Confirmation Bias in Diagnosis

Why this matters

Confirmation bias is the human habit of looking for evidence that supports the answer you already believe and ignoring evidence that contradicts it. In diagnosis it is expensive: you lock onto a theory early, then unconsciously interpret every reading as proof of it, replace the part you already suspected, and miss the real fault sitting in plain view. The best diagnosticians are not the ones who guess right first. They are the ones who do not let their first guess blind them.

How the trap closes

You form a hunch fast, often before you have real data, because your last three jobs looked like this one. From that moment, your brain stops searching neutrally. A reading that fits the hunch feels like confirmation. A reading that does not fit gets explained away ("the meter must be off," "that is probably fine"). You run the tests likely to confirm your theory and skip the ones that might kill it. By the time the part is replaced and the problem persists, you have spent the visit gathering support for a wrong answer instead of finding the right one.

The tells that you are in it

• You stopped considering alternatives. Once one theory feels obvious, the others quietly vanish. If you cannot name a second possible cause, you may be locked in.
• You are explaining away contradictions. Every time you find yourself dismissing a reading that does not fit ("that does not matter here"), pause. The thing you are dismissing might be the answer.
• You only ran confirming tests. If every test you chose was one that could only support your theory, you tested your belief, not the system.
• You are sure before the data. Certainty that arrives before the evidence is a hunch wearing a diagnosis costume.

Try to disprove your own theory

The strongest single habit against confirmation bias is to deliberately attack your leading theory instead of defending it. Ask: what test, if it came back a certain way, would prove me wrong? Then run that test. If your theory is right, it survives the attack and you are now genuinely confident. If it is wrong, you find out in one test instead of after a wasted hour and a needless part. A theory that has survived an honest attempt to disprove it is worth trusting. One you only ever tried to confirm is not.

Let the readings lead, not your expectation

Take measurements before you commit to a conclusion, and read them for what they say, not for what you hoped. The discipline is to follow the evidence even when it walks away from your favorite suspect. If the numbers point somewhere you did not expect, that is not noise to be explained away. That is the diagnosis trying to tell you something. The system does not have an opinion about what failed. It just shows you the truth if you let it.

Watch for the patterns that prime the bias

Two situations load the bias heaviest. The first is the familiar-looking job: it resembles your recent work, so you assume the same cause and skip the diagnosis. The second is the recommended-by-someone diagnosis: a prior tech, a customer, or a manual said it was X, and you spend the visit confirming X instead of checking it. In both cases the cure is the same. Treat the prior conclusion as one hypothesis among several, not as the answer, and verify it like you would any other.

The honest reset

When a diagnosis is not resolving, the professional move is to set down your theory entirely and start over from the evidence, as if you just walked in. Not "how do I prove I was right," but "what does the system actually show." Letting go of a theory you are invested in is hard, which is exactly why it is the skill that separates techs who flounder from techs who solve it. See related: The Assumption That Cost You an Hour; When to Get a Second Opinion on a Diagnosis.

References

• Trade-standard practice on hypothesis-driven, evidence-led troubleshooting.
• General references on cognitive bias in technical decision-making.
• See related: The Assumption That Cost You an Hour; When to Get a Second Opinion on a Diagnosis.