IFAC PROCEEDINGS SERIES Editor-in-Chief JANOS GERTLER, Department of Computer and Electrical Engineering, George Mason University, Fairfax, Virginia, USA GERTLER 8c KEVICZKY {General Editors): A Bridge Between Control Science 8c Technology (Ninth Triennial World Congress, in 6 volumes) Analysis and Synthesis of Control Systems {1985, No. 1) Identification, Adaptive and Stochastic Control (1985, No. 2) Large-scale Systems, Decision-making, Mathematics of Control (1985, No. 3) Process Industries, Power Systems (1985, No. 4) Manufacturing, Man-Machine Systems, Computers, Components, Traffic Control, Space Applications (1985, No. 5) Biomedical Applications, Water Resources, Environment, Energy Systems, Development, Social Effects, SWIIS, Education (1985, No. 6) BARKER 8c YOUNG: Identification and System Parameter Estimation (1985) (1985, No. 7) NORRIE 8c TURNER: Automation for Mineral Resource Development (1986, No. 1) CHRETIEN: Automatic Control in Space (1986, No. 2) DA CUNHA: Planning and Operation of Electric Energy Systems (1986, No. 3) VALADARES TAVARES 8c EVARISTO DA SILVA: Systems Analysis Applied to Water and Related Land Resources (1986, No. 4) LARSEN 8c HANSEN: Computer Aided Design in Control and Engineering Systems (1986, No. 5) PAUL: Digital Computer Applications to Process Control (1986, No. 6) YANG JIACHI: Control Science 8c Technology for Development (1986, No. 7) MANCINI, JOHANNSEN 8c MARTENSSON: Analysis, Design and Evaluation of Man-Machine Systems (1986, No. 8) BASANEZ, FERRATE & SARIDIS: Robot Control "Syroco '85" (1986, No. 9) JOHNSON: Modelling and Control of Biotechnological Processes (1986, No. 10) TAL': Information Control Problems in Manufacturing Technology (1987, No. 1) SINHA 8c TELKSNYS: Stochastic Control (1987, No. 2) RAUCH: Control of Distributed Parameter Systems (1987, No. 3) HAASE: Software for Computer Control (1987, No. 4) MARTOS, PAU 8c ZIERMANN: Modelling and Control of National Economies (1987, No. 5) GENSER: Control in Transportation Systems (1987, No. 6) KUZUGU & TUNALI: Microcomputer Application in Process Control (1987, No. 7) WANG PINGYANG: Power Systems and Power Control (1987, No. 8) Β ALCHEN: Automation in Aquaculture (1987, No. 9) YOSHITANI: Automation in Mining, Mineral and Metal Processing (1987, No. 10) GEERING 8c MANSOUR: Large Scale Systems; Theory and Applications (1987, No. 11) ROOS: Economics and Artificial Intelligence (1987, No. 12) TROCH, KOPACEK 8c BREITENECKER: Simulation of Control Systems (1987, No. 13) KAYA 8c WILLIAMS: Instrumentation and Automation in the Paper, Rubber, Plastic and Polymerization Industries (1987, No. 14) NOTICE TO READERS If your library is not already a standing/continuation order customer or subscriber to this series, may we recommend that you place a standing/ continuation or subscription order to receive immediately upon publication all new volumes. Should you find that these volumes no longer serve your needs your order can be cancelled at any time without notice. Copies of all previously published volumes are available. A fully descriptive catalogue will be gladly sent on request. ROBERT MAXWELL Publisher CONTROL IN TRANSPORTATION SYSTEMS (1986) Proceedings of the 5th IFACIIFIPIIFORS Conference Vienna, Austria, 8—11 July 1986 Edited by R. GENSER Bundesministerium für öffentliche Wirtschaft und Verkehr, Vienna, Austria M. M. ETSCHMAIER University of Lotoell, Lowell, MA, USA T. HASEGAWA Kyoto University, Kyoto, Japan and H. STROBEL Hochschule für Verkehrswesen Friedrich List, Dresden, GDR Published for the INTERNATIONAL FEDERATION OF AUTOMATIC CONTROL by PERGAMON PRESS OXFORD · NEW YORK · BEIJING · FRANKFURT SÄO PAULO · SYDNEY · TOKYO · TORONTO U.K. Pergamon Press, Headington Hill Hall, Oxford OX3 OBW, England U.S.A. Pergamon Press, Maxwell House, Fairview Park, Elmsford, New York 10523, U.S.A. PEOPLE'S REPUBLIC Pergamon Press, Room 4037, Qianmen Hotel, Beijing, People's Republic of China OF CHINA FEDERAL REPUBLIC Pergamon Press, Hammerweg 6, D-6242 Kronberg, Federal Republic of Germany OF GERMANY BRAZIL Pergamon Editora, Rua Eça de Queiros, 346, CEP 04011, Paraiso, Sâo Paulo, Brazil AUSTRALIA Pergamon Press Australia, P.O. Box 544, Potts Point, N.S.W. 2011, Australia JAPAN Pergamon Press, 8th Floor, Matsuoka Central Building, 1-7-1 Nishishiniuku, Shiniuku-ku Tokyo 160, Japan J CANADA Pergamon Press Canada, Suite No. 271, 253 College Street, Toronto, Ontario, Canada M5T 1R5 Copyright © 1987 IFAC All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means: electronic, electrostatic, magnetic tape, mechanical, photocopying, recording or other- wise, without permission in writing from the copyright holders. First edition 1987 British Library Cataloguing in Publication Data Control in transportation systems 1986: proceedings of the 5th IFAC/IFIP/IFORS conference, Vienna, Austria, 8-11 July 1986—(IFAC Proceedings series; 1987, no. 6) 1. Transportation 2. Automatic control 1. Genser, R. II. International Federation of Automatic Control III. Series 380.5 HE336.A/ ISBN 0-08-033438-5 These proceedings were reproduced by means of the photo-offset process using the manuscripts supplied by the authors of the different papers. The manuscripts have been typed using different typewriters and typefaces. The fay-out, figures and tables of some papers did not agree completely with the standard requirements: consequently the reproduction does not display complete uniformity. To ensure rapid publication this discrepancy could not be changed: nor could the English be checked completely. Therefore, the readers are asked to excuse any deficiencies of this publication which may be due to the above mentioned reasons. The Editors Printed in Great Britain by A. Wheaton &f Co. Ltd., Exeter 5TH IFAC/IFIP/IFORS CONFERENCE ON CONTROL IN TRANSPORTATION SYSTEMS Sponsored by International Federation of Automatic Control (IFAC) Technical Committee for Systems Engineering (SECOM) Cosponsored by IFAC Technical Committees of Applications (APCOM) Biomedical Engineering (BIOMED) Computers (COMPUT) Developing Countries (DECOM) Education (EDCOM) Economic and Management Systems (EMSCOM) Manufacturing Technology (MANTECH) International Federation for Information Processing (IFIP) International Federation for Operational Research Societies (IFORS) Organized by Austrian Center for Productivity and Efficiency (ÖPWZ), Vienna International Programme Committee R. Genser, Austria (Chairman) H. Kopetz, Austria A. Artynov, USSR M. Koshi, Japan M. Cremer, FRG P. M. Larsen, Denmark J. P. Crestin, France P. Martens, Belgium H. J. De Norwall, Belgium Y. Ohno, Japan W. D. Ehrenberger, FRG K. Pierick, FRG H. V. Ellingsen, Norway P. Rabischong, France M. M. Etschmaier, USA L. Skyva, Czechoslovakia P. Faller, Austria K. Spera, Austria R. E. Fenton, USA W. Schaufelberger, Switzerland H. Frey, Switzerland G. Schweizer, FRG A. Hmamed, MA B.J. Sterner, Sweden T. Hasegawa, Japan A. Straszak, Poland J.J. Henry, France H. Strobel, GDR G. Johannsen, FRG D. Tabak, Israel H. Knoflacher, Austria P. Kopacek, Austria National Organizing Committee P. Kopacek (Chairman) J. Hähnel (Secretary) G. Becker R. Genser M. Horvath W. Stejskal Copyright © IFAC Control in Transportation Systems, Vienna, Austria, 1986 PREFACE The fifth of a series of conferences by was also recognized in the interface bet- the international federations on Automatic ween the human and the various machines Control (IFAC), Information Processing and facilities used in transportation. (IFIP), and Operational Research Societies Specifically, it was concluded that users (IFORS) was held on July 8-11, 1986 in and operators of transportation systems Vienna, Austria. 110 delegates from a to- are often not considered adequately. It tal of 23 countries met to discuss a broad was suggested that the sponsor organiza- range of topics. tions of the conference might be in a po- sition to fill the void, as is already In addition to the challenges and opportu- done through the IFAC Workshops on Safe nities posed by the rapid developments in Computer Systems (Safecomp). microelectronics and information techno- logies, the subject of national transporta- The use of robotics in transportation is tion systems planning was given particular seen to be concentrated in the manufacture attention. One of the motivating factors of equipment. Only to a limited extent is for emphasizing the subject of national robotics being used in maintenance. A con transportation systems planning is the siderable potential for the application of fact that the General Transportation Sy- robotics is recognized in the operational stems Plan of the host country, Austria, area of a transportation system. The fea- is now in its final phase of development. tures of a robot considered most important The plan has reached a state in which in- in such applications are mobility and the teresting results could be reported in the ability to recognize visual images. form of a case study and of several indi- vidual research reports. Reports on simi- At the round table on the use of Videotex lar endeavors completed or under way in (BTX) for logistics and passenger services other countries were sollicited by the In- it was pointed out that mulitfunctional ternational Program Committee to round out use of infrastructure already available the discussions. Also sollicited were re- might be more beneficial than new systems ports on more limited projects of model- which, even at large expense, would offer ling for management and decision making less access than does the available infra- regarding environmental, social, and eco- structure. Also, it was suggested that, nomic aspects of transportation systems. even if videotex systems are designed in a Survey papers and round table discussions user-friendly manner, infrequent users or were organized to illustrate the potential users with more complicated needs might benefits which transportation community not be able to use the system entirely on can derive from such fields as logistics, their own. Hierarchies of systems might robotics, and telecommunications. be used to maintain the role of the trans- portation professional as an intermediary The view of logistics suggests that for a to the user in these situations. shipper transportation efficiency is secon- dary as an objective to the overall effi- The presentation of a psychologist working ciency of the logistics system. Conver- in the field of highway safety showed the sely, it may not be possible to justify limitations of learning and training. In expensive new transportation systems un- the subsequent discussion it was pointed less the shippers can be motivated to out that electronic equipment already change their logisitcs system so that they available might be used to a greater ex- can take best advantage of the opportuni- tent to make traffic more safe and human. ties it presents. Similarly, a shippers needs are rarely satisfied by just one The emphasis of the series of conferences transportation link or one mode of trans- on control in transportation has been on portation. A mulit-modal approach to the transfer of knowledge about transpor- transportation is therefore called for. tation between researchers and developers, and between users and scientists. This conference points at the great need for Standardization is recognized as an im- further work. portant prerequisite for effective multi- modal transportation. While in the past M.M. Etschmaier standardization efforts were mostly direc- R. Genser ted towards physical aspects of the modal T. Hasegawa interfaces, the current need is seen most- H. Strobel ly in the areas of information processing and informatics. A lack of standardization Editors vi Copyright IQ 1FAC Control