Part III: Contents Of The Process Handbook
Chapter 9: Let a Thousand Gardeners Prune — Cultivating Distributed Design in Complex
9.4 Example — Normal Accidents (Perrow 1984)
Perrow's monograph on high-risk technologies (1984)[8] includes specific recommendations about taking action to address the risks of certain complex systems.
These recommendations for action can be construed as a kind of design text for high-risk systems.
My analysis of Perrow illustrates an interesting variant on the modeling process I have described so far: I first discovered a generalization of Perrow's design method and then moved to a more specific model via process specialization. This analysis also illustrates how one must sometimes read between the lines to construct an account of design from a text whose primary concerns are elsewhere.
Perrow's principal focus in the book is on understanding the nature and cause of the risks associated with certain complex technologies. It is in the concluding chapter, ''Living with High-Risk Systems,''that Perrow addresses the question of what can be done to mitigate these risks, and thus it is here that I will focus the analysis.
9.4.1 The Text
Perrow's main concern is with understanding the nature of the risks associated with complex tightly coupled systems. He offers a theoretical framework in which risks are associated with the underlying structure of the system. Much of the book is a case-by-case presentation of various systems and their risks. Along the way Perrow develops a framework in which he classifies systems according to the degree of coupling (tightly coupled systems vs. loosely coupled systems) and the nature of the interactions (linear systems vs. complex systems).
The degree of coupling is largely a matter of the amount of buffering between components in a system. In a tightly coupled system a local event produces rapid effects on other
components in the system and therefore requires a rapid response. The distinction between linear and complex systems seems to be mainly a question of whether the interactions are unexpected. However, Perrow also talks about feedback loops and one-to-many causal links.
In addressing the design of systems in his concluding chapter on living with high-risk systems, (pp. 304-52), Perrow's main proposal is that systems should be evaluated along two dimensions:
For their ''catastrophic potential''(which depends both on the nature of the system and the current level of organizational effectiveness associated with such systems).
1.
For the cost of alternative systems.
2.
Perrow then groups systems into three categories, each of which requires a different intervention:
Systems where catastrophic potential is high relative to the cost of alternatives. These should be abandoned in favor of one of the alternatives.
1.
At the other extreme, systems where the cost of alternatives is high relative to the catastrophic potential. In this case the existing system should be improved.
2.
In between, systems that cannot be easily replaced but pose significant threats.
Perrow proposes that these systems be restricted to situations where the risks are lowest and/or the benefits are greatest.
3.
9.4.2 A Generalization
My first step in analyzing this process was to adopt a point of view. Point of view is especially critical in a case like this where the text is not explicitly an account of a design process.
Where Davenport provided an explicit map of the innovation process, and Hammer and Champy provided an informal description of the reengineering process, Perrow is primarily concerned with establishing a framework for classifying high-risk systems. Here we want to look more closely to identify where and how he speaks to the process of designing such systems. The outcome of this close reading must initially be a point of view which then guides the further development of the process map.
The point of view I developed follows from Perrow's proposal for living with high-risk systems (Perrow, ch. 9). At the core of this proposal is the classification of such systems into three categories each of which warrants a different intervention (Perrow, p. 349). I chose to construe this as a kind of ''design using classification.''This point of view then immediately suggested to me a very simple process map which I represent in figure 9.14.[9]
Figure 9.14: Design using classification
9.4.3 Dependency Diagram
Having developed this general scheme for framing Perrow's texts as a design story, I then developed a specialization by decomposing APPLY APPROPRIATE INTERVENTION into the three specific interventions that Perrow proposes. As a consequence the single flow from CLASSIFY to APPLY APPROPRIATE INTERVENTION must be decomposed into three flows corresponding to the three types of system that Perrow matches to the three interventions. The resulting dependency diagram is given in figure 9.15.
Figure 9.15: Design of high-risk systems
[8]Unless otherwise noted, all references to ''Perrow''refer to (Perrow 1984).
[9]This process characterization is my own as are the activity names (unless otherwise
noted). Note that in this analysis I show the dependency diagrams without going through the intermediate phase of resource ?ow graphs. This was possible because of the relative simplicity of the generalization that anchors this analysis.