The Field Guide › Paper
Pelrine brings social complexity science, and specifically Snowden's Cynefin framework, to bear on Agile software development, and the argument doubles as a general case about knowledge work. His claim is that building software is not merely complicated but complex, a wicked problem in Rittel and Webber's sense, with incomplete and shifting requirements whose solutions cannot be specified up front; and that Agile methods work, when they work, precisely because they are complexity-management methods in disguise. The foil is the machine paradigm: the Newtonian, Waterfall assumption that the whole is the sum of its parts, that everything can be planned in advance, and that people are interchangeable units on a production line. That model, Pelrine argues, is fundamentally incapable of dealing with the change and interdependence real projects exhibit, and its persistence in management thought long after physics abandoned it is itself part of the problem. Against it he sets Cynefin's probe-sense-respond: in the complex domain you cannot analyse your way to the answer, so you set boundaries, run many small safe-to-fail probes, make sense of what comes back, and amplify what works while dampening what does not, which he notes is exactly the apply-inspect-adapt loop at the heart of Scrum. The paper's sharpest idea is retrospective coherence, the observation that in a complex system causality only becomes legible after the fact: you can explain afterwards why a project succeeded, but you cannot guarantee the same result by repeating the same steps, which is a precise diagnosis of why best-practice cargo-culting and by-the-rules Agile so often fail. For a corpus about psychological safety the connective tissue is explicit in Pelrine's own rules of thumb, the first of which, we do not make mistakes, we learn, is a plain call for a safe-fail environment in which it is acceptable to be wrong and to correct course on the strength of that learning; his self-organising teams, his insistence that whoever holds the risk makes the decision, and his probe-sense-respond loop all presuppose people who feel safe enough to surface failure, question assumptions and decide. The paper thus sits at the junction of the map's complexity cluster and its account of team practice, arguing that safety to fail is not a soft add-on but a structural requirement of working in the complex domain. Its limits are those of a reflective practitioner essay: the evidence is experience and a sense-making exercise run with several hundred Agilists rather than controlled data, and the frame is software, so the wider transfer is by argument rather than demonstration. (Text drawn from the 2011 Emergence: Complexity & Organization essay, 13(1-2), pp. 26-37.)