in Transportation TUTORIAL Systems, Vienna, Austria, 1986 INTERACTIVE ALTERNATIVE ORIENTED DECISION SUPPORT SYSTEM FOR COMMITTEE DECISION MAKING A. Lewandowski International Institute for Applied Systems Analysis, Laxenburg, Austria Abstract. The concept of Decision Support System is presented in the paper. One of the most impor- tant problems in decision analysis relates to the situation, where the committee (group of decision makers) has to select the best alternative from a given, finite set. In most cases, alternatives are evaluated on the basis of several quality factors. In the paper, the author presents the new approach, based on the principle of aspiration-led decision making. This approach ensures proper structuralization of the decision process and allows proper balance of opinions of the group member. The methodological principles of the approach as well as problems relating to computer implementation and applications are presented in the paper. Keywords. Decision analysis; decision theory; utility function; achievement function; scalarizing function; decision support system; group decision making. INTRODUCTION some discussions abut the understanding of this term and Decision Making is one of the most complicated form its relation to other fields of Operations Research (OR) of the manager's or designers (since engineering design and Management Science (MS), especially to Management can be also interpreted in the decision- making context) Information Systems (MIS). Rather deep analysis of this activity. The following elements of the decision making problem was recently performed by Parker and Ai- Utabi process cause certain difficulties: (1986). The authors reviewed about 350 papers relating to - certain (rather high) knowledge about the problem this subject trying to find the most appropriate being analyzed is necessary, characterization of Decision Support Systems. The follow- - the effects of the decision on the system outcome ing set of characteristics was defined by the authors: and performance must be predicted by the decision The DSS should: maker; this is a rather difficult task, because the quality - assist managers in their decision processes in semi of the decision can be characterized by many perfor- structured tasks, mance criteria, support and enhance rather than replace - certain level of uncertainty occurs during the deci- managerial judgment, sion making process, this uncertainty must be taken into account when making decision, - improve the effectiveness of decision making rather the efficiency, - complex structural and logical relationships exist between the elements of the system being analyzed, deci- - to combine the use of models or analytical tech- sion undertaken, system outcome and behavior, niques with traditional data access and retrieval func- tions, - decision making process is in the fact the informa- tion processing task; therefore the decision maker must - specifically focus on features which make them easy handle large amount of information of different nature to use by non-computer people in interactive mode, and usually of poor quality. - emphasize flexibility and adaptability to accommo- Due to the facts listed above, decision making pro- date changes in the environment and the decision making cess requires certain amount of time and resources; approach of the user. moreover, frequently the decision made is rather far from Following the same source, we should consider the optimal, or even satisfactory. Usually, the complexity of characterization of the decision process formulated by the decision making process does not allow to analyze all Simon. According to his qualification, the decision pro- possible options - only a small subset can be effectively cess consists of the following three steps: analyzed. Because of the lack of a'priori knowledge and - Intelligence: searching the environment for oppor- the presence of uncertainty, it can happen, that a poten- tunities calling for a decision, tially good decisions are rejected on a very early stage of - Design: inventing, developing and analyzing possi- the decision process. ble courses of action, The informational and logical aspect of the decision - Choice: selecting a particular course of action from making process suggests, that the computer with its abil- those available. ity of handling and processing large amounts of informa- tion and analyzing complex logical relations could be a Applying the above characterization it is possible to proper tool to support this activity. conclude, that the MIS have made major contributions to the intelligence phase; MS/OR have been useful mostly for There are many commonly accepted definitions of the the choice phase. The design phase could be one of the decision support system - all these definitions agree, that primary contributions of DSS. It should be pointed out, such system must aid a decision maker in solving unstruc- that in the case of MIS the design stage starts from the tured (or "semistructured"), complex problems. data available from the data processing system and moves The term "decision support system" (DSS) became upwards to the decision maker. The DSS in the contrary, rather popular during several years. There are currently starts with the problem to be solved and and works down 1 2 A. Lewandowski with the computer selecting the appropriate subsets of The main differences between various approaches to data and information stored in the computer, necessary to interactive decision support are concerned with the way find the solution. The other aspects of decision support in which the decision maker specifies hies information and systems was specified by Keen and Scott-Morton (1978). requirements about the problem which could be used for According to their specification, the main area of impact estimation the complete preordering, as