Decision Aiding for Managers

The study of the paradigms underlying decisions and their analytical bases encompasses what is known as Decision Aiding. The main aim of this text is to present these paradigms and fundamentals to the reader in a non-technical way. It seeks to satisfy this objective by offering a frame of reference which permits the reader to view a wide range of possibilities concerning the use of Decision Aiding in complex decision making processes, notably in various fields of human experience, in human sciences, applied to society or health, in technology etc., as well as in our personal lives and in our civic activities.
In the next step, some basic questions will be proposed and then answers attempted to give an understanding of what Decision Aiding is.
The first question to be put is the following: What is to be understood by decision? Decision is the process which leads - directly or indirectly – to the choice of at least one from among different alternatives, all of which are candidates to solve a specific problem. In this way, one makes a decision when selecting one candidate from several for a job vacancy. One also makes a decision when classifying these candidates as good, average or not satisfactory. While in the first example there is a direct relationship between the process and the act of choosing itself, in the second example the classification obtained is nothing more than the preamble to the act of choosing – which, however, may or may not materialize.
The second question to be put is the following: Why study decisions? Because (i) the way of thinking implicit in the process which constitutes the decision is part of the daily routine of human beings; and (ii) frequently the practical results of that way of thinking are considered extremely important by them. These two aspects mean that special attention needs to be paid to the decision as the subject of study, justifying it as a field of scientific interest.
On the other hand, although decisions are a constant presence in human activity, at times, people generally considered highly intelligent make decisions which lead to consequences which are very different from those expected. In fact, someone’s performance as a decision maker does not only depend on intelligence: it also depends on the degree of adjustment within the organizational culture – company, political party, family etc. – where the decision is made and the psychological style in decision making. As regards this last aspect, it can be observed that there are, for example, people who prefer to make important decisions alone, after hearing all of the relevant points of view, while others always prefer to work in a group.
A decision is seen, in turn, through the following three dimensions: (i) importance, in terms of satisfaction of values; (ii) the speed required; and (iii) the degree of individuality.
Practically no-one questions the importance of an excellent performance for a decision maker in skills such as leadership, negotiation, entrepreneurship and administration in general. In management, for example, decisions are made because important problems need to be resolved; in this type of decision, decisions are frequently made in groups. In general, but not always, there must be or should be sufficient time to organize the ideas well before the decision itself. This organization of ideas is normally the most important element of the decision making process, called problem structuring.
Next comes the basic question of Decision Aiding: How to make a good decision? It was precisely in the efforts to answer this question that Decision Aiding established itself as a field of scientific knowledge. Even so, it is not difficult to see that a decision which may seem excellent today may tomorrow be revealed as catastrophic. This suggests that the notion of what is a good decision is only valid for a particular scenario, including here the values of the client of the decision, be it a person, a group of people or an organization.
As a result, it can be stated that Decision Aiding arose from the need to support that human activity which consists of making good decisions. There is a consensus among students of Decision Aiding that the path to a good decision usually includes the following stages, not necessarily in sequence:
· Be sure that one is trying to solve the right problem.
· Think sufficiently about the problem, seeking to maintain a distance from possible emotional involvements, never taking as truths the opinions of others and thus avoiding the so-called psychological traps.
· Seek out all the relevant information.
· Clearly identify what effectively matters, in other words, the crux of the decision.
· Consider explicitly the commitments of a moral and ethical nature.
· Generate the widest possible set of viable alternatives.
· List the objectives of the decision making process, both quantitative – to find the solution to the lowest possible annual cost, or minimize the total annual cost – and qualitative – to find the best solution from an aesthetic point of view, for example, or maximize the aesthetic; note that the objectives are always formulated using the infinitive of the verbs.
· For each of the objectives listed, make the decision criteria clear – thus, an objective such as maximize the social importance of the project can be covered in the criteria (i) meeting the most urgent necessities of the population in need and (ii) promotion of social mobility; the criteria are always formulated as nouns.
· Make the consequences of each alternative clear in relation to each of the decision criteria, together with an estimate of the probability of each one of these consequences in fact materializing – the best way to do this is by the construction of a table, in which the lines are associated with the alternatives and the columns correspond to the criteria; the information contained in the intersection of each line with each column will provide the calculations, judgment values, and/or consultations with experts.
· Starting from the nine stages above, although feeding new information as necessary, use one of the various analytical methods available in the literature on Decision Aiding – the multicriteria methods – to select, rank, classify or describe in detail the alternatives on which the decision will be made – the feeding of new information is necessary due to the fact that, during this technical analysis, some aspect of the problem might arise which had not been considered during the previous nine stages, thus generating, for example, new alternatives or new criteria.
