【正文】 by the system. The user interaction with the puterized system is required at the following instances: (1) relevant data extraction from the past project performance data (to be used in the DPM)。 ., the influence pattern is a shadow work of attributes and these attributes are significant only when they attain the active state. This would imply that there has been a change in the status or value of the attribute from what was assumed at the estimating stage. This change in state of an attribute would thus directly influence the cost of certain line items that were estimated based on an assumed status or value of the attribute. These assumptions collectively provide a structured environment for modeling the plex interrelationship between the attributes and to make the DSS more responsive to the user. THE DSS COMPASS A DSS is defined as a puterbased system for decision support, with an ability to improve the effectiveness and productivity of the decision maker by utilizing the builtin analytical, situation modeling, and database management facilities (Ghiaseddin 1987). Accordingly, COMPASS was developed in three modules (refer to Fig. 2): (1) module Ito isolate pertinent information from past project performance data and to calibrate the data for a new project with respect to the project characteristics。 (2) the interrelationships between attributes established in the influence pattern。 ., attribute F (refer to Fig. 1) can attain the active state (., F = 1) only if at least one of its preceding attributes C or D is in the active state (., C = 1 or D = 1). However, this constraint is not applicable for quantifiable attributes, ., X, Y, and Z (refer to Fig. I), because quantifiable attributes, apart from being influenced by their preceding attributes, are also directly related with certain line items (., quantifiable attribute total material cost would be related with material cost associated with various other line items), some of which might be influenced by other active attributes that would define the state of that quantifiable attribute (., total material cost) as active Assumption 3 Only the starting attributes, ., A and B (refer to Fig. 1), can be influenced by factors external to the system, whereas other attributes within the system can only be influenced by attributes preceding them in the influence pattern (refer to Fig. 1). The system represents all of the attributes included in the influence pattern Assumption 4 There is a probability that, although an attribute is in the active state, the attributes influenced by it might not get into the active state, ., C = 1 and D = 1 but F =0 (refer to Fig. 1) A corollary to assumption 4 would be that the active state probability of an attribute is a function of the independent influence of its preceding attributes, as defined in the influence pattern, ., p(F =1) =f{p[(C =1) n (F =1)], p[(D =1) n (F = I)]}. It is important to note that the accuracy of the active state probability of attributes is contingent upon the interrelationships defined in the influence pattern by the user. For example, if attribute F were influenced by a third attribute (say, H) in addition to C and D (as defined in Fig. 1), then p(F = 1) =f{p[(C =1) n (F = 1)], p[(D =1) n (F =1)], p[(H =1) n (F = I)]). However, since only C and D have been defined as the attributes preceding F, p(F = 1) will only reflect the influence of C and D. Assumption 5 If an attribute gets into the active state, it has an independent capacity to cause a certain percentage cost escalation (% CE) in the estimated project cost, ., if an attribute gets into the active state, it might influence the attributes following it, and also independently cause a % CE by influencing certain line items that were estimated based on an assumed state of the attribute. All the assumptions have been carefully considered to provide an ease in putation and modeling of the plex nature of the first assumption is necessary to create a situation that would provide ease in puting the active state probability of attributes and in modeling the interrelationship between the attributes. It might be argued that in the construction context, all the attributes are interrelated under one situation or another and are thus dependent. However, it is putationally tedious and unproductive to consider the labyrinth of relationships existing between the attributes. Thus, it is imperative to define a structured and putationally manageable approach, as defined in the assumption. The second and third assumptions are derived from (1) the definition of the system (defined earlier。COMPASSNEW PARADIGM FOR PROJECT COST CONTROL STRATEGY AND PLANNING By Makarand Hastak/ Associate Member, ASCE, Daniel W. Halpin,2 Member, ASCE, and Jorge Vanegas/ Associate Member, ASCE ABSTRACT: The need to remain petitive while generating profit requires management to develop innovative. cost management strategies that will allow them to distinguish and control earlyon factors that might adversely. impact the cost of a project. This paper describes a decision support system, COMPASS (Cost Management Planning Support System) for project cost control strategy and planning. Throughout the life cycle of a project, COMPASS methodology assists management in evaluating the potential degree of cost escalation. It also identifies attributes such as management errors, regulatory approval, and error/rework, that might be the cause for project cost escalation. Furthermore, COMPASS assists management in formulating a cost control strategy while utilizing their experience and past project performance data. The attributes identified by the cost control strategy, if controlled, would minimize the expected loss. INTRODUCTION Project ost sca