node&norm commits to traceable evidence, transparent methods and accountable publication. The standards below guide research conduct and reporting; each study must disclose how they were applied and what remains unresolved.
Research commitments
Claims have boundaries
Distinguish observations, interpretations and hypotheses. Report uncertainty, contradictory findings and the conditions under which a conclusion holds. Technical conformance alone does not establish practical effectiveness.
Missing evidence remains missing
An incomplete record may leave a question indeterminate. Absence from a search result does not establish absence from the underlying record.
Evidence has provenance
Record the source, date, scope, relationship to the claim, and known limitations. Identify derived analysis separately from source observations.
Results retain their method
Preserve the protocol, method version, evidence packet, code and decision rules needed to examine a result. Record departures from the planned method and distinguish exploratory analysis from prespecified tests.
Corrections preserve history
A material correction identifies the earlier state, what changed, why it changed, and which dependent claims were affected.
Conformance is bounded
A passing schema or test suite demonstrates only the properties it examines. Other claims require separate testing.
Independence is specified
Describe who examined the work, their role, relevant interests and access to evidence. External review, independent scoring and replication are distinct activities; report which occurred.
Human responsibility remains visible
Credit substantive contributions and identify responsibility for the work. Disclose material AI assistance, including its purpose and the human checks performed. Human authors remain responsible for sources, analysis and publication.
AI use & disclosure
Program practice statement · Updated · Prepared with Codex assistance
node&norm uses AI assistance in research development, publication and website work. Documented contributions extend beyond proofreading. Their scope varies by project; each output needs its own account of the assistance that affected its methods, evidence, interpretation or presentation.
| Area | Documented assistance |
|---|---|
| Research development | Proposing questions, assessment rules and methods; examining alternative interpretations; drafting research specifications and critiques. |
| Sources and evidence maps | Discovery, source inspection, extraction, screening proposals, summaries, classifications and research-to-source connections. Metadata checks and full-source reading have different scopes. |
| Writing and presentation | Manuscript and essay drafting, substantive revision, plain-language explanations, tables, figures and website design. |
| Code and checks | Implementation, test development and execution, calculation checks and error analysis. Results retain their tested conditions and unresolved limits. |
Documented tools vary by project. Codex (OpenAI) is recorded for website preparation and selected research tasks. Historical CDFI and GDI records also identify Claude (Anthropic). Other records identify an OpenAI assistant without establishing its underlying model or version. Some project experiments have separately used TypeSafe AI’s Jev to generate judgments under a specified protocol. A model used in an experiment is distinct from an assistant that helped develop the study. Exact underlying models and a complete inventory of earlier assistant sessions are not established for every task; unavailable details remain unknown.
Human responsibility and verification limits
The responsible human researchers remain accountable for adopted claims, sources, interpretations and publication. Each project preserves its contributor and review roles. AI systems are not credited as authors.
Source inspection, software tests, model outputs, AI-assisted critique and human review are reported separately. A separate AI-assisted review is not independent human verification, scholarly peer review or empirical replication. Publication approval does not by itself establish that a person reread every source, reproduced every retrieval or verified every generated judgment. A human check is attributed only to the material and activity the record supports.
Disclosure attached to each work
Project and publication records should identify the tool and model where known, dates of use, substantive tasks, affected materials, how outputs were adopted or corrected, the human checks actually performed and remaining gaps. Earlier records are not retrospectively made complete by this statement. AI suggestions and synthetic references remain identifiable; exposure to an AI suggestion must be disclosed when interpreting a human rating.
Examples: TAE’s versioned AI-assistance and author-review disclosure; CEC’s AI-use policy and development disclosure; and the map’s batch review record.
Supporting records and publication requirements
Preserve relevant decisions, versions, prompts, outputs, tests and corrections where available and permitted. Public disclosure can summarize restricted records. Credentials, unnecessary personal information, confidential manuscripts and protected evaluation material belong under appropriate access controls. Selected public records do not constitute a complete transcript of research development.
Publisher and venue requirements must be checked for each submission. ACM’s May 14, 2026 policy requires detailed Methods reporting for AI used in research and exempts writing assistance from its disclosure requirement. IEEE guidance places generated-content disclosures in acknowledgments and recommends disclosure of editing assistance. Nature’s policy distinguishes documented LLM use from narrowly defined copyediting. A website statement does not replace the disclosure required in a manuscript.
Policy references checked November 3, 2026. ACM and Nature were examined through official indexed text where direct access was restricted. These are publisher-specific policies, not a universal standard or a certification of compliance. The Disclosure is not documentation article offers a proposed documentation framework; it is not a mandatory reporting standard.
Funding, questions & findings
Commissioning an evaluation should not transfer authority over its evidentiary conclusion. Factual review, confidentiality review, and approval of a finding are separate powers.
| Field | What the record should disclose |
|---|---|
| Question authority | Who selected the question and success criteria |
| Funding provenance | Who funded the work and what financial or non-financial interests may affect it |
| Method authority | Who selected or changed the method |
| Evidence access | What could be inspected and what remained unavailable |
| Review rights | Who could correct facts, request redactions, or challenge interpretations |
| Publication rights | Who could restrict publication and on what grounds |
| Conclusion authority | Who retained responsibility for the evidentiary judgment |
Evidence access & limitations
Describe the material available for examination, the search process and the restrictions that affect a conclusion. Share supporting materials where permissions, privacy and confidentiality allow; explain restrictions and any route to controlled access.
| Investigation condition | Required disclosure |
|---|---|
| System and environment access | Full, partial, reproduced, documentary only, or unavailable |
| Transcripts and logs | Access scope, sampling, missingness, and known integrity limits |
| Model and personnel access | What could be queried or independently examined |
| Search process | Queries, tools, sampling, and limits of AI-assisted evidence discovery |
| Redactions and withheld evidence | Known restrictions and their effects on supportable conclusions |
| Resources and authority | Time, personnel, scope, publication rights, and conclusion control |
Separate research states
| State dimension | Question |
|---|---|
| Publication | What version has been frozen or released? |
| Technical | What specification or implementation exists? |
| Research maturity | What empirical progress has occurred? |
| Claim | What does the evidence support? |
| Review | What independent examination has occurred? |
| Adoption | What external institutional use is documented? |
Correction policy
A substantive correction should state the affected version, the error, the evidence for the change and its effect on the conclusions. Preserve the earlier record with a visible correction notice. Findings that can no longer be supported should be marked as withdrawn, with the reason and date. Changes in an external source should be distinguished from errors in the program’s analysis.
Ethics, concerns & accountability
Research involving people or sensitive information requires appropriate consent, privacy safeguards and any applicable ethics review before the work begins. Describe the review obtained and its scope in the study record.
Readers can raise reproducibility problems or propose corrections through the relevant research repository. Include the version, passage or result, supporting evidence and the proposed correction. Public issues should contain no confidential or identifying information. Allegations of misconduct require the appropriate institutional or publisher process, with confidentiality and fair consideration for those involved.
Guidance informing these commitments
The Singapore Statement on Research Integrity (2010) sets out responsibilities for methods, records, authorship, interests and correction of the research record. ALLEA’s European Code of Conduct for Research Integrity (2023) provides a cross-disciplinary framework for responsible research practice.
These are reference frameworks. Their inclusion does not imply external certification or endorsement of this research.