Aiding Decisions with Multiple Criteria Essays in Honor of Bernard R

  • PDF / 2,308,639 Bytes
  • 21 Pages / 439 x 666 pts Page_size
  • 35 Downloads / 158 Views

DOWNLOAD

REPORT


Abstract:

In decentralized decisions systems, coordination and efficiency encounter major difficulties. Risk management systems are particularly important cases in corporations with multiple plants. To solve the problem, it is argued, the analyst needs to raise a cognitive representation question, in particular the question of the criteria according to which the problem at hand is being assessed in view of the whole organization. This, in tum, raises the issue of how these criteria are evaluated by the different individuals. An example based on a subset of the risk management system, namely the maintenance system in nuclear power plants, is used throughout the text. The paper argues that generalizing MAUT to rank dependent risk treatment is of utmost importance in order to deal with such problems. One additional theorem is proved in that perspective and an appropriate software reported upon and illustrated on an example. Beyond the technical problems examined in the paper, the art to use the decision analysis framework is discussed.

Key words:

Cognition; Decision analysis; Industrial maintenance; Multicriteria decision making; Rank dependent model; Risk management

1.

Introduction

In group decision analysis, conflicts and fairness issues are not the only types of problems to be dealt with. Indeed, one of the frequently encountered problems in modem global corporations is the question of cognitive coordination of individuals. Yet solving such coordination questions turns out to be increasingly difficult because individuals have to work more and more on common tasks while having different representations of the task, indeed often different backgrounds on which they approach the task.

D. Bouyssou et al. (eds.), Aiding Decisions with Multiple Criteria © Kluwer Academic Publishers 2002

358

AIDING DECISIONS WITH MULTIPLE CRITERIA

The example on which this paper will rely all along refers to a subset of the risk management system within modem corporations, namely to the maintenance system in nuclear power plants, but a similar analysis could be relevant to quality control and to many other types of multiple agents systems as long as (i) risk considerations play an important role for the system and (ii) the system's members display some heterogeneity in terms of personal risk attitude and to some extent in terms of corporate culture, the latter being often very important. Typically, indeed, a risk management system entails specialists in production processes, others in organizational reliability, still others in workforce health safety, who all make risk assessments of their specific problems and try to take 'self protection' steps to the best of their knowledge. For example, in the production process, machine inspection is reinforced, or maintenance made more focused on reliability. At the same time, organizational routines undergo substantial changes as a result of more stringent reliability requirements [Weick, 1987]. And, finally, at the 'end' of the system, 'risk managers' deal with the question of how to fi