well as with the of DSS related to the decision problems, in which ".. assumptions about the preordering, its properties and there is sufficient structure for computer and analytic interpretations. aids to be of value, but where managers judgment is The most strongly established framework is based on essential". The main payoff to organization caused by the value function or utility function maximization. application of DSS is "... the extending the range and Under some very general assumptions the ordering can be capacity of manager's decision process to help them to represented by the utility function u: E -* R1 such that improve their effectiveness". Finally, the main relevance y by maximizing this function over χ €JSÇ, we can select the for managers in decision making process was specified as decision which is most preferable to the decision maker. "creation of supportive tools, under their own control, The only difficulty relates to the identification of the util- which does not attempt to automate the decision ity function - this function is determined nonuniquely, is process, predefine objectives or impose solutions". affected by uncertainty of the decision making environ- The last statement requires special attention - the ment an of the decision maker behavior, and what consti- role of the decision support system is to provide support; tutes the most essential trouble, is non stationary in the human is still the most important element of the deci- time due to the learning and adaptive nature of the sion making process and the final decision is in his hands. decision making process. Recently, some more detail The only purpose of the DSS is to increase the under- analysis of the relevance of this approach was performed standing of the decision problem being solved through - see, for example Fisher (1979), Horsky and Rao (1984). deep analysis of the possible consequences of the decision According to the results of the research presented in this made, as well as simplifying the access to all information paper, the utility function approach can be used in a and knowledge necessary to find satisfactory solution. rather simple, laboratory environment, but can fail in Rather deep discussion of other aspects of the role and more complex situations. The same opinion was justified by principles of DSS and comparison of DSS methodology and Eriandson (1981). other ones in the field of MS and OR can be found in the quoted above book. The second framework for rational decision making is the principle of satisficing decisions. This approach Discussion related to methodological and organiza- was firstly formulated by Simon (1969) and further tional concepts of DSS was also performed by Naylor extended by Wierzbicki (1982, 1983, 1984). Deep formali- (1982), Watson and Hill (1983), Vazsonyi (1982) and Wynne zation of this concept was recently done by Eriandson (1982). Similarly to the definitions of the DSS specified (1981). above, the conclusion which follows from these papers are the following: The basic methodological assumptions of this princi- ple are as follows: - DSS, unlike OR/MS, does not replace decision mak- ers, but supports them; hence DSS is a mind supporting - the decision maker evaluates possible decisions on system, the basis of set (or vector) of performance factors. These factors can be expressed in numerical scale - DSS allows for the introduction of judgment, while (quantitatively) or in verbal scale (qualitatively), like OR/MS are normative/prescriptive, "bad", "good" or "excellent", - DSS deals with unstructured problems, while OR/MS - for each performance factor the decision maker applies mostly to structured problems, specifies his aspiration levels; this is a value of the fac- - OR/MS comes up with solutions and recommenda- tor, which can be accepted as reasonable or satisfactory. tions, while DSS does not. In more complex situations the decision maker can specify Very deep analysis of various organizational and two levels for each performance factor - aspiration level methodological aspects of DSS was recently published by as defined above, and reservation level. The last one Sage (1981). The analysis of the concept of DSS from the describes the lowest acceptable level for a given perfor- point of view of various disciplines - including psychology, mance factor. organizational behavior and design, information science, On the basis of the information specified by the deci- management science and computer science. The basis for sion maker (i.e. set of performance factors, aspiration investigation is the critical revue of over 400 up to date and preservation levels) the satisficing decision principle publications relating to decision support systems. can be formulated as follows: Case 1: the decision maker overestimated the system possibilities - there exists no decision ensuring values of THE SATISFYING DECISION PRINCIPLE all performance factors exactly equal to the aspiration BASIC ASSUMPTIONS levels. In this case it is possible to propose the decision, for which values of performance factors are as close to the aspiration levels, as possible. The most important problems related to the develop- ment of a decision support system are the assumptions Case Z\ the decision maker underestimated the sys- about the mathematical structure of the class of problems tem possibilities - there exists the decision which ensures to which the DSS will apply. Such structure usually con- the values of performance factors exactly equal to the sists of the following elements (see Wierzbicki, 1984): specified aspiration levels. In this case it is possible to propose the decision, which improves all the components - a space of decisions (alternatives, options, con- of performance vector as much as possible. trols, designs etc.) denoted by E, x If the decision specified as the outcome