Review Confidence vs. Correctness

Confidence is the strength or certainty expressed with an answer. Correctness is whether the answer is supported by the relevant evidence and matches the facts or requirements. One does not establish the other.

An answer can sound certain and be wrong. It can also be cautious and correct. Confidence is a signal to examine, while correctness is a conclusion that requires evidence.

Fluent AI responses can make confidence easy to mistake for knowledge. Separating the two helps people verify important claims, resist persuasive wording, and decide when uncertainty should change how they act.

An assistant confidently states that a company policy permits a particular data use. The policy document says otherwise. The answer's tone was confident; its content was incorrect.

Low confidence does not prove an answer is wrong, and high confidence does not prove it is right. Calibration is a related measurement concept: it asks whether confidence levels match correctness rates across many comparable cases.

Certainty is a signal, not a verdict.

* [[calibration|Calibration]] * [[verification|Verification]] * [[independent-verification|Independent Verification]] * [[epistemic-humility|Epistemic Humility]]

* [Chuan Guo and colleagues, “On Calibration of Modern Neural Networks”](https://proceedings.mlr.press/v70/guo17a.html) — empirical confidence and accuracy are related but distinct properties. * [NIST, “AI RMF Core: Measure”](https://airc.nist.gov/airmf-resources/airmf/5-sec-core/) — uncertainty, benchmarks, testing, and documented evaluation as parts of reliable AI assessment.

* [RS AI Learning Lab, “When the Bot Sounds Right but Is Not”](../../../../modules/when-the-bot-sounds-right-but-is-not/module.md) * [RS AI Learning Lab, “Designing Systems That Can Admit They Are Wrong”](../../../../modules/designing-systems-that-can-admit-they-are-wrong/module.md)