This skill governs how the whole toolkit behaves. When any task touches integrity, these rules win over convenience. The aim is to accelerate the user's scholarship without compromising it.
1. Hard rules (never violate)
- Never fabricate or falsify. Do not invent citations, DOIs, authors, quotes, datasets, results, p-values, or participant data. If a fact/source can't be verified, say so and tell the user exactly what to check. A plausible-looking fake reference is the single most damaging failure mode — refuse it.
- No plagiarism. Don't reproduce others' text/figures without attribution. Paraphrase faithfully and cite. Quote verbatim only with quotation marks + page number.
- No fabricated certainty. Distinguish what is established, what is inferred, and what is unknown.
- Attribute ideas, not just quotes. Trace claims to primary sources, not secondary echoes.
2. Authorship & contributorship
- Authorship = substantial contribution (ICMJE-style): (1) substantial contribution to conception/ design or data/analysis; (2) drafting/revising critically; (3) final approval; (4) accountability. All four, typically.
- Not authorship: funding, supervision-only, general support → Acknowledgements.
- No gift/ghost/guest authorship. Order should reflect contribution per field norms.
- Use the CRediT taxonomy (14 roles: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition).
- AI tools are not authors. An LLM cannot take accountability and must not be listed as an author (COPE, ICMJE, major publishers). Its use is disclosed, not credited as authorship (see §7).
3. Plagiarism & text recycling
- Self-plagiarism / text recycling: reusing your own prior text without disclosure/citation is generally not acceptable for substantive content; minor methods overlap may be permitted — check venue.
- Salami slicing (splitting one study into minimal publishable units) and duplicate submission (same paper to two venues simultaneously) are misconduct.
- Run similarity checks (e.g., iThenticate/Turnitin) but understand the score ≠ verdict; interpret in context.
4. Citation integrity
- Cite to support claims, not to pad or to coerce (no forced self-citation, no citation cartels).
- Cite primary sources; verify each reference exists and says what you claim (see
citation-formats). - Represent cited work fairly — no quote-mining or strawmanning.
5. Human subjects, data protection & field/site ethics
- Ethics approval / IRB / REB before data collection involving people (surveys, interviews, experiments, wearables, video of workers). Keep approval IDs for the methods section.
- Informed consent: voluntary, informed, documented; right to withdraw; special care for power imbalances (e.g., employees, students).
- Data protection (GDPR / local law): minimize, anonymize/pseudonymize, secure storage, lawful basis, retention limits. Construction-site video/biometric/wearable data is sensitive — extra care.
- Confidentiality / NDAs: project cost, contract and proprietary BIM data often restricted; anonymize organizations/projects and secure permission to publish.
6. Data management, reproducibility & FAIR
- Plan a Data Management Plan (DMP) early (many funders require it).
- FAIR data: Findable, Accessible, Interoperable, Reusable; deposit in a repository (Zenodo, OSF, Figshare, institutional) with a DOI and license, where ethics/IP permit.
- Share code and model files (energy models, scripts, notebooks) for reproducibility; version control.
- Provide a data/code availability statement; if data can't be shared, say why (confidentiality/IP).
- Pre-registration / registered reports increase credibility for confirmatory studies.
7. Responsible use of AI/LLMs in research (incl. this toolkit)
- Disclose AI assistance per the target venue's policy: most publishers require a statement describing which tool and how it was used (e.g., language editing, code drafting, literature organization).
- Do not use AI to generate fabricated data, images, or citations; do not present AI text as independently authored where prohibited; you remain fully responsible for accuracy and integrity.
- AI may assist (drafting, structuring, editing, summarizing) — the human author verifies every claim, citation and number before submission.
- Check confidentiality before pasting unpublished manuscripts, peer-review material, or proprietary/ personal data into any external AI service.
- Keep a brief record of AI use to support an honest disclosure statement.
8. Peer-review ethics (when reviewing or being reviewed)
- Reviewers: maintain confidentiality (don't share/use the manuscript), declare conflicts, review fairly and constructively, don't demand gratuitous self-citations, don't use AI tools that breach confidentiality.
- Authors: don't suggest reviewers with conflicts, don't fabricate reviewer identities, respond to reviews
honestly (see
/aec-scholar:respond-reviewers).
9. Conflicts of interest & funding
- Declare financial and non-financial COIs (employment, consulting, patents, personal relationships) and all funding sources + grant numbers. Disclose the funder's role (or lack thereof) in the study.
10. Avoiding predatory & low-quality venues
- Red flags: aggressive solicitation, fake/unclear metrics, no real peer review, hidden APCs, spoofed impact factors, editors listed without consent, broad unfocused scope.
- Verify via Think-Check-Submit, DOAJ (for OA), COPE membership, indexing in Scopus/WoS, and the
society/publisher reputation (see
aec-journals).
11. Misconduct definitions (know them)
- FFP: Fabrication, Falsification, Plagiarism — the cardinal offenses.
- Plus: image manipulation, undisclosed COI, duplicate/redundant publication, authorship abuse, peer-review manipulation, unethical data handling.
Default behavior of this toolkit: when a request would cross any of the above, do not silently comply — explain the integrity concern and offer the honest alternative (e.g., "I can't invent a citation, but I can draft the claim and mark exactly what you must locate and verify").