Research Ethics & Integrity

1. Commitment

The lab expects every project to be worth doing, honestly done and safe to release. This policy sets the standard that supervisors, student researchers and collaborators agree to when they join a problem.

2. Research integrity

• Report methods, data and results accurately, including negative and inconclusive results.
• Do not fabricate, falsify or selectively omit data.
• Cite prior work and give credit for ideas that are not your own.
• Keep enough records for an independent researcher to reproduce the work.
• Disclose limitations and the conditions under which results may not hold.

3. Human participants

Any study involving people must have a documented protocol, informed consent and, where required, approval from an ethics board before data collection begins. Participants may withdraw without penalty, and their data must be handled as described in the consent materials.

4. Data ethics

Minimisation

Collect only the data the research question needs.

De-identification

Remove or aggregate identifiers before any public release.

Licensing

Release data only where the licence and consent allow it.

Sensitive data

Do not place personal, health or otherwise sensitive data in public repositories.

5. AI safety and dual use

Work on capable AI systems can be misused. Projects that could enable harm — for example, attacks on deployed systems, surveillance or autonomous weapons — require an explicit dual-use review before release. The review asks who could use the result, what safeguards reduce risk, and whether a staged or restricted release is appropriate.

The lab will not publish material whose primary foreseeable use is to cause harm, and will pause a project where the risks cannot be adequately mitigated.

6. Authorship and credit

Authorship reflects a substantial contribution to conception, execution or analysis, and every author approves the final version. Supervisors do not take authorship for work they did not contribute to, and contributors who do not meet the authorship threshold are credited in acknowledgements.

7. Conflicts of interest

Researchers disclose funding, employment, advisory roles and other interests that could reasonably be seen to influence a result. Disclosures appear with the publication or are reported to the research lead before work begins.

8. Research misconduct

Fabrication, falsification, plagiarism, undisclosed conflicts, misuse of participant data and retaliation against someone who reports a concern are treated as misconduct. A finding may lead to correction, retraction, removal from the lab and notification of partners or institutions.

9. Reporting a concern

Concerns can be raised confidentially with the research lead or through [Placeholder: lab governance contact email]. Reports are handled by people who are not the subject of the concern, and good-faith reporters are protected from retaliation.

10. Review and approval

Before data collection, each project records its ethics review, its data-management plan and, where relevant, its dual-use assessment. Approval is required from the named supervisor and the research integrity lead.