Skip to content

Ethics · E-03

Research Ethics Guidelines

Integrity standards for the research we publish.

Adoptedv1.0Download PDF
Deliverable
E-03
Document ID
LS-ETH-003
Version
1.0
Effective
July 23, 2026
Owner
Founder
Review cycle
Annual

1Purpose & Scope

LinkScape conducts real AI and NLP research, including work that has resulted in a paper submitted to ACL Rolling Review 2027, decision pending. This policy sets the integrity, consent, data-use, and publication standards that every LinkScape research activity must meet, from a weekend Kaggle experiment to a peer-reviewed submission. It exists so that our research is honest, our participants and data sources are respected, and our published work reflects credit accurately.

This policy applies to all members, fellows, and collaborators — including minors — who design, run, analyze, or publish research under the LinkScape name or using LinkScape resources (notably the 80× NVIDIA H100 compute fleet). It complements E-01 Responsible AI (values and transparency), E-02 Data Privacy (personal-data handling), and E-05 Ethical Review Board (review authority). Where a study touches personal data, E-02 governs; where a study raises novel harm questions, E-05 governs; this document governs research conduct itself.

LinkScape is a fiscally sponsored project of The Hack Foundation dba Hack Club, a 501(c)(3); it is not a separate legal entity. This policy is an internal standard and is not a substitute for an accredited Institutional Review Board (IRB). Studies requiring formal IRB oversight must route through a partner institution — see Section 3.4.

2Research Integrity Standards

Every LinkScape researcher is personally accountable for the honesty of the work they put their name on.

2.1Core commitments

  • Honesty — report methods, data, and results truthfully; never fabricate or falsify data, metrics, or figures.
  • No plagiarism — cite all prior work, code, datasets, and model weights; never present others' work as our own.
  • Reproducibility — retain code, seeds, configs, and data pointers so a reasonable third party could reproduce reported results.
  • Full disclosure of AI assistance — where generative tools helped write code or prose, disclose per venue policy and per E-01.
  • No p-hacking or selective reporting — pre-register the primary hypothesis and report negative and null results.
  • Correct on error — if a mistake is found post-publication, notify co-authors and the venue and issue a correction promptly.

2.2Prohibited conduct

  • Fabrication or falsification of results, benchmarks, or ablations.
  • Duplicate or salami submission of the same result to multiple venues without disclosure.
  • Undisclosed conflicts of interest (see G-05 Conflict of Interest).
  • Misuse of the H100 fleet for undeclared or personal research (see F-03 Compute Resource Usage and S-04 Acceptable Use).

2.3Misconduct handling

Suspected research misconduct is reported to the Founder (research-integrity owner) and, if it involves safety or data exposure, to the CFO who owns safety/security escalation. Findings of misconduct may result in retraction, authorship removal, and offboarding per PP-07. Allegations are handled confidentially and without retaliation.

3Human-Subjects Research & Informed Consent

Some LinkScape research involves people — annotators, survey respondents, workshop participants, or evaluators of AI outputs. Because some LinkScape members are themselves minors, and because our beneficiaries include students, we apply heightened care to consent and youth safety.

3.1When consent is required

  • Any study collecting responses, annotations, or behavioral data from identifiable individuals.
  • Any human evaluation of model outputs where the evaluator's identity or demographics are recorded.
  • Any use of participant-contributed text, audio, or images (e.g., the AI-music project) beyond the original context.

3.2Informed-consent procedure

  1. Provide a plain-language description of the study purpose, what data is collected, how it will be used, and how long it is retained.
  2. State that participation is voluntary and can be withdrawn at any time without penalty, and give a contact (research owner via ceo@linkscape.app).
  3. Obtain explicit, recorded consent (checkbox, signed form, or logged verbal consent) before any data is collected.
  4. For participants under 18, obtain verifiable parental/guardian consent in addition to the minor's assent.
  5. Store consent records separately from research data, referencing E-02 Data Privacy retention rules.
  6. Honor withdrawal requests by deleting the participant's data where technically feasible.

3.3Youth-safety additions

  • No collection of more personal data from minors than the study strictly requires (data minimization).
  • A LinkScape adult member (CEO, CFO, or CTO) must approve any study recruiting minors.
  • Direct-message recruitment of minors outside approved channels (Slack, linkscape.app email) is prohibited.

