Omega 37 (2009) 1100 -- 1108

Contents lists available at ScienceDirect

Omega

journal homepage: www.e lsev ier .com/ locate /omega

Modelling complex ethical decision problemswith operations research

P.L. Kunscha,∗, I. Kavathatzopoulosb, F. Rauschmayerc

aMOSI Department, Vrije Universiteit Brussel, Pleinlaan 2, BE-1050 Brussels, BelgiumbDepartment of IT-HCI, Uppsala University, Box 337, SE-751 05 Uppsala, SwedencHelmholtz Centre for Environmental Research–UFZ, OEKUS-Division of Social Sciences, Permoserstraße 15 04318 Leipzig, Germany

A R T I C L E I N F O A B S T R A C T

Article history:Received 24 September 2007Accepted 1 November 2008Available online 8 January 2009

Keywords:Philosophy of ORDecision-making processGroup decisionsSystems

This paper discusses the practical contribution of operations research (OR) techniques tomodelling decision-making problems with ethical dimensions. Such problems are frequentin the global world: they frequently appear today in sustainability issues, e.g., in conflictsin the triangle of society, economy and environment. We show that the prerequisites forethical problem-modelling are: the definition of moral principles, the evaluation of thedecision context, the participation of stakeholders, the multidisciplinary collection of data,and the understanding of systemic interconnections. Classical OR instruments, mainly usedin logistics and optimisation problems, are not entirely satisfactory for coping with thenew ethical dimensions of sustainability. It is recommended to use and to develop moreadvanced, or combined instruments from the multi-criteria/multi-stakeholder and systemicstreams of OR. It is argued that an important added value of using OR techniques formodelling today ethical issues lies at least as much in the discovery of open questions asin finding closed-form solutions.

© 2009 Elsevier Ltd. All rights reserved.

1. Introduction

In previous papers published in this special issue [1–3],different aspects of promoting ethics in operations research(OR) practice have been developed. [1] is the umbrella in-troduction to all four papers. In [2] it is shown that goodpractice of OR, with the primary objective of quality con-trol regarding the analyst's work already includes ethicalconsiderations. The idea that good practice is necessary, butnot sufficient is developed in [3]. Other dimensions of the

This fourth and last paper in a row published in this issue is a reworkedpart of the results of a working group session in the workshop “PromotingEthics in OR practice”, in April 2003 at INSEAD, Fontainebleau. Participantsof the workshop were: João Clímaco, Iordanis Kavathatzopoulos, PierreKunsch, Marc Le Menestrel, Felix Rauschmayer and Warren Walker.Processed by Editor B. Lev

∗ Corresponding author. Tel./fax: +3226483550.E-mail addresses: pkunsch@vub.ac.be (P.L. Kunsch), Iordanis@it.uu.se

(I. Kavathatzopoulos), Felix.Rauschmayer@ufz.de (F. Rauschmayer).

0305-0483/$ - see front matter © 2009 Elsevier Ltd. All rights reserved.doi:10.1016/j.omega.2008.11.006

ethical process in OR are discussed, evidencing difficultiesand ambiguities in the relationships to be established be-tween the OR practitioners and his/her clients, decision-makers or stakeholders. It shows that neither the analysisand modelling work nor the choice of analytical tools areentirely ethically neutral; incomparability, incommensura-bility and uncertainties must be dealt with, and contributeto the existence of ethical values.

Both articles [2,3] are centred on ethical dimensions to befound in the work and interaction of analysts and decision-makers in solving problems. This article concentrates onif, and how, OR instruments can significantly contribute tosolving ethical problems in modern human societies. Putshortly, `ethics in OR modelling' addressed in [2,3] is com-pleted by the reverse point of view, i.e., `OR modelling forethics', in the present article. This paper primarily discusseshow much, and by which techniques, OR may contribute tosolving ethical challenges of our time. Many of them are lo-cated in the issues of `sustainable development', i.e., accord-ing to [4]: `sustainable development is development that

P.L. Kunsch et al. / Omega 37 (2009) 1100 -- 1108 1101

meets the needs of the present without compromising theability of future generations to meet their own needs'.

The idea that OR techniques can provide a useful con-tribution to important community issues is certainly notnew. The British mathematician Lewis Fry Richardson [5]built the very first mathematical model of conflicts betweennations—the arms race—in the thirties of the last century.This model is comparable to the well-known predator–preymodel of Lotka–Volterra [6], developed at about the sametime with a similar purpose of gaining insight into complexsystems.

The founding fathers of OR during WWII were also verymuch conscious of the social and ethical issues to be ad-dressed by OR techniques and models [7,8]. In the 1970s andearly 1980s in the aftermath of the first oil crisis in 1973many OR papers were produced on energy issues (see [9]with many references on energy planning studies in this pe-riod). Although not yet coming under the label of sustainabledevelopment, this work has to be understood as a desire tocontribute to a crucial issue in modern industrial societies.

Later on, there were significant debates about the socialand ethical role of OR (see for example [10–12]), also as partof the agenda of the Critical Systems Thinking movement(see [13–17]). An important contribution to OR modelling isthe collection of papers in [18], of which several are men-tioned in [2]. A recent paper [19] reviews contributions of ORto ethics, and discusses recent attempts to revive the ethicaldebate within EURO from 2000 on.

With the present article we hope to present a modestcontribution pursuing similar lines of thought of our prede-cessors to better address important societal challenges withquantitative OR techniques.

The article is organised as follows:Section 2 discusses several analyses that are prerequisites

for modelling complex society problemswith OR techniques.These preliminary tasks are made in interaction between ORpractitioners and decision-makers, according to the princi-ple of ethics in modelling detailed in [2,3]. In these steps sev-eral of the complex dimensions of ethical problems shouldbe accounted for: the identification of moral principles; thesocietal context of the decision; the multidisciplinary andmultiple-stakeholders aspects; and the systemic dimensionof the problem.

Section 3 characterises OR techniques that are useful inevaluating decision-making problems, and how they maycontribute in modelling problems with ethical dimensions.

Section 4 gives conclusions.

