New trends in operations research: Introducing behaviour in modellingin modelling

Operational Research(sometimes also called Operations Research, OR) is a discipline, often considered to be a sub-field of applied mathematics, that uses the application of advanced analytical methods for problem solving and decision making. The terms man-agement science and decision science are sometimes used as approximative synonyms.

The methods used include: data analysis, modelling, optimisation, simulation, control theory, network analysis, queuing theory, game theory, and so on. Speaking of OR as of a discipline can be considered misleading because of the interdisciplinary character of OR and of management in general, given its inherent human-technology interaction, because of multicriteria decision and decision under uncertainty with incomplete information, as well as because of its focus on practical applications. Methods of OR are successfully used in social sciences like business, finance and management, in engineering, in psychol-ogy and organisation science for production management, service science, policy making, but also in natural sciences, energy modelling, pollution control, climate policy design and many more domains.

Behavioral Research is a research that involves the application of the behavioral and social sciences to the study of the actions or reactions of persons⁵². Behavioral research refers to the study of the variables that impact the formation of habits and decisions that impact many areas of daily life. In market research, behavioral research is based on "the exhaustive rendering of our conscious and unconscious patterns into data

52definition by American Heritage Dictionary http://ahdictionary.com

New trends in operations research: Introducing behaviour in modelling 57 sets and algorithms" (Duhigg, 2010).

Behavioral Operational Research is the study of behavioural aspects related to the use of operations research methods in modelling, problem solving and decision sup-port.

Operational Research as a tool for management and decision making is intrinsically interdisciplinary. In the areas of risk management, decision analysis and in multi-criteria decision making, there is a strong behavioural component, yet the approach is keeping its rigorous mathematical methodology. OR methods spread into judgment analysis, cogni-tive psychology, organisation theory, game theory and economics. OR is about solving real-life problems and these incorporate human behaviour as a major issue.

The central concept of behaviour is decision making. Decision making has been for-malised in Choice theory which is a part of Operations Research (OR). From the OR side, a choice is the outcome of a sequential decision making process. From the sociology side, a choice is the act of choosing between two or more possibilities that have been compared in a deliberative process. A definition of a decision from the OR side is that it is the process of making choice. In sociology, a decision is a conclusion or resolution reached after deliberation.

In the Choice theory, a choice as an outcome of decision making process has the following attributes:

• Definition of the choice problem - for example, what bulb to buy, or how to get from point A to point B.

• Generation of alternatives - what bulbs are available on the market, or transporta-tion modal choice.

• Evaluation of the attributes of the alternatives - price, luminosity, lifetime of a bulb, or time, price, comfort, flexibility of a transportation mode

• Choice - decision rule

• Implementation - the process of acquiring equipment (investment), or travel in transportation

Choice theory defines

• a decision maker

• alternatives

• attributes of alternatives

• a decision rule

The decision maker can be an individual or a group of persons with no internal inter-activity within the group, which is characterised by their socio-economic attributes: age, gender, income, education, social status, etc. Alternatives may be discrete or continuous,

like a choice of travel itinerary and driving mode that can be anything from defensive to aggressive. The choice set generation is characterised by its availability (availability of a choice) and the awareness (information about the availability of a choice). Attributes can be generic or specific, quantitative of qualitative, based on perception or on objective facts, etc.

A mathematical expression of utility, an important concept of neoclassical economic theory, can be formulated as follows.

Let Cn be a continuous choice set of alternatives a of the individual n and & the preference operator. The properties of reflexivity, transitivity and comparability can be described in the following way.

1. reflexivity: a&a for ∀a∈Cn

2. transitivity: if a&band b&c ⇒ a&c for ∀a, b, c∈C_n 3. comparability: eithera&bor b&a for ∀a, b∈C_n

Then the utility functionUn has the form

∃U_n:C_n→R:a U_n(a) such that

a&b ⇔ Un(a)&Un(b) for ∀a, b∈Cn

Utility is a latent concept and cannot be observed directly. The utility function is important because it defines the decision rules.

We will need this definition of utility function later in section 4.7 when describing the Share-of-Choice MARKAL.

