What we seem to agree on is “if I don’t have much a priori information about system X I scan around the vicinity for clues; if I do, I make adjustments to realign my theory with reality”. I believe this is handled by the statement that we specialize in a few things and apply generalist approaches to understanding everything else.

The rub occurs when we designate generalists and specialists in advance. This is particular in the equivocation of generalist to the proverbial jack of all trades, and the specialist as one in a particular field. My point is if the word is in your job title, it’s going to colour how you tackle a given problem—and I assert that any sufficiently complex problem will possess anatomy for which would-be generalists have ample experience and for which so-called specialists are clueless.

A given problem space can possess an arbitrary number of sub-problems which are too obscure to be named, yet a given individual could well have experience with them, and thus be an ad-hoc specialist. The number of anonymous sub-problems can be surprisingly large, and they can be of crucial importance. If the problem space is a project upon which many people are collaborating, then the issue is exactly political.

For a discussion on the decomposition of design problems, see Christopher Alexander’s book Notes on the Synthesis of Form, or the more concise A City is not a Tree (Part 2). Please also consider the following letter by Edsger Dijkstra.

I agree that this is not about identity politics – that’s why I pointed out Dave first made an important point to say we are all generalists and specialists in different circumstances; it’s about the frame of mind and choice of approach we use when in one or another mode.

What’s wrong with poles? They simply define two ends of a spectrum.

I don’t see specialist/generalist as “identity politics” but I do see them as approaches that are mutually exclusive. For any activity you engage in, professional or not, you can place your approach somewhere along that spectrum. The poles help define the activities.

Specialization is about becoming more efficient. Generalization is about looking at/trying a lot of things, many of which will not deliver results. To the degree that you specialize, you can’t generalize, and vice versa.

Consider the following:
http://poste.posterous.com/another-take-on-that-whole-generalistspeciali

So I’m saying, a generalist might seem less definite on an answer but it’s very likely they will offer a range of options that the expert never could. Then by all means bring in the surgeon, if that makes the most sense.

Every project (or really every system) possesses an anatomy (requirements, behaviours, idiosyncrasies) which is partially known and partially unknown by everyone working on it. This anatomy can be divided any number of ways into partially-overlapping subsystems. Some of these subsystems are going to be familiar and well-understood by certain people and others will be familiar to other people. Certain subsystems will have an unconventional, possibly unique shape and require the people working with them to learn and communicate their properties ad hoc.

When we introduce identity politics into the process, we unnecessarily cleave up the problem space. A “specialist” may be completely inept at the particular application of his or her domain expertise. A “generalist” likewise could easily spot patterns in the system to which others are blind. However, these roles could just as easily be reversed. I submit that ability to solve a problem depends exclusively on the complete acquisition of information, whether it is peculiar to it or applicable in principle.

Versus the generalist, who will seek out a few different frames of reference, perhaps looking at multiple systems – that’s where they spent most of their time first – then leap into narrowing things down, which STILL takes longer as they have more to synthesize.

So yes, I agree. Comparatively, Specialist will spend less time than Generalist overall, but precisely because they apply a different lens to the situation, which narrows the problem and allow them to go deep.