ITM5300 Virtual Lecture for Session SevenMulti-attribute Decision Making; Contra

ITM5300 Virtual Lecture for Session SevenMulti-attribute Decision Making; Contract Management MaturityModel; the Socioeconomic Choice of Contracting or CoercionAlbert DiCanzioThis is the last of my lectures for this course. Our topics are decision making, Garrett Ch. 13and the maturity model for contract management. Recalling again that we began the coursewith Garrett’s idea that excellence with integrity is now the most important aspect of business[Garrett, 9], that is, the ethics of integrity trumps speed and other aspects of doing business, wenow end the course by considering the ethics of socioeconomic contexts for procurement andcontracting. Why? Because integrity entails unconditional adherence to ethics, anduncompromising respect for private property. In other words, we end with the societalorganizational choice whose outcome lies between the polarities of contracting and coercion.Lets first take a quick flashback to Chapter 9, where Garrett wrote about contract sourceselection as a problem in multiple attribute decision making [Garrett, 141], which requiresspecial decision analysis techniques. Though in a footnote he refers you to some of his sources,I am well aware that students often do not have time to chase down and look up everyreference in a textbook. Because this issue of decision making is important, in fact central, inprocurement and contract management, more particularly in source selection, I intend to discussdecision making a little more broadly here.Multi-Attribute decision makingFirst lets discuss decision in general.Based on insight gleaned from various sources, I define a decision as an n-way subjectivevaluation of and choice between identified alternatives. Graphically, it resembles a switchingflow-network node (remember ITM 5100) having input(s) and multiple outputs from which thatnode selects only one pairing or association of input to output link for any particular arrivingunit of flowing commodity based on predetermined criterion or criteria of selection.Decisions are aimed at action. Not the decision but the action affects the world of subjectiveexperience. The decision at hand can only be based on a model of the assumed underlyingreality encompassing it. Or, as Sherlock Holmes said to Dr. Watson:1I can discover facts, Watson, but I cannot change them. Unless some entirely newand unexpected ones come to light I do not see what my client can hope for.There may be nothing more that a decision agent can hope for but to have found and correctlyassessed all the relevant facts. But in today’s world of e-commerce, there is something more.Marketing agents are now armed with big data on customers and online analytical processing1Conan Doyle, The Problem of Thor Bridge in The Complete Sherlock Holmes (Garden City NY: Doubleday & Co.,1930), 1056.Copyright © 2007-2015 by Albert G. DiCanzio(OLAP) that let’s them predict customer behavior, estimate customer lifetime value, and designmarketing campaigns targeted to individual customers or prospects.As to that first step, assessing all the relevant facts, an elementary reference explaining how thisis done is Samuel Richmond.2 Hopefully this dated but seminal book is still available tointerested students. The age of a book is independent of its quality.According to a taxonomy of decision elements by Richmond (whose consistent effort tocommunicate with the reader succeeds most of the time, thus placing his exposition ahead ofthe pack) these facts can be controllable or uncontrollable variables, the latter known as statesof nature. Any configuration of the controllable variables is a course of action. The process ofdeciding is the selection of an optimal course of action.3Given Richmonds characterization, I am led to suppose that every rational decision entails anoptimization and, at some level of generality, resembles a linear programming problem in whichan objective function of the courses of action is optimized subject to constraints which includethe states of nature. Any constraint on a state of nature either reflects a law of nature or it doesnot. If it does not, then I conclude there is risk in the statement of the constraint since bydefinition a state of nature is an uncontrollable variable.With this caveat in mind, assume that we have a utility function such as the one we used (orrejected) in a recent assignment where we calculated (or not) a figure of merit to help usselect among competing contract bids. Because a decision outcome can be represented as afunction of two variables (states of nature and courses of action), a two-dimensional matrixwhose elements are the decision outcomes will model the decision. If the value of each outcomecan be calculated from the utility function, then substituting for the outcome its calculated valuetransforms the matrix to a payoff matrix from which the maximum expected