· Carry out a criticism of the results obtained in this tenth stage, attempting to place oneself in the position of the decision maker as well as those who will live with the direct and indirect consequences of the decision – on occasions, as a result of this criticism, it may be necessary to redo the tenth stage.
· Produce objective recommendations for the decision maker, here including a proposal of the decision itself and the best way to implement it, guaranteeing the transparent documentation of all the stages, with a view to organizational learning. The perception of the viability of implementing each of the alternative candidates should, in other words, permeate all of the process described above, which may, in many cases, constitute one of the decision criteria.
The first nine of these twelve stages constitute what is usually called problem structuring. The tenth and the eleventh of the stages constitute the decision analysis while the last stage is the synthesis. Throughout this twelve stage process values, alternatives, criteria, consequences, possible risks and trade-offs between alternatives and between criteria are elicited. As a general principle, equal attention should be paid to each of these twelve stages. One of the most common errors – and one which frequently leads to bad results – is to give less attention to the structuring of the problem than to the decision analysis and synthesis. In addition, one should not naively believe that this is a purely rational process, as intuition is always present when performing it. Other difficulties which may arise in carrying this out are related to the presence of different points of view, even among experts; to the usual lack of perfect and complete information; to ever-present uncertainty and imprecision; and, without doubt, to the size of the problem – that is, the range of relationships which may exist among elements of what seems to be the decision problem – which together, are called the decision system – and other components – although outside that system – of the context in which the decision should be made.
Due to their technical importance in the process, the seven – non-sequential, although interactive – phases of the decision analysis must be clearly identified. They are as follows:
· Phase 1 – Identification of the decision agents and the decision maker.
· Phase 2 – Listing of the alternatives, all them being viable candidates to solve the problem in question. In some cases, it will be easy to identify the alternatives; in others, however, it will be necessary to define them progressively. There will also be cases in which it may be necessary to reduce a long list of alternatives to a smaller list simpler to administrate; in principle, this can be done in various ways, such as, for example, eliminating the alternatives which do not satisfy some of the criteria in any way, in this way selecting a basic and representative set of the alternatives or, even determining a relatively small number of critical criteria for the evaluation and selection of those alternatives which perform better according to these criteria. Although using one of these techniques, there is no theoretical limit for the number of alternatives to be evaluated. It is considered that the collection of information for a large number of alternatives may be an exhausting task, especially if the number of criteria is relatively large.
· Phase 3 – Definition of the effectively relevant criteria. The definition of the alternatives and criteria, as seen above, will usually be an interactive process, in which new alternatives can suggest new criteria and vice-versa. Eventually a criteria hierarchy is formed; the criteria hierarchy most often used is linear and takes the form of a tree, in which each criterion is progressively decomposed, starting from the highest branch (or criterion) to those located beneath – through this technique, a family of criteria is formed starting from a father-criterion (the highest branch in the hierarchy). It is observed that there are few formal procedures which help in the structuring of a criteria hierarchy; this is a skill which is acquired with practice. In other words, there is no “correct” hierarchy for any problem in particular, and it is possible to develop alternative criteria structures. However, soon after the construction of a tree (or hierarchy) of criteria, it can be judged whether this representation is useful to the decision analyst, using the five factors suggested by Keeney and Raiffa (1976): (i) Completeness: if the tree is complete, all the criteria which interest the decision maker will be included in it. (ii) Operationality: all of the criteria at the lowest level of the tree should be sufficiently specific to permit the decision analyst to use them in the process of solving the problem (iii) Decomposability: it must be possible to evaluate the performance of an alternative in relation to a criterion, independent of its performance in relation to the other criteria. (iv) Absence of redundancy: if two criteria – partially or totally – reflect the same reality (for example, environmental impact and sound pollution) then one of them is clearly redundant; the danger of redundancy lies in the fact that it may cause double counting with the result that the recommendation generally tends to be spurious. One practical manner of identifying redundancy consists of establishing if it is possible to modify in some way the recommendation reached – by using a multicriteria method - if a specific criterion is eliminated from the tree. If the elimination of the criterion does not alter the choice of the best alternative then it will not be necessary to include it in the analysis. (v) Minimum size: if the tree is relatively large, any decision analysis will be impossible from a practical point of view. To ensure that this does not occur, criteria should not be decomposed beyond the level at which they can be evaluated. Common sense should always prevail. At times, the size of the tree can be reduced through the elimination of criteria which do not permit distinctions to be established between the alternatives.
· Phase 4 – Evaluation of the alternatives in relation to the criteria. There are various ways to carry out this phase, depending on the multicriteria method employed. In this phase, scales will be used to represent the consequences of each alternative in relation to each of the criteria, whether quantitative (such as, for a project, the value of the internal rate of return) or not (such as the importance of the point of view of mitigation of environmental impact).