3.4IRB routing

LinkScape cannot issue an IRB approval. Studies that would require IRB review at a university (e.g., clinical, vulnerable-population, or high-risk behavioral research) must be run under a partner institution's IRB, with LinkScape as collaborator. When in doubt, the Ethical Review Board (E-05) decides whether external IRB routing is required before data collection begins.

4Data Use Agreements & Dataset Provenance

Research is only as ethical as the data behind it. Every dataset used in LinkScape research must have a clear, lawful provenance and a license that permits our use.

4.1Dataset intake checklist

FieldRequirement
SourceOrigin URL, institution, or collection method
LicenseExplicit license and permitted uses (research/commercial)
Personal data?Yes/No; if yes, E-02 Data Privacy applies
Consent basisHow the underlying data was lawfully obtained
Attribution requiredCitation string to include in publications
RetentionStorage location and deletion date

4.2Data use agreements (DUAs)

  • Third-party datasets with restricted terms require a written DUA before download to the compute fleet.
  • DUAs are signed on behalf of LinkScape by the Founder or CTO; because LinkScape is not a legal entity, agreements requiring a legal signatory route through Hack Club / HCB per G-06.
  • Scraped or user-generated data must respect the source's terms of service and robots directives; no circumvention of access controls.
  • Datasets may not be redistributed unless their license explicitly permits it.

Any DUA that imposes legal or financial obligations on LinkScape must be reviewed via Hack Club Bank per G-06 HCB Fiscal Sponsorship before signature. LinkScape members cannot bind the 501(c)(3) on their own authority.

5Responsible Use of the Compute Fleet

The 80× NVIDIA H100 fleet is LinkScape's most significant resource, and research is its primary consumer. Ethical research is also resource-responsible research.

  • Compute allocation for research follows F-03 Compute Resource Usage; access is tiered and logged.
  • Researchers estimate and record the compute budget of large training runs and prefer efficient methods (fine-tuning, LoRA, smaller models) where they suffice.
  • Environmental impact of large runs is tracked per E-07 Environmental Impact.
  • No research workload may run on the fleet without an approved project (see P-02 Project Approval).

6Publication & Authorship

6.1Authorship criteria

Authorship reflects substantial intellectual contribution — study design, implementation, analysis, or writing. Contributing compute or funding alone does not warrant authorship; such contributions are acknowledged. Every listed author must approve the final manuscript and agree to be accountable for it.

  • Author order is agreed among contributors before submission and documented in the project's Notion page.
  • Student and minor contributors receive fair credit and are never omitted for seniority reasons.
  • Honorary or ghost authorship is prohibited.
  • Disputes are mediated by the Founder as research owner.

6.2Publication standards

  • Default to open access and open-source release of code and, where licensing permits, data and weights (LinkScape defaults to open-source).
  • Disclose funding and compute sources, including that the work is a fiscally sponsored project of Hack Club.
  • Include a limitations and broader-impacts statement consistent with venue norms (e.g., ACL responsible-research checklist) and E-01.
  • Release accompanying model documentation using the E-04 Model Card Template.

6.3Pre-submission review

  1. Author confirms integrity checklist (Section 2) and consent/data checks (Sections 3–4).
  2. For work with elevated ethical risk, obtain Ethical Review Board sign-off (E-05).
  3. Founder approves external submission under the LinkScape name.
  4. Attach the Model Card (E-04) and confirm IP terms with PP-02 IP Assignment before public release.

7Review & Ownership

The Founder owns this policy and reviews it at least annually, or after any research-integrity incident. Proposed changes with constitutional weight are decided by the Founder and Co-founders collectively. Questions about research ethics should go to ceo@linkscape.app; safety-related concerns escalate to the CFO (cfo@linkscape.app) per E-08 AI Incident Response Protocol.

Approval and Adoption

Adopted upon signature by the officers below. Pending ratification at Sprint 0 (see LS-REG-001).

Document Maintenance Log

VersionDateAuthorChanges
1.0July 23, 2026LinkScape LeadershipInitial release

LinkScape runs as a fiscally sponsored project of The Hack Foundation dba Hack Club, a 501(c)(3) nonprofit. Hack Club holds the charitable status and every dollar moves through Hack Club Bank.

This page and the PDF are both generated from the source document in LinkScape's organizational pack. Cover furniture and approval blocks are omitted here; the text of the policy is reproduced in full.