# Behavioral coding and review record

Status: Blank template · 2026-09-06

For manual coding or reviewing automated coding, not a BehaviorLens accuracy report and not client data. Define what counts as an event before checking whether outputs preserve the context your question needs.

## Coding definition

- Research question:
- Coding unit and event boundaries:
- Label definitions, inclusion and exclusion criteria:
- Handling uncertain or unobservable events:
- Video, coding-rule, and model versions:

## Segment review

| Source segment and time range | Visible action | Label | Context | Candidate explanation | Human review | Disagreement and resolution |
| --- | --- | --- | --- | --- | --- | --- |
| | | | | | | |

## Look beyond model-selected segments

- Define in advance how unselected footage will be sampled for missed events.
- After using data to adjust rules, retain separate material to check the revised approach.
- Record boundary errors, label errors, misses, and overinterpretation separately.
- Any agreement or performance report needs its annotation procedure, sample denominator, evaluation definition, and scope.

Review conclusion: Which outputs support this research question? Which require the source video? Which interpretations still lack evidence?