· Phase 5 – To determine the relative importance of the criteria. This phase of the decision analysis consists of giving the criteria weights. As in Phase 4, there are many ways to carry out this weighting process, depending on the multicriteria method selected. It is important that the measurements of the relative weightings of the criteria are expressions of trade-offs between criteria: for example, one criterion may be considered twice as important as any other, which would bring with it consequences for the calculations to be carried out (by a multicriteria method). These weightings reflect, from the point of view of the decision maker, how much one is prepared to compromise regarding losses in terms of one criterion provided that there is a gain in another criterion, thus providing the idea of a trade-off relationship.
· Phase 6 – Determining satisfactory solutions. These, as seen above, will be the results of a selection procedure (of at least one best alternative for the final choice by the decision maker, or, possibly, a subset of the better alternatives), of a ranking (in which the set of viable alternatives is ranked from best to worst), of a classification (in which the alternatives are classified in pre-established categories) or, simply, of a detailed description of the alternatives (this description, frequently expressed in logical rules, may be used as a preliminary for a selection, a ranking or a classification).
· Phase 7 – Sensitivity Analysis. In this last stage of the decision analysis, as well as playing the role of devil’s advocate, the analyst seeks to introduce realistic modifications (that is, ones which may in fact materialize) in the variables and parameters used by the multicriteria method used, so as to test how robust the results obtained are. Occasionally the analyst may introduce these modifications in order to simulate possible changes in the decision maker’s preferences.
In other words, the preferences are nothing more than binary relationships between two objects – the alternatives. There are four main categories of preferences, which are:
· Indifference – when there are clear and positive reasons which justify equivalence between two alternatives.
· Strong preference (or strict) – when there are clear and positive reasons which justify a significant preference in favor of one of the two alternatives as opposed to the other.
· Weak preference – when there are clear and positive reasons which do not imply a strict preference in favor of one of the two alternatives as opposed to the other, but these reasons are insufficient to deduce if it is a strong preference or indifference between these two alternatives (thus the reasons do not permit one of the two preceding situations – indifference and strict preference - to be isolated as being the only appropriate one).
· Incomparability A – when there are no clear and positive reasons justifying one of the three preceding situations.
These four main categories of preference relationships are present in the carrying out of decision analysis. There are multicriteria methods – the ELECTRE methods – which make intensive use of refinements of them (Roy and Bouyssou, 1993). The twelve stage process is called, as a whole, Decision Aiding. It is fitting here to make a consideration about the terminology. As one makes use of at least two conflicting criteria to resolve all and any decision problems, Decision Aiding is appropriately called Multicriteria Decision Aiding (or Decision Making with Multiple Objectives).
Consequently, it is said that Multicriteria Decision Aiding is Decision Aiding put into practice. This practice, however, although it should be carried out in the best possible way, cannot always guarantee reaching a good decision. It is easy to understand the reason for this: the context – or scenario – in which the decision is made can change with time. Therefore, a new cast of values might arise, making obsolete the initial set of values on which the practice of Multicriteria Decision Aiding was based. In addition to this, new information might arise, as time passes, which may, by means of the introduction of new parameters, invalidate the recommendations which were reached at the end of the initial process.
It can now be stated, recognizing that the decision generally occurs in the presence of a dynamic scenario, that is one which evolves with time, that a good decision is one which solves a problem based on Multicriteria Decision Aiding; as the scenario changes, better decisions, founded on that same base, may materialize.
Therefore, Multicriteria Decision Aiding plays a crucial role, of an extremely technical nature, in the making of the decision concerning complex decision making processes. It illuminates, by means of ample structuring of the problem and analytical focus, through the application of methods, the search for a good solution of the problem. As one is dealing simultaneously with multiple – and conflicting – decision criteria, it can be imagined that this good solution which is sought will meet in different degrees the various objectives which characterize the decision making problem. Thus, based on the ideas of Simon (1982), it is said that one is seeking a satisfactory solution which represents the best possible compromise among multiple decision criteria. According to this last author it is recognized that rationality in decision making is always limited by three main factors, inherent to the participants in making the decision: (i) their cognitive capacities are not infinite; (ii) their personal values and motivations do not always coincide with those of the organization in which they are placed as decision makers; and (iii) their knowledge of the problem which they are attempting to solve is usually partial. In this way, it can be understood why one does not work towards an ideally best possible solution according to the decision criteria, but instead towards at least a satisfactory solution.
At this point it becomes indispensable to define some of the principal participants involved in the practice of Decision Aiding:
· The decision maker – is the last person responsible for the decision to be made, or, simply, the decider; this may be one single person or a group of people, the individual/s for whom the recommendation is produced on which decision should be made.
· The decision agent – is the individual or group or individuals, who, directly or indirectly, carry out calculations, generate estimates and elicit preferences and value judgments which are used during the decision analysis.