2. Prerequisites for modelling ethical decision problems

Much information and data must be made available priorto modelling with OR techniques. This is true in general forany problem. In the case of `ethical problems'—several exam-ple of such problems will be provided in our text—additionalanalyses are needed, however.

Most of the time OR analysts are called in to provide assis-tance to decision-makers, or policy-makers, in solving `well-defined problems', generally optimisation problems, like insupply-chain management, transportation logistics, location

problems, etc. The moral values, or opinions, of differentstakeholders are neglected, or of minor importance. Mod-els are mainly, or entirely static, and only loosely, or not atall, connected to other problems, like the availability of rawmaterial, the CO2 emissions, etc. To solve these problemsmathematical technicalities and skills aremainly required. Inethical problems new dimensions are present. Many prob-lems of this kind appear in conflicts between economic, soci-etal, and environmental aspects, as set out in [3]. An exampleof such conflicts would be the decision to extend a local air-port on which low-cost airplane companies would operate.The conflict is between employment, travellers' enjoyment,etc., on one hand, and increased CO2, fuel consumption, noisepollution, etc., on the other hand. In such problems themoralvalues and opinions of decision-makers and stakeholders'are central. In addition many connections exist with otheraspects, which cannot be ignored in the modelling, becausethey may induce consequences on the society as a wholeduring a long time frame.

Thus ethical problems are much harder to address thanlogistical problems, and different skills and techniques areneeded. The main issues are not primarily the technicalitiesor heuristics. The definition of the human context, the identi-fication of stakeholders and their moral values, the systemicanalysis of all connections and entanglement with societyimpose in-depth recurrent analyses. Table 1 gives a sum-mary of these analyses and the sub-sections in which theyare addressed.

2.1. Defining moral principles

Analysts and decision-makers must agree on moral prin-ciples, which lead the decision-making process. First let usdefine what is understood under moral principles. Moraleand ethics are practically synonymous words. Ethics refersto the search of `good' and `fair' attitudes in human conducts,while the science of Morale details the set of principles re-quired for a `moral' behaviour. All the adjectives betweenquotes may actually have different meanings in space andtime as will be further explained in Section 2.2 (see refer-ences in [6]). The discovery process of moral principles ismost of the time quite tedious and difficult. The main reasonis that moral principles in a given context are too generalin character to deduce right courses of action in particularreal life situations. Knowing what is `right' and `wrong' ata general level is not that useful in solving concrete ethicaldilemmas. Even if we knewwhat themorally right principlesactually are, it would not be enough to guide our decisions.Moreover, stable conditions are necessary for the construc-tion of functioning moral principles, even at a general level.This is not the case in the world today, given the high paceof change.

Under such conditions, individual decision-makers, aswell as groups and organisations, must acquire high ethi-cal competence and confidence in handling significant moralproblems that may arise in professional activities. That isnecessary in order to solve moral problems, and to makemoral decisions in accordance with relevant values, princi-ples and interests. Spontaneous subjective reactions tomoral

1102 P.L. Kunsch et al. / Omega 37 (2009) 1100 -- 1108

Table 1The in-depth recurrent tasks to modelling ethical problems with OR techniques.

Step Definition Sub-section in text

Defining moral principles Elaborating moral principles and ethical values in thedecision-making process

Section 2.1

Evaluating the decision context Taking into account the context aspects in time andspace for elaborating the moral principles and ethicalvalues

Section 2.2

Associating stakeholders Dialoguing with all stakeholders in order to learn abouttheir values, commitments and interests in the decision-making process

Section 2.3

Collecting inter-disciplinary knowledge Taking into account all available knowledge, includinguncertainties and ignorance gaps, in all disciplines re-lated to the decision-making process

Section 2.4

Understanding the systemic interconnections Taking into account the systemic aspects in any humanorganisation, or society, and accounting for existing linkswith their environment

Section 2.5

issues may give solutions to problems, which probably sat-isfy one's moral feelings. However, with such more or lessemotional reactions only, the relevant factors of the particu-lar moral problem are certainly not fully taken into account.For example, empathy is not a good strategy to achieve sat-isfying solutions in negotiations compared to the cognitivecapacity of perspective taking [20].

What is needed is a psychological approach to ethicalcompetence implying high ethical awareness, adaptive prob-lem solving and decision-making abilities at personal andorganisational levels, effective ethical argumentation skills,and high ethical confidence. It appears then that to knowhow to handle moral conflicts and to acquire ethical com-petence is what really matters. Certainly making decisionsconcerning technical problems or issues of nature is not easyeither. There is a lot of evidence in psychological researchdescribing people's deficient rationality in handling techni-cal, logical or mathematical problems. However, rational, orautonomous ethical decision-making (in the sense definedin [3]), is difficult, much more difficult than autonomousthinking on technical problems [21] due to the followingconditions [22,23]:

(1) In moral problems it is difficult to agree on the context-dependent and controversial definition of `right' and`wrong'. Persons or groups that benefit from somemoralprinciple often find it morally right, whereas those whodo not benefit from it regard the same principle as be-ing morally wrong. Facing a personal ethical dilemmaimplies that arguments for and against a certain moralprinciple can be concurrently valid in the thoughts ofone and the same individual. This is an illustration ofthe non-universality of ethical values to be discussed inSection 2.2.

(2) Solutions to moral problems may be in conflict withother moral values pertaining to the same situation.Then a decision-maker has to choose one principle overthe other. Moreover, under certain conditions, doublestandards and hypocrisy may be morally necessary.

(3) Real-life moral problems are often accompanied bystrong emotions, not present in technical problems. Ifa decision-maker cannot distance him/her from emo-tions, there is a great risk that his/her decision-making

process may be biased [24]. Distancing from emotionsis not easy since decision-making is associated with theactivity of emotional centres in the brain, as confirmedby NMR (nuclear-magnetic resonance) measurements[25,26]. Furthermore, emotions are necessary in achiev-ing rational decisions and they should not be excludedfrom the decision-making process [27]. A problem-solving process loaded with too many strong feelingswill have difficulties in finding the right emotional bal-ance, and in being objective and rational, however. See[28–31] for a discussion on emotions and multi-criteriaanalysis.

