Research & Evidence · Field note 04
Documenting AI use in research and scholarship
Direct answer
A grounded place to begin
Material AI use should leave a trace that identifies the task, approved system, date, permitted input category, relevant output, human verification, corrections, retained decisions, and required disclosure. Documentation supports accountability while human judgment remains responsible for scientific meaning and final claims.
Five-part practice
Carry the idea into daily work.
- 01
Define the permitted task
State the bounded purpose for AI assistance and the decisions that remain with qualified people.
- 02
Protect the inputs
Confirm authorization and keep protected health information, identifiable participant data, confidential research material, and third-party secrets outside unapproved systems.
- 03
Record the system and date
Preserve the tool, model or service when known, access date, configuration, and material instructions needed to understand the interaction.
- 04
Verify every material output
Check claims, citations, calculations, interpretations, code, and recommendations against authoritative sources and project requirements.
- 05
Disclose proportionately
Follow journal, institutional, funder, protocol, contractual, and professional requirements for acknowledgment or disclosure.
Questions and answers
Questions that deepen the work
Can confidential research data be entered into an AI system?
Only when the data use is authorized and the selected system, agreements, security controls, protocol, and institutional requirements expressly support that processing.
Can AI qualify as an author?
Authorship standards generally require responsibilities that an AI system cannot assume. Human authors remain accountable for the work and should follow the relevant venue policy.
What AI activity is material?
Treat assistance as material when it shapes analysis, code, interpretation, claims, structure, language, figures, decisions, or other content that meaningfully enters the scholarly record.
Related pathways