Conclusion

Abstract

This book was on rational decision making with the aide of artificial intelligence. The classical definition of a rational agent is an agent which acts to maximize its utility. Utility is a difficult concept to grasp. It is classically defined as the ability of an object to satisfy needs. Much has been written about utility and its derivative expected utility. One definition of utility which dominates the economics field is that it is a representation of the preference of some goods or services. Samuelson attempted to quantify utility as a measure of the willingness of the people to pay for a particular good. In this book we view utility as a measure of the value of some goods less the cost associated with the acquisition of such goods. This simply means that we define utility as the value that is derived from a good minus the cost of that good. Thus if the good is quite valuable but is also equally expensive then its utility is zero because its value and cost balance out.

Conclusions and Further Work

This book proposes that rational decision making is executed by using two primary drivers and these are causality and correlation. In fact embedded in artificial intelligence methods for prediction which is the basis of rational decision making using artificial intelligence are the correlation and causal machines. This book defines a causal function as the function which maps the input to the output where there is a flow of information from the cause to the effect. In philosophy this is what is termed as the transmission theory of causality . The importance of causality on understanding many vital systems has been studied in fields such as child development (Kegel and Bus 2014), wind turbines (Tippmann and Scalea 2014), heart studies (Seiler et al. 2013) and in philosophy (Chaigneau and Puebla 2013).

This book formulated causal function within the context of rational decision making. The causal function was implemented using rough sets to maximize the attainment of an optimal decision. The rough sets which is a type of artificial intelligence was used to identify the causal relationship between the militarized interstate dispute (MID) variables (causes) and conflict status (effects) such that whenever the MID variables were given the probability of conflict was then estimated. Rough sets has modelled diverse problems successfully such as rural sustainability development (Boggia et al. 2014), in web search (Yahyaoui et al. 2014), fault classification of gas turbines (Li et al. 2014), customer segmentation (Dhandayudam and Krishnamurthi 2014) and hepatitis (Srimathi and Sairam 2014). As a way forward other artificial intelligence approaches should be explored to classify them in terms of effectiveness on the construction of a causal function given a particular application and hopefully a generalized conclusion can then be drawn.

Another important area that is studied in this book is the concept of correlation function which is a function which maps the input to the output where there is no flow of information from the cause to the effect. The concept of correlation has been applied in many areas such as constructing a function generator (Li et al. 2014), to study the mechanism of cell aggregation (Agnew et al. 2014), in crystal studies (Jaiswal et al. 2014) and in damage detection (Ni et al. 2014). The correlation function is implemented using support vector machine and succesfully applied to identify the correlation relationship between the electroencephalogram (EEG) signal and eliptic activity of patients. Support vector machines have been used successfully in the past to model EEG signal by Yin and Zhang (2014), Kumar et al. (2014), Yu et al. (2013), Panavaranan and Wongsawat (2013) as well as Bhattacharyya et al. (2013). For future work other artificial intelligence techniques should be applied in other complex examples to model a correlation function.

This book also introduced the concept of missing data estimation as a mechanism for rational decision making. This assumed that there is a fixed topological characteristic between the variables required to make a rational decision and the actual rational decision. This technique was successfully operationalized using an autoassociative multi-layer perceptron network trained using the scaled conjugate method (Chen 2010; Mistry et al. 2008; Duma et al. 2012) and the missing data were estimated using genetic algorithm (Aydilek and Arslan 2013; Azadeh et al. 2013; Canessa et al. 2012; Devi Priya and Kuppuswami 2012; Hengpraphrom et al. 2011). This technique was successful used to predict HIV status of a subject given the demographic characteristics. For the future other missing data estimation techniques should be used for rational decision making.

This book introduced the concept of rational countefactuals which is an idea of identifying a counterfactual from the factual and knowledge of the laws that govern the relationships between the antecedent and the consequent, that maximizes the attainment of the desired consequent (Hausman and Woutersen 2014; Scholl and Sassenberg 2014; Vaidman 2014; Van Hoeck et al. 2014; Wang and Ma 2014). This is primarily intended to either avoid previous mistakes or reinforce previous successes. In this chapter in order to build rational counterfactuals neuro-fuzzy model (Airaksinen 2001, 2004) and genetic algorithm were applied (Martin and Quinn 1996; De Faria and Phelps 2011). The theory of rational counterfactuals was applied to identify the antecedent that assure the attainment of peaceful outcome in a conflict situation. For the future the concept of rational counterfactuals should be applied to other complex problems such as in economics and engineering sciences.

This book advanced the theory of flexibly bounded rationality which is an extension of Herbert Simon’s theory of bounded rationality where rationality is bounded because of inadequate information to make a decision, limited processing capability and limited brain power (Ding et al. 2014a, b; Shi et al. 2014). In flexibly bounded rationality inadequate information is variable because of advances in missing data estimation techniques, processing power is variable because of Moore’s Law where computational limitations are advanced continuously and limited brain power which is more and more being replaced with artificial intelligence techniques (Frenzel 2014). The multi-layer perceptron network and particle swarm optimization were applied to implement the theory of flexibly bounded rationality in the problem of interstate conflict. For the future work, this technique should be tested in other complex problems and using other artificial intelligence techniques.

This book dealt with the concept of using relevant information as a basis of rational decision making. It proposed three methods for making rational decisions by either marginalizing irrelevant information or not using irrelevant information. In this regard four techniques were considered and these were marginalization of irrationality approach, automatic relevance determination (Smyrnakis and Evans 2007; Fu and Browne 2007), principal component analysis (Evans and Kennedy 2014; Yan et. al. 2014) and independent component analysis (Zhan et al. 2014; Chen and Yu 2014). These techniques were applied to condition monitoring, credit scoring, interstate conflict and face recognition. For the future these techniques should be applied to other more complex problems.

This book considered the concept of group decision making and how artificial intelligence is used to facilitate decision making in a group (Bolón-Canedo et al. 2014; Hailat et al. 2014). Four group based decision making techniques were considered and these were ensemble of support vector machines which were applied to land cover mapping, condition monitoring, incremental learning using genetic algorithm which was applied to optical character recognition, dynamically weighted mixtures of experts which were applied to platinum price prediction as well as the Learn ++ which was applied to wine recognition. For the future, other forms of ensemble such as products of experts should be considered for further studies.