utility over thedecision space can be determined by visual inspection (if it is not too large) or by an easilyprogrammed numeric sort such as can be found today in standard software libraries.That entire process reduces to the following steps: decision problem analysis, measurement ofthe variables, determination of payoffs based on value assignments, and construction of themodel. The model then becomes the feedback point for iterations of Galilean scientificmethodology: it generates predictions which are then compared with real-world observations.These, in turn, generate revisions to the model. If there has been a error of estimation for anystate of nature, the process control system built into the Galilean method will ferret it out as aconsequence of comparing the model with the measurements. Though Richmond did not pointout the estimation risk as to a given state of nature, he thus describes a feedback flow loopwhich controls that risk.4A distinction sometimes drawn between operations research and decision theory is whether themain task is to construct a model or to select the optimal course of action. Problems in thedecision sciences often require both disciplines, relying on a model (a decision matrix) and on a2Samuel B. Richmond, Operations Research for Management Decisions (NY: Ronald Press Co., 1968)Ibid., 22.4Ibid. 30. Figure 1-10.5, this is the most important figure in his introductory exposition.3Copyright © 2007-2015 by Albert G. DiCanzioselection (choosing a payoff from the matrix). It is unclear why such a fine distinction is needed,so that here I generally refer to the whole process under the name of decision sciences.It remains to describe the taxonomy of decision-making and associated techniques, done asfollows with reference to Richmond or to Luce-Raiffa as indicated.Decision-making under certaintyFor decision-making under certainty, the states of nature are knowable through the testing ofthe model, and the highest payoff is identifiable in the final matrix.About decision-making under certainty:The bulk of formal theory in economics, psychology, and management sciences canbe classed under this heading. Until quite recently, the mathematical tools used werelargely the calculus to find maxima and minima of functions and the calculus ofvariations to find functions, production schedules, inventory schedules, and so onwhich optimize performance over time dynamic programing, so to speak.Typically, decision-making under certainty reduces to this: Given a set of possibleacts, to choose one (or all) of those which maximize (or minimize) some given index.Very often the heart of the problem is the appropriate choice of the associated index.In many economic contexts profit and loss are suitable indices, but in other contextsno such quantities are readily available. 5To find an alternative with a maximum index is the problem how to define utility. 5Decision-making under certainty includes several well known classes of problems, including theassignment problem of operations research (which of n tasks should be assigned to which of nresources?), the traveling salesman problem (what is the shortest of the n! possible routesconnecting each of n points exactly once?), and linear programming.Linear ProgrammingLinear programming is a popular approach to deciding under certainty that George Dantzigused to minimize the cost of allocating supplies for the Air Force. The general problem is to findan extremum of some objective function subject to a stated set of constraints, expressed asequations or inequalities. The parts of the problem are (1) a set of acts, each of which representsa specific choice of n real numbers, (2) a set of feasibility constraints on the possible acts, and(3) for each act a weighted average of the n numbers, also known as the index or costfunction. The problem is to find a set of acts which minimizes the cost function while meetingthe constraints.6Game theory and linear programming are both derivable from the general theory of convexbodies.7 The convex body that one might risk becoming after eating too many Big Mac burgersor Oreo cookies may illustrate a special case of the concept. In the abstract, it particularizes a56Luce and Raiffa, op. cit., 15.Ibid. 18.Copyright © 2007-2015 by Albert G. DiCanzioconvex set, a point set in a Euclidean space such that whenever two points lie in the set theline segment joining them also lies in the set. The set is bounded if there is a circle about theorigin which includes the whole set if the radius is chosen to be sufficiently large. 