Successful or unsuccessful search for risk defusing operators: Effects on decision behaviour

Abstract

In experiments with quasi-realistic risky scenarios instead of gambles, decision makers are less interested in probability information. Often, they actively search for a risk defusing operator (RDO; an action to be performed in addition to a specific alternative and expected to decrease the risk involved). Examples in daily life are insurance and vaccination. In our experiment, the central independent variable was the successful or unsuccessful RDO search. The central dependent variables were choices and information search behaviour. In order to get information about the applied heuristics, a concurrent Thinking Aloud procedure was employed in addition to the method of Active Information Search. Eighty subjects made a choice in two risky scenarios. The findings confirm that the successful search for an RDO for one of the alternatives is an excellent predictor of choice. If the subject does not search for an RDO or the search is unsuccessful, MAXIMIN was the most frequent heuristic.

Acknowledgements

We would like to thank Odilo W. Huber, Siegfried Macho, Michael Schulte-Mecklenbeck, and three anonymous reviewers for their comments on an earlier version of the manuscript.

Notes

¹The search for probability information is independent of whether a decision maker chooses a risky alternative or a nonrisky one. This independence rules out the alternate explanation that people do not search for probability information because they choose the nonrisky alternative anyway. Furthermore, in part of the scenarios, all alternatives are uncertain. Another alternate explanation would be that people do not search for probability information because they generally introduce or infer probabilities from their background knowledge. Huber and Macho (2001) demonstrated that subjects did not introduce or infer usable probability information in four different quasirealistic tasks. Therefore, this alternate explanation can be discarded as a general one. For a more elaborate discussion, see Huber (2007).

²In gambles and in tasks prestructured like a gamble, in general, outcomes and probabilities are presented to the subjects (and nothing else) and thus a “classical” gamble structure of the decision alternatives is introduced by the experimenter and usually adopted by the participants.

³This category system differs from that used in Huber et al. (1997) or Schulte-Mecklenbeck and Huber (2003). The reason is that in the meantime the different types of RDOs have become clearer.

⁴When using the Lexicographic heuristic (cf. e.g., Svenson, 1979) in a multidimensional choice, a decision maker first selects the most important dimension or attribute and chooses the alternative that is best on this dimension. If no decision is possible on the most important dimension, the procedure is repeated with the next important dimension. With the Weighted Pros heuristic (Huber, 1979) the decision maker forms two sets of dimensions when choosing between alternatives x and y: one set of dimensions, where x is better than y, and another set, where y is better than x. Alternative x is chosen if the set of dimensions speaking for x is more important than that speaking for y.