(4) Authority significantly affects ethical problem solvingand decision-making. Obedience to authority impliesnon-rational or heteronomous thinking (this concept isdefined in [3]), and our proneness towards obedienceis so strong that we can even do things that we findmorally wrong [32].

(5) Individuals in groups or in organisations conform easilyto the majority [33], and they adopt more extreme posi-tions when they are in a group together with other like-minded people [34]. Decision-making in authoritarian,insulated, cohesive and stressful groups suspends crit-ical and systematic thinking, the so-called groupthinkphenomenon [35].

(6) The contents of moral statements dominate our moralperception, and inhibit further investigations on thewaythey were established in the first place. Usually peoplewould spontaneously react to the content of a moralstatement by accepting, or rejecting it; but they wouldfind it very difficult to focus on the thought process atthe origin of the proposed moral solution, as would beimplied by discourse ethics [36]. The situation is reversein technical processes in which more focus is given tothe methodological approach to the problem than to thetechnical conclusion.

All the above conditions heighten the complexity ofmoralproblem solving. The way humans solve everyday problemswith ethical content is far from being fully rational. Adopt-ing a purely philosophical rational/utilitarian/calculatingview on the way people are acting when confronted to reallife ethical dilemmas is extremely reductive [26]. The way

P.L. Kunsch et al. / Omega 37 (2009) 1100 -- 1108 1103

people make decisions is as well-impregnated by rules ofthumb [37], as by deontological or virtue ethics.

We will investigate in Section 3 in which way OR tech-niques can ease the evaluation of moral principles.

2.2. Evaluating the decision context

Ethical values, understood as synonymous to moral prin-ciples discussed in Section 2.1, will be very dependent onthe prevailing geographical, ecological, economic, historical,religious, etc. context. This aspect has been addressed in [3]as part of the difficulties in the social relationships betweendecision-maker, practitioner, and the wider society. This isthe reason why it is impossible to define a universal anddefinitive code of ethical values.

Basic problems of modern societies are the exhaustionof resources, the global change (e.g., of climate), and the so-cial and economic inequalities around the planet. Long-termsurvival of mankind seems to be only possible if the ethicalvalues at the base of sustainable development are adopted(see [6] for a more detailed discussion). For example, solvingtruck transportation logistics problems are very importantsustainability issues in both Europe and Africa, but differentpriorities will be given. In Europe, congestion, accident risks,pollution, etc. are on top of the agenda. In Africa, bringingfood or medical care to isolated villages will be the mostprominent issues. The identification of what is important forthe human beings should come first. Finding out what goeswrong in a particular society is a prerequisite to any ethi-cally caring decision process.

2.3. Associating stakeholders

The identification of moral principles discussed in Sec-tions 2.1 and 2.2 clearly requires that all involved parties beconsulted and allowed to participate in a decision-makingprocess with ethical content, e.g., the establishment of ahigh-level nuclear waste repository, a fast-speed train track,a high-security prison, etc.

Participation is understood here as `forums for exchangethat are organised for the purpose of facilitating communi-cation between government, citizens, stakeholders and in-terest groups, and businesses regarding a specific decision orproblem' [38]. There are three main arguments for partici-pation of those affected by a decision in the decision-makingprocess: (1) instrumental reasons, (2) substantive reasonsand (3) ethical-normative reasons [39,40].

(1) The growing demand for participation translates intolegislation and jurisdiction, with the consequence thatstakeholders increasingly have the formal and informalright to block decisions. There are doubts whether tra-ditional decision-making procedures can include thesenew demands for the recognition and inclusion of inter-ests in such a way that decisions can be implementedeasily. Participatory decision-making strengthens theprobability of the successful implementation of adecision.

(2) The substantive argument for participation mainly liesin the ability of participatory processes to deal with in-certitude and ambiguity (see [3] for a thorough discus-sion of incertitude in the decision process) through theinclusion of idiosyncratic or local knowledge. Indeed,two different types of information to improve the qual-ity of decisions can be distinguished. The first is scien-tific or technical knowledge of the processes involvedand of the probabilities of certain outcomes. This infor-mation comes from different scientific disciplines andneeds to be integrated (see Section 2.4); trade-offs andinteractions need to be assessed (see Section 2.5). Thesecond type is location-specific or idiosyncratic knowl-edge, based on the experience users have acquired witha specific situation in a specific location. Again, this typeof knowledge can be held by different groups of people,often stakeholders in the decision, but others as well.Depending on the kind of decision, the relative impor-tance and availability of these two types of informationvary. Modelling is often used to integrate informationfrom natural or social sciences, but it may prove insuf-ficient in integrating knowledge from both groups ofsciences, and, even more, in integrating research- andexperience-based knowledge. Furthermore, in each ofthese areas, the type and amount of risk, incertitude,and ignorance is different [3], which makes integrationeven more difficult. Participation of lay people in thedecision-making process is the only way to link thesetwo types of information.

(3) Ethical-normative reasons for participation focus onpopular sovereignty, equity and political equality. Di-rect participation is considered a necessary conditionfor good decision-making in discourse ethics [36]. Therecognition of interests of those concerned by thedecision is necessary for all modern ethics [41].

A high degree of transparency seems to be the mostpromising measure to achieve effective and legitimatedecision-making procedures [42]. One effective way toachieve transparency is to include stakeholders or the gen-eral public in the decision process [43]. Using a structuredprocess that is understandable to both participants and non-participants can further augment transparency. But it is noteasy to conceive who will be concerned by a decision andwho should therefore participate in the process. Globalisa-tion and the inter-connectedness of modern societies leadto a higher degree of complexity of the impacts of decisions.More and more, decisions have impacts on people living inother societies, and on future generations. Ethically, there ishardly any reason why one should consider their interestsless than those of present people living in the same society.Participants in the decision-making process need to takethese `distant' interests of future generations into account.