8 A convexbody is a convex set which contains its boundary. This description is clear enough, though whilethe authors defined complex set for a general Euclidean space, they might have been moreconsistent by referring to a locus of points equidistant from the origin instead of using the wordcircle which exists only in a Euclidean plane. In any case, their common antecedent — theseparation theorem of convex bodies — explains why problems couched in game-theoretic termscan often be represented in the form of a linear programming problem.The cost function of the linear programming problem is nothing more than the quantityweighted average of costs ci for each unit i times the number of units x, or ci xi .Decision-making under the uncertainty of conflictFor decision-making under conflict, a form of decision-making under uncertainty, states ofnature on both sides of the conflict are under control of a competitive intellect so that nocompetitor knows the course of action which will be selected by the other. However, analytictechniques exist for educated guessing as to what an adversary will do. The method of analysisbelongs to game theory.9 To some extent, contract source selection and the negotiationsincident thereto resemble this model, though in a successful negotiation, each party looks tocreate advantage for the other party, that is, to exchange with the other party a value perceivedby that party as greater than the value of what is exchanged.Decision-making under the uncertainty of riskFor decision-making under risk, another form of uncertainty, the uncertainty may arise frompatterns of randomness or from ignorance as to the states of nature. Some uncertainty exists incontract source selection, so that a hybrid of this and the preceding model seems appropriate.Chernoff and Moses describe a state of nature differently from, though not necessarilyincompatibly with, that characterized by other authors in this literature review. They see a stateof nature as a description of laws of randomness applicable to the uncertainty arising from theundertaken risk. In a game of coin-flipping, the participants do not know the states of naturebut they know the pattern of randomness to the extent that the behavior of the coin has beeninvestigated. Because the randomness is associated with a known probability distribution, this isclearly better than no knowledge of the probabilities at all. When in this situation, it behoovesthe investigator to perform sufficient tests, as permitted by the project parameters (e.g.,schedule and budget), to replace uncertainty from ignorance with uncertainty from a knownpattern of randomness.107Harold W. Kuhn, in his introduction to Theory of Games and Economic Behavior, writes that By the end of thesummer, we [Gale, Kuhn, and Tucker] had established that, mathematically, linear programming and the theory ofzero-sum two-person games are equivalent. See von Neumann and Morgenstern, op. cit., xi.8Ibid. 117.9Richmond, op. cit. 32.10Chernoff and Moses, op. cit., 1Copyright © 2007-2015 by Albert G. DiCanzioLet’s assume that an agent encounters unknown states of nature with known probabilities. Thenthe MEU principle requires an estimation of the expected value of each course of action. Forindividual decision-making under risk, the question becomes: What is it worth to participate in awager of n outcomes worth a1, a2, , an ounces of gold, respectively, whose probability ofoccurrence is p1, p2, , pn, respectively, where each pi is in the closed interval [0,1] and the pisum to unity?11 As explained by the St. Petersburg paradox and the foregoing discussion of it, themaximum fair price for the wager is not the monetary expected value but rather the numberwhose logarithm to the base 2 is equal to that summation in which log2 2i is substituted for theoutcome values ai. The authors criticize even this interpretation; as their criticism was writtenbefore Graysons empirical results were available, the argument is passed over here, as is thediscussion which merely reformulates the axioms of utility we reviewed elsewhere.Decision-making under ignoranceFor decision-making under ignorance, even the probabilities of the states of nature are unknownthough identifiable. Here, the pyschological state of the agent is a dominant decision factor.Having estimated payoffs and entered them into the decision matrix, each row of the matrixrepresenting a course of action, a conservative or pessimistic agent might use the maximinapproach, attributed to Abraham Wald, in which the course of action or row selected is the onewhich maximizes the minimum payoff of the row. A less risk-averse agent might choose the rowwith the maximum payoff, gambling that the associated state of nature, an a priori unknown,will be the one that occurs; he gambles by accepting a lower payoff than that guaranteed byWalds approach. Intermediate in risk is a criterion which has been attributed to LeonidHurwicz, who would have an agent assign himself an optimism coefficient on a scale from zero(completely pessimistic) to one (completely optimistic). He would then replace the payoffs bythe average of the best and worst payoffs of a given row weighted by the coefficient, or by oneminus the coefficient, respectively.12An approach appealing to still another psychological