severe inconsistencies with the predictions of the rational model (cf., e.g., Refs.…”

Competence effects for choices involving gains and losses

“…Consider a lottery in which the decision maker wins x if event E occurs. The two-stage model posits that individuals first judge the probability of event E, then transform this subjective probability by the probability weighting function for risk (see also Fox and See 2003;Fox and Tversky 1998;Wu and Gonzalez 1999). The two-stage model proposed by Tversky and Fox, however, does not permit source preference, and thus cannot accommodate the Ellsberg Paradox.…”