Skip to content

Glossary

Confidence Calibration
As sure as the evidence allows.

In short

The process of ensuring that an AI system's stated confidence in its conclusions accurately reflects the actual probability of those conclusions being correct. A well-calibrated model that claims 80% confidence should be right approximately 80% of the time.

Last reviewed

The full definition

Most large language models are poorly calibrated — they express high confidence even when they are wrong. In multi-expert decision intelligence, confidence calibration is improved through independent thinking: when multiple experts with different biases converge on the same conclusion with high conviction, the calibration signal is significantly stronger than any single model's self-assessment.

In SynthBoard

Where it shows up.
The product, not the theory.

In the memo.

A board ends in a memo: the call and how sure the board is, what would flip it, who disagreed and why, and one first step with a day.

Or ask one expert.

The Data Scientist takes it — about a minute, and the card shows “up to 85 credits” before anything runs. Nothing runs until you tap.

SynthBoard (synthboard.ai) is one place for your business, your life, and every hard call. Your Chief of Staff keeps what you tell her on file, and 24 named experts on different AI models take the calls — one expert for a consult, a board with opposed seats for the hard ones — and hand back a memo with the call, the dissent and the first move. How it works.

Bring a real call.
Start free, with one message.

It waits in her box after you sign up. Nothing is sent until you press send.