stance is the minimax regret attributed toL. J. Savage. An agent choosing the Savage criterion wants to minimize the regret he would feelif in hindsight it should become apparent that he missed a higher payoff, the difference of thathigher payoff and his actual payoff being the measure of regret. This criterion is operationalizedby substituting the regret for expected payoff in each cell, then choosing the action whichminimizes the maximum regret over the states of nature. The transformation is known as theregret matrix.13In case an agent feels intuitively confident that the various states of nature are equally likely,then a suitable criterion might be the equal likelihood criterion which simply calls for choosingthe course of action for which the row sum is maximum. This criterion is associated withLaPlace as it is seen as a consequence of his principle of insufficient reason denying theexistence of a reason to believe in a difference of likelihood between the states of nature. 1411Luce and Raiffa, op. cit. 20. This paraphrase substitutes ounces of gold for dollars, since there is no apparentreason why the authors should refer to any particular currency.12Richmond, op. cit. 33.13Ibid. 35.14Ibid. 36.Copyright © 2007-2015 by Albert G. DiCanzioAlternatively, the agent might prefer a mixed strategy, choosing between equally likely states ofnature by a coin toss, superimposing another chance process and effectively detaching his willfrom the decision. Richmond notes that all of these approaches are vulnerable to the relativedifference of cell entries not considered in the max or min or random function, or thefailure to differentiate states of nature to an appropriate resolution, as is particularly the casewith the LaPlace criterion. His recommendation is to attempt to modify the position of totalignorance by acquiring some information about the probability of the various states of naturebefore deciding how to decide.A special form of decision-making under ignorance is what is called group decision-making;because of the attention given it, and because of what I consider its dubious status as adiscipline, it warrants a category sui generis.Contract management in any given scenario may entail an element of certainty, of uncertainty,of risk, and of ignorance. Therefore, to model the scenario requires a hybrid model. This is adecision model, not the sort of organizational-structural hierarchy that Garrett refers to as acontract management model, a somewhat overblown term for what it is.How many objectives does contract source selection have? If the selection occurs, then morethan zero, perhaps more than one. Our organization might wish to have a profitablepartnership. It might also wish that the source sees the partnership as profitable and, thoughthese objectives may not be entirely independent, neither are they identical. Decision making inthese circumstances may be both multiobjective and multiattribute.The decision how to select a contract source and the decision how to close a contract are ingeneral multi-objective decisions addressing at least the basic decision rule: 15Given a set of goals and priorities on goals, choose an alternative having theminimum combined deviation from the specified goals.Objectives are desiderata that may be achieved (or not) but toward which the business processor program should be aimed. They also represent criteria against which the value of a givenalternative (or its performance) may be determined. If it is feasible to estimate the degree ofperformance, then the objective is operationally defined. In some cases a suitable technique formultiobjective decision depends on decision rules that are either optimizing or satisficing.According to Changkong, a decision rule that is optimizing orders all available alternatives intoa complete ranking, that is, there exists a best alternative, while one that is satisficing sacrificesoptimality in favor of simplicity and possible substantial savings in time and cost.An attribute is a measurable quantity whose (measured) value reflects the degree ofachievement for a particular objective (to which the attribute is ascribed). 16 Garrett asserts thatsource selection entails multiple such attributes. To associate a set of attributes with anobjective, it is necessary that the attribute be comprehensive and measurable: comprehensivein that its value indicates the degree of attainment of the objective; measurable in that onsome scale a value can be assigned for any given alternative to that attribute. A scale can be:15Vira Chankong and Yacov Y. Haimes, Multiobjective Decision Making: Theory and Methodology (Mineola, NY:Dover Publications, Inc.), 16.16Ibid., 9.Copyright © 2007-2015 by Albert G. DiCanzionominal (a qualitative datum such as type of equipment, geographic location, etc.); ordinal(ranked in some order by relative performance or size; interval (numeric such that the distancebetween two