Individual decision making in static, sequential and dynamic situations.

Abstract

Static, sequential and dynamic models of decision making situations and additive and subjectively expected utility models of decision making behaviour are defined and discussed in chapter 1. Results from conjoint measurement theory are surveyed in chapter 2 and their application to a qualitative functional analysis (QFA) of the information integration models is discussed. An important issue is how to deal with fallible data. In chapter 3 functional measurement for binary choice data by the method of minimum normit chi square is considered with a view to examining information integration models quantitatively. In the last of the 4 theoretical chapters a selective review of experimental work related to some major issues in decision theory is represented. Six -pair comparison experiments are reported in chapters 5 to 7, in five of them choices were observed and in the other (experiment 2) statements of preference were elicited. The subjects were randomly selected university students whose results were analysed individually. The alternatives available to subjects were gambles for small amounts of money, which they actually played in real play situations. Experiments 1 - 3 were set in static decision¬ making situations, 4 and 3 in sequential ones and experiment 6 was set in a dynamic situation. In experiment 2 the role of indifference in decision making was investigated by QFA and found to be minor. In experiments 1, 3 and 5 information integration models were investigated by QFA and functional measurement. Support for SEU models and not additive ones was found. In experiments 4 and 5 the effects on choices of current capital and previous outcome were found to be negligible. Experiment 5 was a study of an additive information integration model in a simple, two stage, dynamic betting game. A functional measurement analysis led to its rejection. In the final chapter the results are discussed in relation to previous work. Methodological difficulties which arose from the use of the analytic methods are considered and partly resolved. It is concluded that they are suitable techniques for the present application. The future of the information integration models is also discussed. It is concluded that as descriptive models of behaviour in static, sequential and dynamic situations they still have a very useful role to play.