The Choice theory in psychology has a different construction. According to Glasser (1998), behaviors we choose are central to our existence. Our behavior (choices) are driven by six genetically driven needs (survival needs): food, clothing, shelter, breathing, personal safety, security. Also, by further four fundamental psychological needs: Belong-ing/connecting/love, Power/significance/competence, Freedom/autonomy, Fun/learning.

Glasser introduces the concept of "Quality World" whose images are our role models of an individual’s "perfect" world of parents, relations, possessions, beliefs, etc. Throughout our lives, each person places significant role models, significant possessions and signifi-cant systems of belief (religion, cultural values, and icons, etc.) into a mostly unconsicous framework Glasser called our "Quality World". Behavior ("Total Behavior" in Glasser’s terms) is made up of these four components: acting, thinking, feeling, and physiology.

An individual has control or choice over the first two of these, but the latter two are

New trends in operations research: Introducing behaviour in modelling 59 more sub- and unconscious.

The ten Axioms of Choice Theory are drawn on the Web site devoted to the works of Dr. William Glasser, http://www.choicetheory.com :

1. The only person whose behaviour we can control is our own.

2. All we can give another person is information.

3. All long-lasting psychological problems are relationship problems.

4. The problem relationship is always part of our present life.

5. What happened in the past has everything to do with what we are today, but we can only satisfy our basic needs right now and plan to continue satisfying them in the future.

6. We can only satisfy our needs by satisfying the pictures in our Quality World.

7. All we do is behave.

8. All behaviour is Total Behaviour and is made up of four components: acting, thinking, feeling and physiology.

9. All Total Behaviour is chosen, but we only have direct control over the acting and thinking components. We can only control our feeling and physiology indirectly through how we choose to act and think.

10. All Total Behaviour is designated by verbs and named by the part that is the most recognisable.

As we see, there are important differences in Choice theory from the applied mathe-matics (OR) perspective and the psychology side. However the methodology is not that different as it is seeking to describe the human behaviour while sticking to the observed reality as much as possible. Engineering models, created to describe objects and their relationships within an energy system, are much more normative.

While the traditional research in OR is still limited to developing mathematical meth-ods, optimisation techniques, convergence of algorithms, there is an emerging stream of modelling effort in behavioural OR. Already Herbert Simon, when setting up the Be-havioural Model of Rational Choice, has drawn a mathematical model of behaviour that opened the door to a new discipline (Simon 1955, pp. 115-118).

Von Winterfeldt and Edwards (1986) synthetise concepts and ideas from economics, statistics, psychology, operations research, and other disciplines, on the platform of be-havioural research relevant to decision analysis. Typically, application of a concept within one discipline does not recognise that other disciplines may have considered the same or similar problems. This is also true in case of decision theory. French and Insua (2000) provides an overview of the main ideas and concepts of statistical decision theory and places it within the broader concept of decision theory, decision analysis and decision sup-port as they are practised in many disciplines beyond statistics economics, operational

research, including artificial intelligence, philosophy and psychology. Recent publica-tions include French et al. (2009) that draws together results and observapublica-tions from decision theory, behavioural and psychological studies, artificial intelligence and infor-mation systems, philosophy, operational research and organisational studies. Franco and Montibeller (2010) bring a formal definition of the process of facilitated modelling and describes its advantages over traditional expert mode when the operational researcher was external to the problem analysis and formulation process and just brought results to the client-formulated problem. Franco and Rouwette (2011) further develops the social aspects and the personal facilitator role, skills and styles as “soft” operational research approach to group decision support research. A review of how Soft OR has dealt with the possibility to use qualitative methods that include subjective beliefs and values to support decision making is in Checkland (2000).

Independently of the operations research, there is a similar technique developed in marketing research, called Conjoint analysis. It is an applied statistical technique to determine how people value attributes (features, functions, benefits) that make up an individual product or service.

The objective of conjoint analysis is to find out what combination of a limited number of key properties, called salient attributes, has the maximum influence on on respondent choice or decision making. The implicit valuation of the individual elements making up a tangible product or an immaterial service is determined by a set of cards shown to respondents. These cards contain a description of properties of potential products or services. The analysis of how consumers make preferences between these products deter-mines the implicit valuations, utilities or part-worths. This represents the core principle of the method called Share of Choice.