Another argument for dialogue and participation withan instrumental, substantive, and normative aspect is thebuilding-up of competence. Those participating in a deci-sion will learn the others' points of view, restrictions on thespace of alternatives, dependencies within and between thedifferent systems, and therefore build up competence. This

1104 P.L. Kunsch et al. / Omega 37 (2009) 1100 -- 1108

gain can be useful for the implementation of the decision,for its acceptance, and for seeking new alternatives openedup by the combination of the new competences of the par-ticipating persons and the `old' knowledge.

Participation can be undertaken in different forms withdifferent foci, e.g., focus groups, citizens juries, consensusconferences, cooperative discourse, dialogue groups, stake-holder workshops, participatory expert workshops, reflec-tion forums, deliberative interviews, voluntary agreements,eco-audits, policy simulation exercises, deliberative fore-sights, concerted environmental management, mediation,regulatory negotiation, consultative forums, deliberativeconflict resolution processes, environmental negotiations,etc. (see [38,43–45]). No special form can be recommendedwithout consideration of the specific elements of the deci-sion situation. [46,47] discuss the use of OR instruments inthe resolution of environmental conflicts (see also [3]).

2.4. Collecting inter-disciplinary knowledge

Because any ethical decision-making process involvesmultiple stakeholders, multiple aspects, and also complexsystemic interconnections, as will be discussed in Section2.5, it will draw knowledge from different disciplines, orscientific fields, relevant for the decision problem [48].As an example, a nuclear plant allocation problem mightuse information coming from economics, sociology, socio-psychology, political sciences, hydrology, geology, etc. Manytimes decision criteria in the decision problem will haveincommensurable units—like for example generating elec-tricity costs, expressed in EUR/kWh, and CO2-emissions,expressed in tons/kWh. These criteria will therefore beevaluated within different disciplines, in this exampleeconomics, and power plant technology. Translating CO2-emissions into monetary values, using, e.g., prices of carbonemission trading systems as done in economic cost-benefitanalyses, biases the view on the problem, and may leadto bad solutions, or deviances from the pursued reductionobjectives. The problematic of incommensurability betweencriteria is discussed in [3]. This example makes clear thatmulti-disciplinary knowledge cannot be conceived of as amere concatenation of information from several disciplines.Symbiosis of several disciplines in one model can aug-ment clarity, and reduce model incommensurability. Groupmodel building techniques [49,50], and problem structuringmethods [51], may be used for easing the preparation of anew interdisciplinary model. Also such models should bedesigned in a modular way so that the incorporation of anew corpus of knowledge can be made easy when suitable.

2.5. Understanding the systemic interconnections

Important sources of complexity related to the ambi-guities of defining ethical values have been discussed inSections 2.1–2.4. An additional complexity layer to be dis-cussed in this sub-section originates from the many systemsinterconnections relevant for problem solving. Systematicanalysis is thus another requirement particularly importantin a global economy and environment. Moreover, systems

relevant for economic decision-making affecting the envi-ronment [3] are characterised by incertitude or ignorancein the far-remote future, often covering several generations.This may induce unexpected long-term detrimental conse-quences of some apparently attractive current decisions. TheGerman philosopher Hans Jonas has discussed the responsi-bility of present generation in future negative consequencesof technological decisions in his celebrated book `The Imper-ative of Responsibility' [52]. This way of thinking lies at thebasis of the precautionary principle for anticipating long-term effects unforeseen at the current decision time. TheRio declaration on environment and development [53] givesthe following precautionary recommendation: `Where thereare threats of serious or irreversible environmental damage,lack of full scientific certainty shall not be used as a reasonfor postponing cost effective measures to prevent environ-mental degradation'.

Regarding systemic analysis, it cannot be ignored thathuman societies and organisations, from families to multi-national companies and states, have common structural anddynamic properties with living systems in general, despitethe human freedom. Recent research has identified the ex-istence of global properties of living systems, which can beclassified under the heading `emergence' (see e.g., [54–56]).Emergence means that the global behaviour of a systemconstituted by many individuals, called `agents', is not de-ducible from the mere addition of individual behaviours. In-fluencing emergent properties by managing is particularlydifficult, and thus it requires specific skills. An importantemergent property is resilience, i.e., the resistance to anychange due the existing non-linear influences between theagents in the system. This has the consequence that thesystem gets trapped in some stable states, called attractors,with more or less desirable properties. Moving the systemto more favourable attractors requires considerable energyand repeated, long-lasting efforts. Every company manager,or politician, is aware of the difficulties of defeating well-established routines, changing consumer's habits in orderto protect the environment, promoting `responsible' drivingattitudes, etc.

Managing complex living systems with emergent prop-erties, and long-term consequences of present actions, thusproves to bemuchmore difficult than controlling engineeredsystems. It requires a lot of understanding of the way com-plex systems react to external influences. Human liberty andpsychological dimensions—stressed in Section 2.1—make thechallenge for managers even harder. In [57] it is argued thatgood managers best cope with resilience in the system theymanage by giving adequate considerations to the psycho-logical attitudes of the system agents. Often the managers'intuition is not enough to be successful, however. Complexsystems behave in counter-intuitive ways, because of non-linear causal relationships and feedback loops at the originof emergent properties [58].

Systemic modelling can considerably strengthen andimprove the understanding of complex-system dynamicbehaviour for managing them better, and for anticipatinglong-term consequences. This point will be further discussedin Section 3.

P.L. Kunsch et al. / Omega 37 (2009) 1100 -- 1108 1105

3. Modelling ethical problems with OR techniques

Section 2 has evidenced various prerequisites for mod-elling ethical problems. We have insisted on the multidi-mensionality aspects related to multiple decision-makersand stakeholders with different ethical values, multiple dis-ciplines, and multiple evaluation criteria. Also the systemicinterconnections within and between complex human sys-tems were pinpointed. These considerations make clear thatOR techniques used exclusively for well-defined, static andscalar optimisation are not really matching the new needs.It is why more adapted techniques must be used, developed,or revived, when necessary, separately, or in combination.

