The quest for optimality

The quest for optimality

Book Remews the abstraction from the 'dirty' reality keeps the space free for concentrating on general technical principles. The book is recommended t...

300KB Sizes 13 Downloads 170 Views

Book Remews the abstraction from the 'dirty' reality keeps the space free for concentrating on general technical principles. The book is recommended to all those who are interested in any branch of combinatorial optimization.

H. M(JLLER.MERBA Ci I Universtt~t Kaiserslautern Kalserslautern, Germany, Fed. Rep Jean H.P. PAELINCK and Paulus H. VOSSEN (eds.)

The Quest for Optimality Volume 1 in: Issues in Interdisciplinary Studies, Gower, Aldershot, 1984, ix + 166 pages, £15.95 This book is the result of a project promoted by Paelinck and Vossen over a period of three years from 1978-1981. The contributions came from the fields of economics, operations research, management science, law, cognitive psychology, and mathematical psychology. The linking theme is that of opttmahty as it relates to the disciplines of the various authors. I found the book to be a very intriguing one, although not always addressing itself to the optimality issue as one might expect from the title. It is however full of related issues, some of whach I feel to be very controversial and, in some cases, at almost every step makes one sit back and examine the significance of the point being made which, in its own right, might justify a chapter. The chapters which I found closest to the central optimality issue were both by Schoemaker. In Chapter 2 he addresses the question of optimati,y principles in science, and the chapter is a rather thought provoking discussion of the way nature oeems ".o be governed by optimality principles such as those of Hamilton's principle of least actton, Farmat's pnnclple of least ttme for light rays, and so on. Central to the discussion is the question of causal and teleological explanations, and the belief that even in the physical sciences a search for a teleological e~planation is hkely to be more productive than a search for mere causahty. Can we think of physical systems being intentional in their behaviour in the light of the optimality endowed m their various principles. Schoemaker takes pmns to point out that optimality is a feature of a model, and only of reality

279

in as much as reality is somehow close to the model. We are thus led to the question of limitations of models, a central theme of most of the other authors. Schoemaker's Chapter 6 is a very compact and thorough discussion of expected uuh~y from various points of view (perhaps, with a rich enough model the expected utility principle is followed by everyone?). This chapter alone could command an equally long review. Taken together with Schoemaker's chapters, the Chapter 7 of Bree, makes a nice companion chapter, although one which I find contentmus (but then, the topic of optmlality is contentmus). As with Schoemaker, optimality res:des within the model, but then this is not exactly reality. Survival is a key feature of systems, and reqmres adaptability, but not optimisation (businesses do not need to e,ptimise ix, order to survive, true, but they prefer to do more than survive, and to do as well as they can). This leads the author to experts and heunsttcs as instruments in the search for good solutions (but I would ask if solutions are good or not good, whaz is good, and is it not linked to the notion of optimal?). Questions of subjective probability and utlhty are ra:sed (can we expect a dec,sion maker to have these stored in his brain for use when the occasmn arises, and ff not, where does utility theory stand - I would argue that the aulhor has a different perspective of utility theory to mine, but not necessarily a less vahd one). There are many other thought provoking points made. These are the chapters which I got most from. The other chapters seem to be a httle less to the main point and somewhat less penetrating m this respect. Paelinck discusses vanous approaches to decision making in terms of what he calls level O, 1, 2 approaches. Rinnooy Kan places has emphasis on a discussmn of dffterem behefs that mathematical models in operatmnal research either do not adequately describe the real world, or, if they do, the solutions do not yield a proper gmdelme for acUon. Finally Wesemann discusses cost benefit analys~s m cnrmnal taw, concentrating on possible legislat:on to deter motorists from speeding. Although it does not fit easily into the rest of the book, where the remaining articles are of a general kind, nor does it really address itself to the opt:o reality theme. It does develop a model, :t does introduce {he ideas of Pareto opt~mahty. It does

280

Book Remews

~3ive an interesting insight into decision making in aenal systems, and raises issues specific to this study, such as should the crimmal ha~e h~s beaefits included in the model, or should his be!aaviour only be part of the model in as much as he ~nfluences the benefits to others?. On the whole a very thought provoking little book which will be of value to social scientists, management scientists and operations researchers providing they are prepared to grapple seriously •,with the many points made, and not just to accept them at a superficial level as a means of reinforcing their own beliefs on the utility of optimality as a guiding concept in their own areas. I for one could not do without the concept. It seems quite possible that, using the postdiction principle, e.veryone is optimising in the light of the way they see their sltuatmn (although it may well be impossible to know exactly how they see it). It is, however, the task of operational research to help individuals do the best they can under the circumstances, using rational argument, and this means building models (whose reality we wilt never be sore of) as a means towards this rationalisation. To me, the survey of concepts used, such as imp,:oved, good, acceptable, satisfac,ory, are inevitably related to the concept of opttmahty, in an appropriate setting. In the final analysis we act on the basis of what we percewe, and to some this is as real as it is possible to be. The operational researcher may seek to change this perception, but, to do so, he has to link th~s change to approprmte concepts, so that such changes in such a perception may be evaluated.

D.J. WHITE Unwerstty of Manchester Manchester, M13 9TL, United Kmgdom Nmll M. FRASER and Keith W. HIPEL Conflict Analysls: Models and Resolutions Volume 11 in: North-Holland Series in System SCence and Engineering, North-Holland, New York, 1984, ~ + 377 pages, $34.50 This book presents a descriptive conflict analysis methodology based on the theory of metagames developed by N. Howard. Although the methodology is si~nple, its use to study conflicts requires in most situations the analysis and classification of lzrge numbers of possible ¢~utcomcs~ making its application tedious.

The methodology consists of identifying all possible courses of action of the conflicting parties. This information is then coded in the form of a vector of O's and l's for each possible outcome. A zero represents that a course of action has not been selected, while a I is assigned if it has. The analysis proceeds by eliminating all infeasible outcomes (e.g., contradictory courses of action), and arranging the remaining ones according to the preferences of the parties. If the information about a party's preferences of the outcomes is not known with certainty hypergames are used to analyze the conflict. The purpose of the methodology is to ldt~ntify those outcomes or courses of action that lead to a stable situation. The analysis is performed with a specific time-frame m mind. If the methodology is used for sequential time-frames the result is known as a supergame. The book consists of 14 chapters divided into two parts, and 6 appendices. Part I (Chapters 1-9) is dedicated to introduce the basic methodology and illustrate its use, while Part II (Chapters 10-14) is of a more theoretical nature and it is used to briefly introduce the reader to metagame theory and compare the methodology with other conflict analysis methods. Part II requires some basic knowledge of game theory. Overall the book is well written and makes good use of the examples to illustrate the me Lhodology. However, a problem with this type of combinatorial methodology is that because most of the examples revolve large numbers of possible outcomes, the analysis and explanations require the reader to go back and forth from table to table to identify the outcomes to which the authors ~'e,fer. Since each outcome is represented by a vector of O's and l's, the authors have attempted to simplify the analysis by transforming the vector from a number in the binary system into a number in the decimal system. This may stmplify the stability analysis of the outcome but may also contribute to the loss of a logical trend in the analysis. A computer program is available for applications, a positive aspect of the book, as the use of the method would be too cumbersome without i~.. Nonetheless, a fundamental problem encountered with tlfis methodology is that it does not allow the parties involved to make tradeoffs among the issues involved m the conflict, and cannot be used to develop possible solutions. It yields outcomes but does not allow for testing possible courses of