of the pro- - a constraint set X c E, 0 x cess is not satisfactory, the aspiration levels are - space of outcomes (attributes, objectives, perfor- corrected and the whole process repeats. The technical mance indices, etc.) denoted by Eyt question arises - how to provide a proper formalization of - an outcome mapping / : E -» E, which also defines the specified above methodological concept. This can be x y the set of attainable outcomes Y = / (x ) c E done in the following way: let the decision making be 0 0 y guided by aspiration levels, and let these levels be speci- - a partial preordering that is usually implied by the fied by the decision maker. The aspiration level w will be decision problem and usually has some obvious interpreta- used as the parameter for certain special utility function tion, like maximization of profit etc. The standard assump- called achievement function. This function is of the form tion is that this preordering is transitive and can be u(y) = s(y,w). This approach provides an explicit and expressed by a positive cone D c Ε. χ constructive means of expressing the dependence on the - a complete preordering, which is usually not given decision environment and proper handling of the nonsta- in any precise mathematical form, but is specified by the tionarity of the decision's maker's preferences. decision maker. Committee Decision Making 3 The theory of achievement function was developed linear optimization problem, when by suitable transforma- and further extended by Wierzbicki (see Wierzbicki, 1986 tion of variables it can be reduced to the single objective for references and detail discussion of the problem). linear optimization problem with additional constraints (see Lewandowski and Grauer, 1982, 1984). Other forms of An achievement function is a function s: E -* R1, y achievement functions together with detailed discussion of with argument y-w, where y eE is an attainable y their properties can be find in the mentioned above paper objective and w is the aspiration level. An achievement by Wierzbicki (1986). function should satisfy the following requirements: Practical experiments with this approach (see, for a. It should be strictly order preserving with respect to example, Lewandowski at all., 1985) have shown, that the y: language of aspirations coincides very well with the deci- y-ytintD -» s (y - w) < s {y - w) sion maker's style of thinking. The information which is 2 1 ± 2 required from him is easy to express, in the contrary to or, if possible, strongly order preserving other approaches based on pairvise comparisons, utility theory, explicit weighting factors etc. V2"Vi ^D/{D η -D) ·* s(y-w) < s(y~w) x 2 b. It should be order representing: \y ε E: s(y - w) & 0 1= w +D ALTERNATIVE BASED COMMITTEE DECISION y SUPPORT SYSTEM or, at least, order approximating for small ε: w + D c ( y e E: s {y - w) j cw +D Due to certain fuziness of the concept of decision y € support system, it is rather difficult to provide a more where detailed taxonomy of the various concepts of DSS organi- D= \y <=.E: dist(y, 2» < ε \\ y ]|) zation. Moreover, such a taxonomy could be performed c y from various points of view - assumptions according to the c. If w € E - D, then the maximization of s(y-w) over decision maker's behavior, formulation of the decision y y e E should represent a concept of either allocation or problem, methods of handling the uncertainty etc. The y maximization of the surplus y — w Ε D; if w j£E -D, most extensive comparison of various methods and y then the same maximization should represent the concept approaches can be found in book by Bonczek (1981). of distance minimization between w and the set The most important criterion for distinguishing \y ZE:E π {y -hD) = φ J between various approaches for DSS will be, for the pur- y y pose of this paper, the method of generating alterna- The most important property of such function is the tives. Practically, two situations can occur: following: a. the set of alternatives is generated by the team of If s(y,w) is strongly order preserving then its maxi- specialists, on the basis of their knowledge about the mal points in y e E satisfy the following condition: decision problem being solved, but without taking into y consideration all the possible quality factors and without y = arg max s(y - w) * Ε., π (q - D) - φ evaluating these factors. This set of alternatives is sub- yeE y y mitted to the decision maker (or group of decision where makers) for evaluation and selection. After performing D — D/ (J) η —D) the evaluation, some alternatives are rejected and possi- bly, new set is generated. Such a decision support system We will not discuss the properties of achievement func- can be called alternative oriented. tions in detail. The usefulness of the achievement func- b. The integral part of the decision support systems tion in building the interactive decision support systems constitutes the mathematical model of the decision situa- follows from the following properties: tion. The primary input to the decision support system are - maximization of the achievement function results in decision maker's aspirations; the system computes the Pareto optimality, decision which in some sense corresponds to the these - if the decision χ e E and the corresponding objec- aspirations. Such a decision support systems can be called tives y e Ey are such, thaxt s =s(y — w) = 0 then w is model oriented. attainable and Pareto optimal, Evidently, in the case of alternative oriented DSS the mathematical model can be used by experts to generate - if s < 0 then w is not attainable, alternatives or evaluate the objectives. This, what differs - if s > 0 then w is attainable, but not Pareto both approaches is the organization of feedback loop optimal. between the decision maker and the mechanism for gen- Therefore, this function