Let us consider multi-criteria decision analysis to addressthe multidimensionality issues discussed in Section 2. Thistechnique, and its further developments to group modelbuilding, group decision support systems, negotiation tools[47,49,50], etc., lies somewhat outside the main optimis-ing OR streams [59,60]. One reason is that these techniqueshave weak mathematical foundations, and they address ill-defined vector-optimisation problems.

As an example for a systemic approach of problems con-sider system dynamics (SD) [61,62]. This technique is stillnot very popular among OR researchers. Perhaps the con-troversial Club of Rome report [63], issued in 1972, justbefore the first oil crisis, has some responsibility. The ap-proximations it may have contained regarding its data andmodelling equations do not condemn the approach as awhole, however. Forrester [64] gives a strong advocacy ofthe usefulness of SD. He claims that the origin of many, if notall, `problems' in any human organisation, or society, mustbe sought for in the internal systemic structure, in whichfeedback loops play a major role. This corresponds to what isknown today about emergent living systems [58], discussedin Section 2.5. Unfortunately, systems and network thinkinghave been, and still are in our opinion, too much ignoredby analysts and decision-makers altogether. This may havecaused or aggravated the bad consequences of many deci-sions in the past [6,65,66] because many complex intercon-nections in time and space were ignored, in opposition tothe requirements exposed in Section 2.5.

It is why we strongly feel that systemic disciplines likeSD [6,61,62], soft systemmodelling [67], non-linear dynamicbiological systems [68], agent-based modelling (ABM) [69],evolutionary computing algorithms [70], small world the-ory [71], adaptive policies [72], etc., should be promoted formodelling the difficult problems with ethical dimensions.

Let us consider in particular `ABM', extensively developedat the Santa Fe Institute [73] as a relatively new, but im-portant, systemic technique. ABM is based on the assump-tion that the behaviour of interacting agents in face of someevents or decisions can be simulated by a set of rules (seethe definition of agents in Section 2.5). As underlined in[73], ABM is different from the common approaches in sci-ence, which are based on induction or deduction. Ratherthe purpose of ABM modelling is to analyse by simulationlargely unpredictable and counterintuitive behaviours in or-der to observe the emergence of collective behavioural pat-terns, and to analyse how they can possibly be influenced or

changed (see the definition of the concept of emergence inSection 2.5).

As an example for ABM use let us consider the urban con-gestion problem developed in [74]. This is a crucial sustain-ability issue in large cities, and it may therefore be labelledas `ethical problem'. It gives an illustration of the ubiquitous`tragedy of the commons' observed in conflicting and non-co-operative societies [3,6,75]. This systemic mechanism isoften held responsible for the collapse of many ancient soci-eties (Mayas, Eastern-Island Civilisation, etc.) and for manyexisting or pending unsustainable evolution in modern so-cieties [6,66].

The purpose of ABM simulations, in such examples of sus-tainability analysis, is not forecasting as in econometrics, oroptimising as in classical OR. Rather simulations are usefulfor a better understanding of how car drivers collectively be-have, and how things could change, to improve the presentsituation. At the initial time in the simulation the car-drivingcommunity is deeply trapped in an unfavourable system at-tractor, i.e., no or unsatisfactory alternatives to car drivingare available in the urban environment. Understanding thesystemic implications of these conditions is a first step todesigning long-term solutions, as advocated in Section 2.5.

This gaining of insight into complex systemic structuresis the main incentive for using ABM. In the same line ofthought Danielson [76] proposes ABM as a powerful toolfor achieving insight into ethical values and behaviours (seeSection 2.1). The `Center for Applied Ethics' of the BritishColumbia University (Canada), aims at enlarging the scopeof traditional game theory with ABM. It compares the be-haviours of `rational agents', with those of `ethical agents'acting in a co-operative way. The hope is to understand therules to be developed for guiding peoples' behaviours withthe objective of improving society. Quoting [76]: `The ob-jective is not to discover eternal truths of human nature,but to construct mechanisms (some new, some familiar) tosupport mutually beneficial and fair ways of interacting '.Axelrod [73] and many other authors [77] also insist onmechanisms of co-operative behaviours, proven to be essen-tial in the survival of living systems in general, of which hu-man organisations, or societies, are particular cases. The ideais first to find out why there are problems in those systems,and how they might get cured.

Another interesting approach for combining methodolo-gies for solving complex sustainability problems has beenproposed in [78–80]. The methodological framework iscalled adaptive control methodology (ACM). ACM combinesSD, multi-criteria decision aid (MCDA), and control tech-niques. The main steps are shown in Fig. 1 (adapted from[80]). The methodology is adaptive in the sense that at eachstep in the flowchart backtracking to any upstream step ispossible.

The first stage (I) of system modelling contains the iden-tification and boundary definition of the analysed problem.It is mainly a development phase producing a mental modelshared by analysts and decision-makers, and providing aconsensus on objectives to be achieved in the long-term.

The second stage (II) starts with the generation ofrepresentative sets of useful and believable policies for

1106 P.L. Kunsch et al. / Omega 37 (2009) 1100 -- 1108

STAGE I

STAGE II

STAGE III

Mental ModelSystem Dynamics

Modelling

Policy MakingMCDA Selection

Policy ImplementationMonitoring and Control

Fig. 1. The adaptive control methodology provides a framework for com-bining system dynamics, multi-criteria decision aid (MCDA), and controltheory for modelling complex systems in the long term (adapted from[80]).

controlling the complex system. A long-term test sets theprecautionary principle in practice. Simulated policies,which awake the suspicion of being unfavourable in thelong-term, do not pass the test, and they are discarded fromthe set of candidate strategies. A `good' strategy is finallyselected by means of MCDA techniques from among theacceptable strategies.

The third stage (III) in Fig. 1 is dedicated to monitor-ing and real-time control of the selected policy. Watchdogvariables are monitored on a real-time basis, during theoften-long periods representative of issues with an ethicalcontent. Regular updating of the policy is foreseen, e.g.,every five years. Watchdogs may signal urgent short-terminterventions. Use is made of the multiple branching op-portunities in the ACM flowchart of Fig. 1 when revisitingpolicies, the model, or both, as appropriate.

