Authors: Doraid Dalalah; Khaled A. Alkhaledi
Addresses: Industrial and Management Systems Engineering Department, College of Engineering and Petroleum, Kuwait University, Kuwait ' Industrial and Management Systems Engineering Department, College of Engineering and Petroleum, Kuwait University, Kuwait
Abstract: When a decision maker encounters a decision making problem of uncertain outcomes, he/she tends to estimate the so called certainty equivalent of the anticipated rewards, which resembles the benefit measure per unit increase in the expected payoff. Such certainty estimation results in underestimation and sometimes overestimation of the possible rewards. Hence, when two decision problems of uncertain outcomes are to be compared, the decision maker may unintentionally prefer one alternative over another due to the risk imposed by the uncertainty of the outcomes, while in reality the best choice is the other way around. In this paper, we present a model to resolve the complications of overestimation/underestimation of an individual's certainty equivalent by introducing a stochastic non-excepted utility model that adds an error component to the estimated certainty equivalent by the decision maker. The error distribution is optimised via training datasets. By stochastic representation of the expected utility and its corresponding certainty equivalent, we can resolve the decision making situations when the decision maker is risk averse/seeker. To demonstrate the merits of the presented model, different datasets are tested; the model shows a remarkable prediction capability of human choices under risk and uncertainty.
Keywords: non-expected utility; stochastic utility; cognitive thinking; decision making; risk; uncertainty; gambling; error modelling; human errors; certainty equivalents; reward estimation; anticipated rewards; human choices.
International Journal of Applied Decision Sciences, 2016 Vol.9 No.3, pp.307 - 319
Received: 01 Mar 2016
Accepted: 30 Jun 2016
Published online: 20 Dec 2016 *