In ecological economics, economists value non-market benefits and consumers’ de-mands in monetary terms and to evaluate the utilities derived from non-marketed goods and intangible services in the metric of markets, and then use them in cost-benefit anal-ysis. Many methods have been developed to estimate this kind of non-market values.

They could be divided into five groups:

• Market price valuation methods

• Surrogate market methods

• Production functions methods

• Stated preference methods

• Cost-based methods

Market price methods, surrogate market methods and production functions methods use relevant market prices and revealed preferences to price the non-market goods and intangible services. Alternatively, stated preference methods rely on asking consumers directly to rank their preferences, obtaining the values from consumers’ willingness to pay for getting them. Cost-based methods estimate the values focusing on the costs of

New trends in operations research: Introducing behaviour in modelling 61 supplying, maintaining or restoring the goods and services (Munasinghe, 1992).

Conjoint analysis is a market research tool to determine which attributes are impor-tant on a product or service by studying the trade-offs consumers are ready to make between attributes. Around 1971, Paul Green who specializes in marketing study recog-nized that the non-marketing paper on conjoint measurement, written by Luce and Tukey (Luce and Tukey, 1964), might be used to solve marketing problems such as what affects consumers’ complex purchase decisions, how to estimate consumers’ preferences and how to predict consumer behavior (Green and Srinivasan, 1978). Green and Rao published their paper in the Journal of Marketing Research (Green and Rao, 1971). This paper focused on full-profile conjoint analysis, which is considered as the beginning of conjoint analysis. The full-profile approach utilized the complete set of factors to construct a set of survey cards, and each card described a product profile, then let respondents evaluate each of them, sort them in the order from best to worst. Researchers gathered data and then deduced the most important attributes from all attributes and the preferred levels for each of the respondents. The full-profile approach performed very well when the number of attributes involved was not too big.

According to Green and Srinivasan (1990), interest in this area took off in the 1970s (Green and Srinivasan, 1978). Researchers started to recognize the power of forecasting customers’ choices and borrowed techniques that were developed by mathematical psy-chologists and statisticians in the 1960s (Green and Rao, 1971; Green and Wind, 1975;

Orme, 2006). Srinivasan and Shocker (1973) were the first to formalize the consumer pref-erence problem as an optimisation Share of Choice programming problem (Srinivasan and Shocker, 1973). They made use of linear programming techniques on a multidimensional joint space to find a product location that maximized the number of consumers who have the new-product location closest to their individual ideal points. The distance measures they considered were the weighted and unweighted Euclidean distances, which are pop-ular metrics used in cluster analysis techniques.

Albers and Brockhoff (1977) showed how the problem can be formulated as a mixed integer nonlinear programming problem (Albers and Brockhoff, 1977). Other more re-cent studies include Camm et al. (2006), who made use of an exact branch and bound algorithm and showed that this approach is very appropriate for large-scale problems as well as cases with part-worths containing estimation errors, and Gruca and Klemz (2003), who presented an innovative method that utilizes genetic algorithms to come up with an optimal positioning strategy. Draganska and Jain (2006) showed that yogurt’s consumers value line attributes more than flavor attributes through a discrete-choice model formed by both a demand-side and a supply-side model (Draganska and Jain, 2006). Other studies contain sophisticated operations research models that take into consideration consumer preferences as well as several tests to check the reliability of re-sults obtained from different models (Camm et al., 2006; Green et al., 1991, 1993).

Despite all the studies that focus on the Share of Choice model, very scarce literature can be found on models that combine all the following factors: resource planning, costing, product of service design, and customer preferences. Fragnière et al. (2006) suggested the use of dollar-based methods to create fair pricing schemes that take into consideration a customer’s perceived value for intangible goods. A more concise application of this approach may be found (Fragnière et al., 2008).

In section 4.7 on page 146, we are using a stated preference (conjoint analysis) mixed with a surrogate method (energy optimisation model) through a Share of Choice directly integrated in the energy optimisation model.

Today’s model-based problem solving is an increasingly important approach when dealing with complex, important and urgent multidisciplinary issues involving group decision making, such as global climate change and its human dimensions and new ways in resource management. Our approach of including behaviour into normative energy-economy model is fully in line with this trend.