The methodological framework insists very much onstage III, i.e., on the obligation for policy-makers to verify ona permanent basis that no deviance, or undesirable systemevolution occurs, with respect to the consensus objectivesdecided in the planning phase of stages I and II (Fig. 1). Thisis of course an important requirement of ethical respon-sibility that things should not just be left going withoutcontrol, after a decision has be made. The ACM frameworkhas been applied to several case studies on sustainableenergy management in [81–85]. It provides an example ofmulti-methodology advocated in [86].

These examples of possible approaches are of course notexhaustive. We plead for continuing effort to update ORtechniques, to develop new ones, or to combine them. Theaim is improving the modelling capabilities when it comesto solving problems with ethical implications, instead of op-timising ethically not so relevant situations.

4. Conclusions

The aspects of `ethics for OR modelling', of good practiceand beyond, the validation and legitimacy of the analysts'work, the ethical ambiguities in the relationship betweenthe analysts, the OR model, the decision-makers, and thestakeholders have been thoroughly discussed in [2,3]. In thepresent paper the authors discuss the reverse issue of `ORmodelling for ethics'. As they have argued today's ethicalproblems get centred on the global problems of mankindencompassed in the broad concept of sustainable devel-opment. It is why the authors are pleading for increasedmodelling efforts in solving sustainability issues, as a ma-jor contribution of OR to ethics. To achieve this ambitiousaim, a number of prerequisites for ethical OR modellinghave been discussed in detail. A central issue is here to beable to apprehend the moral principles to be respected inapproaching ethical problems in complex human systems.These problems have multidisciplinary facets; they involvemultiple decision-makers, stakeholders and criteria, andthey imply complex systemic interconnections in space andtime. The authors argue that less conventional or sometimesinnovative OR tools, should be increasingly used, developed,and/or combined. Many efforts in this direction are certainlyalready made today in excellent OR articles, like [87–89];but this effort must be continued and amplified. Thoughthe new sustainability problems may be from an analyticalpoint of view less attractive, but also often more difficult,than well-defined mainstream OR problems, they must beplaced at the top of the priority agenda. The tools neededfor modelling these oft ill-defined problems are centredon systems thinking, and multi-criteria/multi-stakeholdertechniques.

An important incentive for reviving, developing, or com-bining less conventional OR practices, is that today the an-alytical process of mainstream OR would tend to reduceany complex problem to a structured and solvable form. Insuch approaches ethical concerns are either ignored, or ab-stracted, so that many dimensions are indeed missing inthe analysis. The intention behind the proposed increaseduse of systemic and multi-criteria decision-aiding tools is to

P.L. Kunsch et al. / Omega 37 (2009) 1100 -- 1108 1107

reintroduce into models otherwise hidden and value-loadedmoral principles.

Themain purpose of ORmodelling is, in the authors' opin-ion, to conduct an extensive and exhaustive ethical processwith all decision-makers and stakeholders, placed within amultidisciplinary framework. An essential result of this pro-cess is to discover what was left out, rather than what wascovered with traditional quantitative instruments.

References

[1] Le Menestrel M, Van Wassenhove L. Ethics in operations research andmanagement sciences: a never-ending effort to combine rigor andpassion, Omega, 2009, this issue, doi:10.1016/j.omega.2008.12.009.

[2] Walker WE. Does the best practice of rational style model-basedpolicy analysis already include ethical considerations? Omega, 2009,this issue, doi:10.1016/j.omega.2008.12.006.

[3] Rauschmayer F, Kavathatzopoulos I, Kunsch PL, Le MenestrelM. Why good practice of OR is not enough—ethical challengesfor the OR practitioner, Omega, 2009, this issue, doi:10.1016/j.omega.2008.12.005.

[4] World Commission on Environment and Development (WCED). Ourcommon future. Oxford University Press; 1987. p. 43.

[5] Richardson LF. The collected papers of Lewis Fry Richardson. In:Sutherland I, editor. Quantitative psychology and studies of conflict,vol. 2. Cambridge: Press Syndicate of the University of Cambridge;1993.

[6] Kunsch PL, Theys M, Brans JP. The importance of system thinking inethical and sustainable decision-making. Central European Journal ofOperations Research (CEJOR) 2007;15/3:253–69.

[7] Rosenhead J. Operational research at the crossroads: Cecil Gordon andthe development of post-war OR. Journal of the Operational ResearchSociety 1989;40:3–28.

[8] Horner P. History in the making, OR/MS today, October 2002. p. 30–9.[9] In: Lev B, editor. Energy models and studies, studies in management

science and systems. Amsterdam: North-Holland; 1983.[10] Ackoff R. The social responsibility of operational research. Operational

Research Quarterly 1974;25:361–71.[11] Mingers J. Towards an appropriate social theory for applied systems

thinking: critical theory and soft systems methodology. Journal ofApplied Systems Analysis 1980;7:41–50.

[12] Jackson M. The nature of soft systems thinking: the work ofChurchman, Ackoff, and Checkland. Journal of Applied SystemsAnalysis 1982;9:17–27.

[13] Jackson M. Social systems theory and practice: the need for a criticalapproach. International Journal of General Systems 1985;10:135–51.

[14] Jackson M. The origins and nature of critical systems thinking.Systems Practice 1991;4:131–49.

[15] Mingers J. Subjectivism and soft systems methodology—a critique.Journal of Applied Systems Analysis 1984;11:85–103.

[16] Mingers J. Towards critical pluralism. In: Mingers J, Gill A, editors.Multimethodology: theory and practice of combining managementscience methodologies. Chichester: Wiley; 1997. p. 407–40.

[17] Mingers J. What is it to be critical? Teaching a critical approachto management undergraduates. Management Learning 2000;31:219–37.

[18] Wallace WA. Ethics in modeling. New York: Pergamon; 1994.[19] Brans JP, Gallo G. Ethics in OR/MS: past, present and future. 4OR

2004;2:95–110.[20] Galinsky AD, Maddux W, Gilin D, White JB. Why it pays to

get inside the head of your opponent: the differential effects ofperspective taking and empathy in negotiations. Psychological Science2008;19:378–84.