can be used for computing Pareto erating alternatives. optimal decisions as well as for checking for Pareto Currently, both approaches have been implemented optimality and attainability of given w €.Ey. Moreover, and tested. The family of model oriented systems, named value of the achievement function can be meaningfully DIDAS (Dynamic Interactive Decision Analysis and Sup- interpreted - this can be treated as a kind of qualitative port) have been successfully applied to many practical distance between a given decision y and aspiration level problems - see for example Grauer at all. (1984). The w. alternative oriented system, named SCDAS (Selection Com- The general form of the achievement function pro- mittee Decision Analysis and Support) for supporting posed and analyzed by Wierzbicki for the case, when all decision making in group environment will be described in performance factors should be minimized is as follows sequel (for more detailed presentation see Lewandowski at ( —ν (w —y ) if y —w e D all., 1986). dist(7/-u>, Z>) otherwise (1) Many major decisions in management and engineering are delegated to committees or groups of decision makers. where ν : Rm -* R1 is a strongly monotone function with the The institution of a committee, remains an important property that v{w-y) = 0 for all y - w e D/intD and aspect of many decision processes; the process of commit- any norm in RTO can be taken to define the distance. In the tee decision-making must therefore be improved through above formula D denotes the negative cone D = -Rm. appropriate decision supporting tools. + Another form of achievement function, which is piecevise The problem of selecting one alternative from a finite linear can be expressed by the formula: set of alternatives presented to a committee is one of the s(y,w) = m (2) most basic and classical decision problems and has received much attention in the decision-theoretical max(/o max (y - w), (1/m) J] (y - w)) literature (Mirkin, 1979, Kuzmin, 1982, Pankova, at all. t t t t 1984). There are many variants of such a problem; here, The above function is especially useful when applied to we will consider the following formulation: 4 A. Lewandowski A committee consists of several members (denoted Procedural Framework here by k = 1,...,K)\ each member can have either equal or different voting power (denoted here by a voting One of the basic assumptions of the presented metho- power coefficient v(k)), specified a priori by the commit- dology is structuralization of the decision process. We tee charter. In addition to the committee structure, the assume, that the process consists of several well defined committee charter might specify the purpose of the phases or stages. According to the procedure or the com- committee's work, further procedural details, etc. mittee charter it is possible to advance the decision pro- The problem faced by the committee is to jointly rank cess forward only if all committee members successfully or select one or a few from a set of available decision completed all the previous phases. Details of the pro- alternatives (these might be candidates for a job, propo- cedure must be defined during the initial stage of the sals for R&D projects, alternative transportation routes, decision process. Let us consider in detail all phases of proposed sites of an industrial facility, alternative the decision making process. computer systems, etc.). The list of alternatives need not The first point on the agenda is to define the pro- be complete at the beginning of the committee's work; cedures by which the committee will operate. The ques- during the decision-making process, new alternatives may tions addressed here should include the following: be generated and subsequently evaluated. (a) What is the expected product of the committee's Evaluation of alternatives is performed by the com- work and how does it influence the appropriate pro- mittee by first specifying decision attributes (such as a cedure? The answer to this question depends on the candidate's age, experience, professional reputation, committee's charter and its perceived role. For example, etc.) and then assessing each alternative with respect to if the expected product is a short list of significantly dif- each of these attributes. The list of decision attributes ferent alternatives, procedural rules will be different (denoted by ,7=1,...,«/) might be specified in the from the case when the expected product is a consensus committee's charter or decided upon by the committee. In opinion on the "best" alternative. any case, decision attributes must be specified before (b) What aggregation rules should be adopted, and in alternatives can be evaluated and compared. particular, should outlying opinions be included in or Each alternative (denoted by i =1,...,/) must be excluded from aggregation? evaluated by the committee or its individual members. The (c) Should the committee be allowed to divide and problem consists of proposing a decision process which form coalitions that might present separate assessments together with assessment of various attributes of the of aspirations, attribute scores and thus final rankings of alternatives and aggregation of evaluations across both alternatives? attributes and committee members, leads to a final rank- The second point on the agenda is the problem ing or selection of an alternative(s) in a way that is specification. Neither the list of alternatives, nor their rational, understandable and acceptable to the committee descriptions need be complete at this stage and moreover members. this information should be not known to the committee Several approaches to this problem have been members to avoid the bias in specifying the aspiration developed; most of them are based on the classical multi- levels required in this phase. The most important part of