An Expert Is Not an Oracle

Society swings between blind trust in anyone with a diploma and proud faith in its own research. An expert is not an oracle, but that does not make a random opinion equal to a professional assessment.

Experts differ in education, experience, conscientiousness and the ability to explain. Some tend to over-insure. Some have an interest in selling a service. Some know the subject perfectly and understand your particular problem badly.

Meanwhile the formula “do not trust experts” leaves a person dependent on the most confident non-expert available.

It makes more sense to check:

• education and credentials;
• experience with this specific task;
• independence;
• an explanation of the method;
• possible conflicts of interest;
• alternatives;
• guarantees and liability.

Between the expert and the confident non-expert there is now one more character: a system that can speak in an expert’s voice. It can help you prepare questions, decode a term and compare options, but it gains from this no licence, no access to all the circumstances and no responsibility for the decision.

A reference produced by an AI needs the usual check: does the source exist, is it current, does it apply to the right country, and does it actually support the conclusion drawn. In medicine, law, engineering and finance, a smooth answer is an especially poor substitute for professional liability.

Consulting an AI is fine. But it reflects the picture of the world it was given and does not know your specific situation from an independent source. The higher the cost of a mistake, the more important it is to move from a conversation with a model to a person who can study the facts, ask an unexpected question and put their name to a recommendation.

A second opinion is particularly sensible when:

• the decision is irreversible;
• the price is high;
• the diagnosis or assessment is doubtful;
• experts disagree;
• a complex and expensive service is being proposed;
• the person cannot explain their grounds.

A third opinion does not always add clarity. Sometimes it only lets you pick the answer that is nicer to hear.

A recommendation from someone you know is a good way to find a candidate but no proof of competence. They are judging their own experience, which may have been simple, lucky or based mainly on pleasant conversation.

You can recommend an expert if you state the limit clearly:

He handled a problem like this well for me. I do not know how well he fits yours.

That is more honest than a blanket ban on recommendations and more useful than promising that someone is “the best there is.”

What is worth judging is not confidence but the way a person describes the boundaries. “Most likely A; if it has not cleared up in a week, then it is B, and then we do this” is a testable construction: it has a deadline, a condition and a next step. “It is all clear, let us proceed,” with no answer to “why,” is selling confidence, not competence. It works the same way with a doctor, a lawyer, a mechanic and a roofer; only the cost of the mistake differs.

A second opinion is worth asking for before the first one has acquired a deposit and a dismantled wall. And it is better to ask blind: an expert who is told someone else’s conclusion up front tends to confirm it rather than test it.