[21] Sunstein CR. Moral heuristics. Behavioral and Brain Sciences2005;28:531–73.

[22] Kavathatzopoulos I. The use of information and communicationtechnology in the training for ethical competence in business. Journalof Business Ethics 2003;48:43–51.

[23] Kavathatzopoulos I. Making ethical decisions in professional life. In:Montgomery H, Lipshitz R, Brehmer B, editors. How professionalsmake decisions. Mahwah, NJ: Lawrence Erlbaum Associates; 2004.p. 277–88.

[24] Glucksburg S. The influence of strength of drive on functionalfixedness and perceptual recognition. Journal of ExperimentalPsychology 1962;63:36–41.

[25] De Martino B, Kumaran D, Seymour B, Dolan JR. Frames, biases,and rational decision-making in the human brain. Science 2006;313:684–7.

[26] Greene JD, Nystrom LE, Engell AD, Darey JM, Cohen JD. The neuralbases of cognitive conflict and control in moral judgment. Neuron2004;44:389–400.

[27] Koenings M, Tranel D. Irrational economic decision-making afterventromedial prefrontal damage: evidence from the ultimatum game.The Journal of Neuroscience 2007;27:951–6.

[28] WenstBp F. Mindsets, rationality and emotion in multi-criteriadecision analysis. Journal of Multi-Criteria Decision Analysis2005;13:161–72.

[29] Le Menestrel M. A Comment on rationality, ethical values and emotionin MCDA. Journal of Multi-Criteria Decision Analysis 2005;13:179–82.

[30] Rauschmayer F. Linking emotions to needs. A comment to FredWenstBp's article “Mindsets, rationality and emotion in multi-criteria decision analysis”. Journal of Multi-Criteria Decision Analysis2005;13:187–90.

[31] WenstBp F. Some viewpoints on emotion, consequentialism andmulticriteria decision making. Journal of Multi-Criteria DecisionAnalysis 2005;13:191–8.

[32] Milgram S. Obedience to authority. New York: Harper and Row; 1974.[33] Asch SE. Studies of independence and conformity: a minority

of one against a unanimous majority. Psychological Monographs1956;70:416.

[34] Sia CL, Tan BCY, Wei KK. Group polarization and computer mediatedcommunication: effects of communication cues, social presence, andanonymity. Information Systems Research 2002;13:70–90.

[35] Janis L. Groupthink: psychological studies of policy decisions andfiascos. 2nd ed., Houghton Mifflin: Boston; 1982.

[36] Habermas J. Erlauterungen zur Diskursethik (Notes on discourseethics). Frankfurt/M: Suhrkamp; 1991.

[37] Gigerenzer G, Todd PM, Group AR. Simple heuristics that make ussmart. Oxford: University Press; 1999.

[38] In: Renn O, Webler T, Wiedemann P, editors. Fairness and competencein citizen participation, evaluating models for environmentaldiscourse (risk, governance and society). Dordrecht: Kluwer AcademicPublisher; 1995.

[39] Fiorino D. Citizen participation and environmental risk: a survey ofinstitutional mechanisms. Science, Technology and Human Values1990;15:226–43.

[40] Stirling A. Analysis, participation and power: justification and closurein participatory multi-criteria analysis. Land Use Policy, Special issue“Resolving Environmental Conflicts: Combining Participation & Multi-Criteria Analysis” 2006;23(1):95–107.

[41] Webler T, Renn O. A brief primer on participation: philosophy andpractice. In: Renn O, Webler T, Wiedemann P, editors. Fairnessand competence in citizen participation. Dordrecht: Kluwer; 1995.p. 17–34.

[42] O'Neill J. Representing people, representing nature, representingthe world. Environment and planning C—government and policy2001;19:483–500.

[43] Lafferty WM, Meadowcroft J. Democracy and the environment:problems and prospects. Aldershot: E. Elgar; 1996.

[44] van den Hove S. Participatory approaches to environmental policy-making: the European commission climate policy process as a casestudy. Ecological Economics 2000;33:457–72.

[45] de Marchi B, Ravetz J. Participatory approaches to environmentalpolicy, policy research brief N10, EVE concerted action. Cambridge(UK): Cambridge Research for the Environment; 2001.

[46] Wittmer H, Rauschmayer F, Klauer B. How to select instruments forthe resolution of environmental conflicts. Land Use Policy, Specialissue “Resolving Environmental Conflicts: Combining Participation &Multi-Criteria Analysis” (2006);23(1):1–9.

[47] Rauschmayer F, Wittmer H. Evaluating participatory and multi-criteria methods for the resolution of environmental conflicts.Land Use Policy, Special issue “Resolving Environmental Conflicts:Combining Participation & Multi-Criteria Analysis” 2006;23(1):108–22.

[48] Muller-Merbach H. The interdisciplinary generalist. Omega 2009;37/3:495–6.

[49] Rouwette E, Vennix J, Mullekom T. Group model buildingeffectiveness: a review of assessment studies. System DynamicsReview 2002;18/1:5–45.

[50] van den Belt M. Mediated modeling. A system dynamics approach toenvironmental consensus building. Covelo: Island Press; 2004.

[51] In: Rosenhead J, Mingers J, editors. Rational analysis for a problematicworld revisited: problem structuring methods for complexity,uncertainty and conflict. 2nd ed., Chichester, England: Wiley; 2001.

1108 P.L. Kunsch et al. / Omega 37 (2009) 1100 -- 1108

[52] Jonas H. The imperative of responsibility. In search of an ethics forthe technological age. Chicago: University of Chicago Press; 1984.

[53] Rio declaration on environment and development, made at UNCED1992, ISBN 9 21 100509 4, 〈http://www.unep.org/Documents/Default.asp?DocumentID = 78&ArticleID = 1163〉.

[54] Holland JH. Emergence—from chaos to order. Cambridge, MA: ThePerseus Books Group; 1999.

[55] Kauffman S. At home in the universe—the search for laws of self-organization and complexity. Oxford, UK: Oxford University Press;1995.

