Zen or the art of motorcycle maintenance An inquiry into values by Robert M. Pirsig First Edition: 1974 The body of literature effectively cited or referred to, from within the data processing professional literature or Usenet newsgroups, is not very eclectic. Not providing any statistics, I would claim that few authors are mostly cited. Come to my mind Lewis Carroll; some well-established philosophers like Plato or Aristotle, including some oriental masters like Confucius or Lao-tse; a few science-fiction authors like Douglas Adams; a couple of scientists, possible sources for insightful analogies, the architect Christopher Alexander or the cognitive linguist George Lakoff, the physicist Stephen Hawking; and closer yet to computer science proper, some famous cognitive scientists: Herbert Simon, Marvin Minsky, Terry Winograd, Douglas Hofstadter... Among them, Robert Pirsig has had his share for over 20 years, for a book which is after all, mostly an autobiography. The topic I'd like to focus onto, which is indeed central in the book, is this of quality. I consider an interesting paradox the fact that despite the appeal the book exerts over computer scientists, Pirsig's thesis concerning quality contradicts so deeply the conception Software Quality Assurance experts build upon. 1. Transcendental quality In a paper on software quality [1], an author categorizes Pirsig's conception as "transcendental" and reproaches it its "lack of rigor", which he equates to a lack of objectivity ("intensely personal [...] experiences"). The binding between scientific rigor and objectivity is quite traditional. It is nevertheless not immediate. Science as we understand it was born roughly with Descartes' rationalism and its reaction to the scholastic tradition. Rationalism was very much based on experience, and this one authorized by the Cogito. This revolution gave soon birth to empirism, personalized among others by Hume, which led Kant to formulate his "Critique of Pure Reason". This is covered by Pirsig (p 114 and following, until the end of chapter 11), and is a cornerstone of his conceptions: the meaning of the world is conditioned by our subjective experience. Now, Pirsig follows Kant in accepting a few a priori "pre-conceptual" categories, such as time and space, which allow us to build up our understanding. These categories are common to all human beings, and therefore "objective". This is what constitutes the foundation for an objective science. Although Pirsig doesn't explicitly state it as such, I understand that he considers quality as part of this pre-conceptual baggage: "What I mean (and everybody else means) by the word quality cannot be broken down into subjects and predicates. This is not because Quality is so mysterious but because Quality is so simple, immediate and direct." (p 225) Quality being objective doesn't make it an object itself, which could be the source of sense data, and thus quantified, and measured. While this allows us to communicate our experiences, quality cannot itself be their object. We have a pre-conceptual, hence objective, notion of quality, not a conceptual one. A conceptual notion would be subjective: "It's quite a machine, this a priori motorcycle. If you stop to think about it long enough you'll see that it is the main thing. The sense data confirm it but the sense data aren't it." (p 118) The motorcycle can be measured, precisely because it is being conceptualized. Quality is not. "Quality is not a thing. It is an event. It is the event at which the subject becomes aware of the object." (p 215) Approaches which in the name of objectivity attempt to get rid of subjects, are reductionist: "Caring about what you are doing is considered either unimportant or taken for granted." (p 25) Quality is the satisfaction with which one feels that one controls the world, that sense data confirm one's a priori understanding, and with which this understanding adapts to new sense data. Quality is the feeling of one's own fit into the world. 2. The shortcomings of classical quality The paper already cited went on (before loosing sight of any short term applicability in a statistical study of the many "dimensions" of software quality) reviewing classical approaches to flesh up quality with an objective structure. The four categories it came with were: - Product-based: a generative view, based on attributes of the products. - User-based: relative to customer expectations and their satisfaction, measurable through market analysis. - Process-based: relying on metrics to assert the compliance of results with specifications. - Value-based: taking cost efficiency into account, and weighing according to marginal returns on investments. The mere fact that these approaches are all indeed reasonable casts a shadow of doubt over their goal to find out an objective criteria. The largest common factor they share is precisely their project to evacuate any "subject" from the evaluation (if only by melting it into statistical nothingness). All these attempts have been so far non conclusive, to the point that it is good to formulate once again the actual needs: - Predictability. This is needed in order to make investments, to affect resources. - Means of validation. - Efficiency. These are indeed aspects of a feeling of control, and thus naturally bind to quality as we saw earlier. The question of objectivity is not as clear, however. Everybody is more after competitive advantages than after progresses of science. It is important to be convincing towards one's customers, not in front of the posterity. The birth of global markets could bring an answer, but it is still a marginal explanation. The issue, getting back to Pirsig's framework, is more one of subject: who should experience the quality? Be in control? This is the question: the requirement for an objective quality criteria comes from a social separation of tasks. Quality is sought for as an interface. What is expressly wanted is confidence without caring! Pirsig showed that this is a nonsense. Control over a process is only increased by getting involved. Delegation doesn't work. The conclusion is that the need for increased sophistication, especially in software, will require a modification of the social structures of production. The quest for objective definitions of quality is only a desperate move to preserve a status quo. This leaves us with a subsidiary question: is there an objective reality? The history of philosophy did not stop with Kant. The categories he claimed as pre-conceptual and objective have suffered a lot during the two last centuries: after Darwin, and even more after Einstein, our conception of time has changed significantly. In fact time and space cannot any more be considered distinct pre-conceptual categories for structuring the world. Heidegger came actually to the same conclusions, speaking of the "temporalization" of the Dasein. This purification of "existence" from any objective structure, and thus from any "essential" meaning, only confirms the contradiction there is in trying to make decisions, without involving one's responsibility. Conclusion Following Pirsig, I believe that one is misguided to look for objective assistance to decision making under the concept of quality. The issue is however -unfortunately?- not only one of terminology, but is symptomatic of a deeper problem with the classical conception of quality. Appendix [1] "Dimensions of Software Quality", by Michael W. Usrey, in Quality Management Journal, Vol 3, Issue 3, pp 67-86, Milwaukee, WI: ASQC, July, 1996. [2] http://dana.ucc.nau.edu/~twh/pirsig.html A web site, devoted to large quotes to the book, classified by topic. A query among Usenet postings in "Deja News", a database of Usenet postings, found 44 matches, for the query: "zen motorcycle pirsig" between 97/02/08 and 97/03/13 (motorcycle: 12435, pirsig: 339, zen: 9825).