A REVIEW OF BIAS IN DECISION-MAKING MODELS

Abstract

A decision-making model solution is a dependent variable derived from independent variables, parameters and forcing functions. Independent variables collected in linguistic form require intuition which can be potentially biased. A collection of qualitative research papers on bias in models was perused to identify the causes of bias. Decision-making in the manufacturing, finance, law, and management industries require solutions from a complex assortment of data. The popularity of combining decision-making with artificial intelligence (AI) for intelligent systems causes concern, as it can be a predisposition to a true solution. A true solution avoids impartiality and maintains repeated results from a natural phenomenon without favoritism or discrimination. This paper appraised the development of the decision-making environment to identify the path and effect of bias on the variables used in models. The literature reviewed was associated with the design of a decision-making criterion rationalizing the application of variables. The influences on variables were observed with respect to the available resources, environment, and people. This list was further extended to consider the constraints of the resource, customer, network, and regulation fed to the structure. The involvement of bias was founded because of the need for rational decision making, cognitive misperceptions, and psychological principles. The study of variables showed the opportunity for a conscious bias from unethical actions during the development of a decision-making environment. In principle, bias may be best reduced with continuous model monitoring and fair adjustments. Ignoring these implications increases the chance of a bias decision-making model. It also influences the decision result and may be avoided with an ethical and fair quality review. The paper increases the awareness of bias in decision-making and guides actors to the identification and avoidance/reduction of bias effects. This may be a guide for the reduction of the model error to achieve a true solution.