attribute utility theory (see e.g. Keeney and Raiffa, this process that requires discussion and specification by 1976), but there are also alternative approaches, such as the entire committee is the definition of the attributes of the analytical hierarchy of Saaty (1982), the orderings of the decision and their scales of assessment. Roy (1971) or aggregation principle by Jacquet-Lagreze Various studies in decision theory suggest that a rea- (1982). Some of these approaches have been also imple- sonable number of attributes should not exceed seven to mented as microcomputer-based decision support systems: nine (see e.g. Dinkelbach, 1982); if more attributes are an interesting implementation is that of analytical hierar- suggested, they should be aggregated. For example, chy (EXPERT CHOICE, 1983) or the non-procedural pack- there might be a large number of qualitative indicators age DEMOS (1982) used for probabilistic evaluation of that are all related to the ecological impacts of the alternatives. Another commercially available implementa- planned investments; instead of using all these indicators, tion (LIGHTYEAR, 1984), based on utility theory and it is better to ask committee members to evaluate subjec- weighting coefficients specified by the user, employs a tively the attribute "ecology", that is, to translate the rather primitive decision process and is restricted to information about all these indicators into one assess- only one user, hence it is not applicable in committee ment, given originally on a verbal scale from "unaccept- decisions. able" to "excellent", into a quantitative scale, say from 0 Most of these approaches rely on either user- to 10. The other possible option is to introduce the supplied rankings of attributes and alternatives for each hierarchy in the set of objectives. This makes however attribute, pairvise comparisons of alternatives, or some the procedure much more complicated and the uncertainty equivalence principle (e.g. comparisons to a corresponding theory and software is currently not fully lottery). This information is rather difficult to obtain dur- developed. ing discussions with the committee members, therefore the During the third point on the agenda, aspiration proposed approach utilizes the principle of satisfaction and/or reservation levels for all attributes are deter- decision making based on the concept of achievement mined separately by each committee member. After these function. values are entered into the system, all necessary indica- Additionally to the aspiration level which expresses tors (disagreement indicators, dominant weighting fac- the satisfactory (or acceptable) values of all objectives, tors) can be computed. members of the committee can specify the reservation The fourth point is the analysis and discussion of level. The reservation level represents a minimum accept- aspirations by the entire committee. These discussions able level for each attribute (e.g. investment cost should are supported by the computed indicators and their not exceed some given a'priori value) whereas an aspira- graphic interpretations. tion level reflects a higher desired level of expertise. If an alternative is evaluated below the reservation level on In these discussions, the committee might address the even one attribute, it is considered unacceptable, and if it following questions: is evaluated at least equal to aspiration levels for all (a) Do the dominant weighting factors accurately attributes, it is considered highly desirable. The exten- reflect the perceptions of individual committee members sion of aspiration-led decision process to committee deci- about the relative importance of various attributes (if sion making was first proposed by Johnson (1984); details not, should the aspirations or reservations be of the procedure, theoretical background and principles corrected)? of implementation are presented in more details in the (b) What are the relevant differences and do they paper by Lewandowski at all. (1986). represent an essential disagreement about decision prin- ciples? Committee Decision Making 5 Formalization of the Decision Process (c) Does the entire committee agree to use joint, aggregated aspirations (reservations), or will there be several separate sub-group aggregations? In the previous section we presented the general structure of the decision making process. The decision The fifth point on the agenda is a survey of alterna- support system supervises the progress of the discussion tives. Discussions might center on the following: within the committee - its role is to process all the infor- (a) Are the available descriptions of alternatives mation necessary to perform the discussion, compute all adequate for judging them according to the accepted list necessary informative indicators, display graphic infor- of attributes? If the answer is negative, additional infor- mation and ensure proper structuralization of the pro- mation should be gathered by sending out questionnaires, cess. In the sequel we will consider in detail all the consulting experts etc. actions undertaken by the system during each phase of (b) Which of the available alternatives are irrelevant the process. and should be deleted from the list? This kind of cursory screening can be done in various ways. The committee might define some screening attributes and reservation Setting and discussing aspirations levels for them (of a quantitative or simple logical structure): for example, we do not accept investments Most judgmental decision processes require a choice which are more expensive than a given limit. of scales of evaluation for each decision attribute. The The sixth point on the agenda is the individual scales are often qualitative, such as unacceptable, bad, assessment of alternatives. The assignment of scores for acceptable, good, very good, excellent, though