[56] Prigogine, Stengers I. La nouvelle alliance (The new alliance). Paris:Folio Essais, Gallimard; 1986.

[57] Theys M, Kunsch PL. New Paradigms for decision-making in complexliving network systems. In: EURO XII conference, Prague, CzechRepublic, 2007.

[58] Thomas R, Kaufman M. Multistationarity, the basis of celldifferentiation and memory. Chaos Review 2001;11/1:170–95.

[59] Roy B. Méthodologie multicritère d'aide à la decision. Paris:Economica; 2000.

[60] Bouyssou D, Marchant T, Pirlot M, Perny P, Tsoukias A, Vincke Ph.Evaluation and decision models: a critical perspective. Dordrecht:Kluwer; 2000.

[61] Randers J. Guidelines for model conceptualisation. In: Randers J,editor. Elements of the system dynamics method. Cambridge, MA:Productivity Press; 1980. p. 117–38 [chapter 6].

[62] Sterman JD. Business dynamics, systems thinking and modeling fora complex world. Boston: McGraw-Hill; 2000.

[63] Meadows DH, Meadows DL, Randers J, Behrens III WW. The limitsto growth (Club of Rome report). New York: Universe Books; 1972.

[64] Forrester JW. Policies, decisions and information sources formodeling. In: Morecroft JDW, Sterman JD, editors. European Journalof Operational Research, Special Issue “Modelling for Learning”1992;59:42–63.

[65] Dorner D. Die Logik des Misslingens. Strategisches Denken inkomplexen Situationen. (The logic of failure—strategic thinking incomplex situations). Reinbek bei Hamburg: Rowohlt TaschenbuchVerlag; 2004.

[66] Diamond J. Collapse. How societies choose to fail or to survive.London: Penguin Books; 2005.

[67] Checkland P. Systems thinking, systems practice. Chichester: Wiley;1993.

[68] Murray JD. Mathematical biology. 2nd ed., New York: Springer; 1993.[69] Ferber J. Multiagents systems: an introduction to distributed artificial

intelligence. Reading, MA: Addison-Wesley; 1999.[70] Goldberg DE. Genetic algorithm in search, optimization and machine

learning. Reading, MA: Addison-Wesley; 1989.[71] Watts DJ, Strogatz SH. Collective dynamics of `small-world' networks.

Nature 1998;39:440–2.[72] Walker WE, Rahman SA, Cave J. Adaptive policies, policy analysis,

and policymaking. European Journal of Operational Research2001;128/2:282–9.

[73] Axelrod R. The complexity of cooperation. Princeton: PrincetonUniversity Press; 1997.

[74] De Smet Y, Springael J, Kunsch PL. Towards statistical multicriteriadecision modelling: a model and its application. Journal ofMulticriteria Decision Aid 2002;11/6:305–13.

[75] Hardin G. The tragedy of the commons. Science 1968;16:1243–8.[76] Danielson D. Modeling ethical mechanisms. Center for Applied Ethics,

University of British Columbia, Canada; 2002 〈www.ethics.ubc.ca〉.[77] Theys M, Kunsch PL. The importance of co-operation for ethical

decision-making with OR. European Journal of Operational Research2004;153:485–8.

[78] Brans JP, Macharis C, Kunsch PL, Chevalier A, Schwaninger M.Combining multicriteria decision aid and system dynamics for thecontrol of socioeconomic processes. An iterative real-time procedure.European Journal of Operational Research 1998;109:428–41.

[79] Brans JP, Kunsch PL, Mareschal B. Management of the future. In:Bouyssou, Jacquet-Lagrèze, Perny, Slowinski, Vanderpooten, Vincke,editors. Aiding decisions with multiple criteria, essays in honorof Bernard Roy. Boston, MA: Kluwer Academic Publishers; 2002.p. 483–502.

[80] Kunsch PL. Adaptive control methodology and some applicationsin environmental modelling. In: Findley PN, editor. Environmentalmodelling new research. Nova Science Publishers; 2008. (4th quarter,still formatting).

[81] Kunsch PL, Springael J, Brans JP. An adaptive multicriteria controlmethodology in sustainable development—case study: a CO2 ecotax,In: Colorni A, Parrucini M, Roy B, editors. Advances in multicriteriadecision aid. Report of the Commission of the European Communities,Joint Research Center, Ispra, Italy: EUR-report 19808, 2001. p. 145–61.

[82] Kunsch PL, Springael J, Brans JP. The zero-emission certificates: anovel CO2-pollution reduction instrument applied to the electricitymarket. European Journal of Operational Research 2004;153/2:386–99.

[83] Kunsch PL, Springael J. A fuzzy methodology for evaluating a marketof tradable CO2-permits. In: Loulou R, Waaub J-Ph, Zaccour G, editors.Energy and Environment. Gerad 25th Anniversary Series. New York:Springer Science+Business Media; 2005. p. 149–73 [chapter 6].

[84] Kunsch PL, Springael J. Simulation with system dynamics and fuzzycontrol of the reduction of CO2-emissions in the residential sector.European Journal of Operational Research 2008;185:1285–99.

[85] Ben Maalla EM, Kunsch PL. Simulation of micro-CHP diffusion bymeans of system dynamics. In: Antunes CH, Gomes A, editors. EnergyPolicy. Special section: operational research models and methods inthe energy sector 2008; 36(7): 2308–19.

[86] In: Mingers J, Gill A, editors. Multimethodology. The theory andpractice of combining management science methodologies. New York:Wiley; 1997.

[87] Govindaraj T. Characterizing performance in socio-technical systems:a modeling framework in the domain of nuclear power. Omega2008;36/1:10–21.

[88] Wiecek MM, Ehrgott M, Fadel G, Figueira JR, editors, Special issueon multiple criteria decision making for engineering. Omega 2008;36(3): 337–504.

[89] Papadopoulos A, Karagiannidis A. Application of the multi-criteriaanalysis method Electre III for the optimisation of decentralisedenergy systems. Omega 2008;36/5:766–76.