· The decision aider (or decision analyst) – is the professional versed in the principles and methods of Decision Aiding to whom is attributed the tasks of administrating and structuring the problem, its analysis and the production of recommendations for the decision maker; it can also be said that the modeling and the solution of the problem are the essential activities of the decision analyst, who constantly interacts with the decision agents and with the decision maker himself/herself. Therefore, the functions performed by the decision maker and the decision analyst are complementary, even though, in the end, the direct responsibility for the decision is assumed by the first and not the second.
It is also said that there are two possible focuses to Multicriteria Decision Aiding: the constructive focus (or constructivist) and the prescriptive focus. According to the constructive focus, the structuring of the problem advances in an interactive way – that is, by means of interaction between the decision analyst (versed in Decision Aiding) and the other participants in the decision making process (in other words, the decision agents) – in a way which is coherent with the values, objectives, consequent criteria and preferences of these agents and of the decision maker himself/herself. The prescriptive focus, meanwhile, consists of starting from a description of all the elements relevant to the problem, including here a description of the preferences of the decision maker, proposing – via the decision analyst – prescriptions to the decision maker, based on normative hypotheses. In the prescriptive focus, the involvement of the actors (or decision agents) in the process is restricted to the structuring of the problem. Due to the greater facility of adapting it to constantly evolving scenarios, the constructive focus to Multicriteria Decision Aiding has increased considerably in relative importance in recent years which, at the same time, has relegated the prescriptive focus to a second plane.
Multicriteria Decision Aiding therefore does not seek an optimum solution to a specific problem, as occurs in traditional Operations Research, but instead a compromise solution, where a consensus must preferentially prevail among the parties involved. From this viewpoint, the criteria used, as well as the importance attributed to them, have an essential role in the results obtained. This type of analysis allows the decision making process to be dealt with in a more transparent way thus increasing its credibility. However, it must be noted that this approach to the decision problem, from the Multicriteria Decision Aiding perspective, does not seek to present the decision maker with a definitive solution to the problem, electing a single truth represented by the selected alternative. This approach instead seeks to support the decision making process with a recommendation for actions which are in line with the preferences expressed by the many decision agents.
Therefore, a multicriteria approach applied to a complex decision making process generally results in the following advantages: (i) the construction of a base for the dialogue between the different decision agents; (ii) a concrete possibility of working with the subjectivities, uncertainties and imprecision always present in such a process; (iii) a visualization of each potential satisfactory solution as a compromise between the distinct points of view in conflict.
While Expected Utility Theory (VON NEUMANN and MORGENSTERN, 1944) reflects a normative vision of the decision, Prospect Theory (KAHNEMAN and TVERSKY, 1979) describes how decisions are made in the presence of risk. As all and any decision implies running some type of risk – at least the risk of not meeting, at a minimally adequate level, the objectives of the problem -, it can be said that both of the theories are decisional paradigms in competition with one another: while the first establishes the norm according to which the rational being seeks, when deciding, to maximize a measure of utility expected by him/her, the second, based on empirical observations, hopes that the rationality of the decision maker, confronted by risk, will reflect on the relative gains and losses, always defined in relation to a point of reference. Prospect Theory provides indeed a wider framework than Expected Utility Theory, based on a description of how to make decisions effectively in the face of risk. However, a decision always occurs in the presence of some type of uncertainty and, therefore, risk. When studying decisions one cannot therefore set out to ignore human behavior in the face of risk. Thus, it is natural that a decision paradigm such as Prospect Theory has established itself as a base for Multicriteria Decision Aiding. At least one multicriteria method exists which has in its base Prospect Theory: the TODIM method (Gomes and Rangel, 2007).

Bibliographical references

GOMES, L.F.A.M.; RANGEL, L.A.D. (2007) “An application of the TODIM method to the multicriteria rental evaluation of residential properties”, doi:10.1016/j.ejor.2007.10.046, paper available through www.sciencedirect.com. To be published in European Journal of Operational Research.
KAHNEMAN, D., TVERSKY, A. (1979) “Prospect Aiding: An Analysis of Decision Under Risk”. Econometrica, vol. 47, p.263-292.
KEENEY, R.L.; RAIFFA, H. (1976) Decisions with Multiple Objectives: Preferences and Value Tradeoffs. New York: Wiley.
ROY, B.; BOUYSSOU, D. (1993) Aide Multicritère à la Décision: Methods et Cas. Paris: Economica.
SIMON, H. (1982) Models of Bounded Rationality. 3 volumes. Cambridge: The MIT Press.
VON NEUMANN, J.; MORGENSTERN, O. (1944) Aiding of Games and Economic Behavior. 3rd ed. 1953 edition. Princeton: Princeton University Press.