CK-23 Confidence calibration check (reliability diagram; expected calibration error; temperature scaling; P(True); verbalised confidence)
Check that a stated confidence of 80 per cent matches being right about 80 per cent of the time, rescale confidence if not, and produce a calibration curve and error.
What it does
Checks that a stated confidence of 80 per cent corresponds to being right about 80 per cent of the time, and rescales confidence if not. Artefact: a calibration curve and error. Read it before using confidence to route items to human review.
When to use it
Use when confidence scores will be used to select items for human review or to decide which outputs to trust.
When it does not fit
Needs labelled correctness data. Misfit: confidence is a prompt-generated number with no labelled data to test it. Neither study concerns free-text extraction.
Result format
The result has a calibration curve showing stated confidence against the observed rate of correct answers, and a calibration error figure. It states whether confidence can be used to route items to human review and gives the rescaling applied if it was not calibrated.
Requirements on the material
| Property | Requirement type | Requirement |
|---|---|---|
| Content type | Always applies | requires=text; requires=tabular |
| Unit of analysis | Always applies | requires=claim; requires=record |
| Level of detail available | Always applies | requires=record-level |
| Appraises source credibility | Applies in some cases | no |
Not restricted on: Structure, Variable types, Time structure, How the sample was chosen, Number of sources, Whether sources can be joined, Amount of material, Known data problems, Producer interest, Source dependence, Origin, Document genre, Language, Personal information, Security classification, Licence.
| Minimum material | labelled correctness data for at least several hundred items |
|---|---|
| Additional sources needed | none |
| Not for (rules) | none |
| Not for (text) | Wrong choice when confidence is a prompt-generated number and there is no labelled correctness data to test it. The studies cited for it do not cover free-text extraction. |
Choosing and ordering
| Kind of technique | Verification technique, runs in step 6 |
|---|---|
| Phase tag | validate |
| Other phase tags | none |
| Kind of finding | risk |
| Question answered | how much |
| Strongest statement it supports | describe |
| Needs as input |
|
| Makes as output |
|
| Must exist first | None |
| Loop target | return to gather |
Library checks by stakes
Each stakes level includes the checks of the level below it. Checks that need a person apply at high stakes only, so the process leaves them out at lower stakes (decision 38). Repeatability: deterministic.
| Stakes | Check | Person involved | Counts as a different method |
|---|---|---|---|
| exploratory | input validation | none | no |
| decision-informing | input validation | none | no |
| decision-informing | code execution for arithmetic | none | no |
| decision-informing | spot check | optional | no |
| high-stakes | input validation | none | no |
| high-stakes | code execution for arithmetic | none | no |
| high-stakes | spot check | optional | no |
| high-stakes | gold standard agreement | conditional | no |
| high-stakes | independent review | needed | no |
Failure modes and limits
| Model suitability | model plus code |
|---|---|
| Typical model failure | Verbalised confidence contains no information, or answers are overconfident. |
| Error sources | input data; language model; statistical assumptions |
| Certainty limits | Certainty drops when the calibration set differs from the target data (indirectness) and when bins contain few items (imprecision). |
Related techniques
- should follow MM-08 LLM coding and zero-shot classification of text (LLM annotation; LLM-assisted deductive coding; few-shot classification; natural language inference classification): model codes with stated confidence are what is calibrated
- should follow PE-10 Schema-based structured extraction with an LLM (LLM (large language model) information extraction; form filling from documents): extracted fields with confidence are what is calibrated
- pairs with CK-26 Independent human review of model output (expert review; adjudication; human oversight): calibrated confidence decides which items go to human review
Evidence and tag basis
Tested. Modern neural classifiers were poorly calibrated and temperature scaling worked well (2017). Larger models were well calibrated on multiple-choice and true or false questions in the right format, with weaker calibration on new tasks (2022). (Guo et al. 2017, arXiv 1706.04599; Kadavath et al. 2022, arXiv 2207.05221)
Tag basis: input profile inference; selection inference; sequencing inference; robustness inference. All tag values are inference unless a source is named.