Diagnostic Thinking

A good mechanic listens to an engine and hears information. That knock at 2000 RPM. The hesitation when cold. The smell of the exhaust. Symptoms form a pattern, the pattern suggests hypotheses, hypotheses get tested against reality. This is epistemology enacted with wrenches.

Matthew Crawford’s Shop Class as Soulcraft argues that repair is cognitive work of the highest order. Unlike assembly (follow instructions), repair requires diagnosis — figuring out what’s wrong when the machine can’t tell you. The problem space is constrained by physics but unbounded by procedure.

Diagnostic thinking follows a structure: generate hypotheses from symptoms, rank by probability and testability, test the cheapest and most likely first, update based on results. Experienced mechanics do this so fast it looks like intuition. It isn’t — it’s pattern recognition built from thousands of cases.

The novice sees symptoms as isolated facts. The expert sees them as nodes in a probability network. “Won’t start” plus “clicks when turning key” plus “lights dim” equals battery or alternator — not starter, not fuel, not ignition. Each symptom rules out branches, narrowing the space.

What diagnostic thinking requires: a model of how the system works when healthy, maintained through direct contact with functioning examples. You can’t diagnose without knowing normal. This is why the best diagnosticians have both tacit knowledge (accumulated feel) and explicit knowledge (how things work in principle).

The diagnostic stance is humble and curious. The machine knows something you don’t. Your job is to ask the right questions and listen to the answers. Force doesn’t help — a systematic approach beats a confident guess. The feedback loops between hypothesis and test is the whole game.