they can each attribute to each alternative is the main input of be transformed into quantitative scales for computational committee members into the system. Each member speci- purposes. When asked to specify aspiration and reserva- fies scores; the system supports this by displaying those tion levels on these scales at an early stage of the deci- assignments already made and those still to be entered. sion process, the decision maker is better prepared to The seventh point on the agenda is individual make consistent evaluations across alternatives. How- analysis of alternatives, based on calculations of the ever, we cannot expect and we should not require full achievement function which lead to a ranking of ail alter- consistency in any judgmental decision process, since not natives for each committee member. This ranking is the all relevant attributes might be evaluated and the main source of learning about the distribution of alterna- relevant information an alternatives is never completely tives relative to aspirations. shared by all committee members. If each committee member is asked independently to specify his aspiration The questions addressed by each member at this point and (or) reservation levels for each attribute, a com- might be as follows: parison of such results across the committee and across (a) Do the rankings along each attribute correctly attributes serves several purposes: represent the individual's evaluations of alternatives; does the achievement ranking, based on individual aspira- (a) the relative importance of each attribute for each committee member and across the committee, as tions, correctly represent the aggregate evaluation (if implied by the more or less attainable levels, becomes not, should the scores be modified)? apparent, as discussed below. (c) If the committee member agrees with the indivi- (b) the division of opinions among the committee dual achievement ranking proposed by the system, what members can be discussed: if a significant subset of the are the differences between this ranking and that based committee has high aspirations (reservations) for an on individual scores but related to committee aggregated attribute and another subset has low aspirations (reser- aspirations? Are these differences significant, or can he vations), it is a case of a clear disagreement on decision accept them as the result of agreement on joint decision principles. The committee might then discuss this principles? disagreement and come to a consensus; or agree to The eighth point on the agenda is a committee dis- disagree by allowing the formation of coalitions that rally cussion of the essential differences in scoring and for the importance of various attributes (for example, disagreements about the preliminary ranking of alterna- when deciding on siting an industrial facility, a part of the tives aggregated across the committee. These discussions committee might be more concerned with environmental are supported by the system; the system computes indica- impacts, another more concerned with economic impacts). tors of differences of opinion and prepares a preliminary aggregated ranking. (c) if the discussion shows that the reason for disagreement stems from different perceptions by various The questions addressed by the committee at this committee members about the exact meaning of a particu- point might be the following: lar attribute and its scale of evaluation, the result might (a) On which attributes are the largest differences in be a better specification of, or at least corrections in, scoring between committee members observed? Do these the list of attributes. differences represent essential differences in information (d) if the committee (or a coalition inside the commit- about the same alternative? tee) agrees to use averaged aspiration and (or) reserva- (b) What is the essential information (or uncertainty tion levels, each committee member has a better percep- about such information) that causes these differences? tion of the anchor points to be used when evaluating alter- Should additional information be gathered, or can certain natives. committee members supply this information? In order to support these discussions, a number of (c) Would the results of these discussions and possi- indicators can be computed. Denote the individually ble changes of scoring influence the preliminary aggre- specified aspiration levels for attribute j by the commit- gated ranking list proposed by the system? This can be tee member k by p(j,k) and the corresponding reserva- tested by applying simple sensitivity analyses. tion levels by r(j,k). Then the committee "voting" pro- (d) Does the preliminary ranking proposed by the cedure might specify an averaging of individual inputs, system correctly represent prevalent committee prefer- weighted by the voting power coefficients as follows: ences? Κ Κ After these discussions, a return to any previous PU) = Σ v{k)p(j,k)/ Σ v{k) (3a) points on the agenda is possible. If the committee decides k =1 k=\ that the decision problem has been sufficiently clarified, Κ Κ it can proceed to the final, ninth point on the agenda: rU) = Σ w<*)r(j\*)/ Σ <3b> agreement on the aggregated ranking or selection of one k. =1 k =1 or more alternatives. It is important to stress again that Such an average is subject to manipulations by com- the committee need not stick to the ranking proposed by mittee members who have an incentive to distort their the system, since the purpose of this ranking - as well as true aspirations in order to influence the entire commit- of all information presented by the system - is to clarify tee. A classical remedy, successfully used in subjective the decision situation rather than to prescribe the action evaluations of certain sport performances (e.g. ice skat- that should be taken by the committee.