elements is measurable, such as stock prices on different dates); ratio (numericsuch that both distances between elements and ratios of elements have meaning, such as pricesin dollars, or speed of performance in operations per second.)17It is important that the set of attributes for a given decision problem be complete; i.e., that setrepresents all aspects of the decision that are of interest to the decision maker.Now we can quickly visit decision making in the particular case of implementing capitalistethical strategy in free markets. Because freedom requires production, moral behavior must beengineered by the same free markets; i.e., history has shown that any market system that hasassumed and relied on the morality of human contracting partners for its functioning ultimatelyfails.18 What is required is a mechanism by which unethical behavior cuts off participation in thefree market, in much the same way that when X cheats Y in business, Y is unlikely to continuedoing business with X and may spread the word about X, causing others to refrain from doingbusiness with X. Credit rating systems now operate this way with respect to borrowing but, Ihave observed, not with respect to contractual activity that is unethical or unconstitutional. Still,the first attack has the strategic advantage due to surprise, in this case, to no adverse history.What is a general way to counter the advantage of strategic surprise than an attacker has? Ioffer up to ten points extra credit for whoever may work and support what I consider to be thebest answer into his/her Chapter 13 Assignment.Contract Management Maturity Model (CMMM as described by Garrett)One of the benefits of Garrett as a textbook for this course is that it is intensive in contractingdetails and has relatively short chapters. Another benefit is that its flaws make it a good modelwith which to inculcate caution in students. The flip side of this benefit is that your currentinstructor has a duty to comment on the more subtle deficiencies that he finds in your textbook.NOTE: page numbers in Garrett in the range of the following references are 4th ed. references,5th ed. delta -2; e.g., a reference here to p. 215 corresponds to p. 213 in 5th edition.Garrett introduces the so-called maturity model with a dictionary definition of maturity (p.214), but the reader should note that when it comes to defining maturity in the context of thatmodel, he defines it as a measure of effectiveness in any specific process, which does not havemuch to do with the dictionary definition; it is independent of the degree of development of theprocess, since relatively undeveloped processes can turn out to be more effective than relativelydeveloped ones. It is important that you understand this terminological confusion.He claims that the CMMM creates a vision of excellence to help buyers and sellers focus on thekey area of process improvement (214). He does not support this claim with experimentaldata. The reader might better interpret it as an intent rather than a measured result.17See Aczel, Amir D. Complete Business Statistics (Boston: Irwin/McGraw-Hill, 1999), Ch. 1. (This is the textbook Ihave used when teaching Websters BUSN 5760 course in business statistics.)18See Timothy J. Barnett, http://www.truthsavvy.com/content/are-free-markets-really-free-without-moralityCopyright © 2007-2015 by Albert G. DiCanzioThere are some other definitional issues of which you would better be aware. For example, thephrases causally related and effective or superior in the definition of competency are vagueand confusing. Competence is the ability to do what you say that you are going to do, and hasnothing to do with being superior to anything. Furthermore, Garrett failed to include Curtis,Hefley, and Miller, whom he credits with this definition in his Bibliography. The result is thatwe cannot easily ascertain the context for which these authors came up with the definition.As if this were not a weak enough definitional foundation for what follows, one finds processcapability defined as the inherent ability of a process to produce planned results, whichGarrett attributes to Ahern, Clouse, and Turner again omitting the bibliography inclusion.We dont even know the title of this work. If this ability were to inhere in the process, then itwould not be a capability but a reality. As a result it is hard to know what this may mean.On. p. 215, after introducing the set of models reviewed, Garrett wrote: Thus, the majority ofmaturity models reviewed were project management maturity models. Is this the right level?Look at the definitions of program management and project management in the glossary.Which one do you think better fits contract management?Garretts entire concept of maturity depends on the existence of levels or plateaus. But where ishis existence demonstration? If contract management is more or less a continuous progression,then what sense does a level…