AI-Assisted Research Decision Log
Use only for decisions that materially affect the research. If an AI suggestion changes nothing important, no formal log entry is required. Do not overwrite an original entry with a later judgement. An entry written later cannot be presented as a contemporaneous record. Retain only material you are permitted to store.
A. Initial decision record (made at the time, where possible)
Project:
Decision ID:
Date/time entered (or dated document version):
Recorded by:
Research stage:
Source/reference for trigger:
Decision under consideration
What consequential research move was proposed or made? State the question, comparison, interpretation or boundary that might change.
Trigger and traceable origin
What suggested or required the move? Record the identifiable source/version where permitted; choose only what the surviving record can support.
- Researcher
- Generative AI
- Collaborator
- Literature
- Reviewer/editor
- Mixed
- Indeterminate
Evidence available when the decision was made
What was actually known at that point? Distinguish an AI proposal or planned check from an inspected source, executed analysis or supported conclusion.
Researcher decision and reason at that point
- Accepted
- Modified
- Rejected
- Deferred
- Requested further analysis
Decision and reason:
What could change this decision:
B. Later updates (append, do not replace A)
Use one row per consequential new challenge, evidential check, changed judgement or finalisation. Enter dates or traceable versions and identify who made the update. Leave unperformed checks explicitly pending.
| Date recorded / version | Recorded by | New challenge or evidence (source/reference; completed or proposed?) | Changed decision and reason | Status after update |
|---|---|---|---|---|
Status may be pending, retained, narrowed, reframed, withdrawn or replaced. Do not put a later status into Section A.
C. Retrospective reconstruction (only when needed)
Date reconstructed:
Prepared by:
Surviving sources or versions consulted:
What is remembered or inferred rather than contemporaneously shown:
Uncertainty and alternative possible histories:
A retrospective account may be useful, but it is not proof of when or by whom the original reason was expressed.
Attribution boundary at the point of assessment
Date/source of assessment:
What limited account of contribution is supported? Which alternative origination or decision histories cannot be excluded? Separate recorded proposal, researcher evaluation and successful empirical execution.
Completed example
See the retrospective PISA example, which does not claim to be a preserved original form.