Conflict of Interest

A scientific registry that maintains credibility has to be visibly honest about who decides what. This policy describes when moderators and endorsers must disclose conflicts and when they must recuse.

Why this matters

A registry is a small social system. The same people who develop agents are likely to know each other, share funding, co-author papers, and sit on each other's grant committees. That's normal in any field. What's not normal — and what destroys trust — is when those relationships shape moderation decisions invisibly.

The fix is disclosure + recusal, not banning relationships.

When you have a conflict

You have a conflict of interest with respect to a specific submission, flag, or appeal if any of the following are true:

Relationship	Conflict?
You are a co-author of the agent / tool	Yes — recuse
You are a co-author on a recent paper (≤3 years) with the submitter	Yes — recuse
You are at the same institution as the submitter	Yes — disclose; recuse if same lab or same direct supervisor
You are funded by the same grant as the submitter	Yes — recuse
You are the submitter's direct supervisor, advisor, or supervisee	Yes — recuse
You have a current commercial relationship with the submitter	Yes — recuse
You endorsed the submitter to join the registry	Disclose; recuse from moderating that author's submissions
You're in the same scientific field but don't know the submitter personally	No — no action
You disagree with the submitter's scientific positions in published work	No — but the disagreement itself isn't grounds to act

When in doubt, disclose. The cost of disclosure is low; the cost of an undisclosed conflict surfacing later is high.

How to recuse

In the moderation tool, the Recuse action:

1. Reassigns the item to the next available moderator who doesn't share the conflict.
2. Records a public note: "Moderator [ORCID] recused — conflict declared."
3. The specific nature of the conflict is recorded internally but only the fact of recusal is public, unless the conflict is itself part of the moderation decision (e.g., the conflict is being investigated).

You don't need approval to recuse. You should err toward recusing.

How to disclose (when not recusing)

For situations where you have a real but non-blocking conflict — same institution but different lab, professional acquaintance — you can:

1. Disclose the conflict in the moderation note ("disclosure: same institution, different department").
2. Proceed with the action.
3. The disclosure is part of the public audit log.

Disclosures don't reverse decisions on their own, but they make the decision auditable. If a pattern of disclosures-without-recusal looks like soft bias, that's grounds for review by UAH admins.

Special case: endorsers

Endorsement creates a long-term conflict. The registry tracks endorsement chains, and:

• Your public profile shows everyone you've endorsed.
• You may not moderate submissions, flags, or appeals from authors you've endorsed.
• This is enforced by the moderation tool: items from your endorsees auto-route to other moderators.

This is the same rule arXiv applies and for the same reason. Endorsement is supposed to be a low-friction trust signal, not a permanent advocacy relationship.

Special case: institutional affiliations

UAH is the registry's host. UAH-affiliated moderators are not blanket-recused from UAH submissions — that would make the registry unworkable — but:

• Same-lab and same-direct-supervisor relationships always trigger recusal.
• When a UAH-affiliated moderator approves a UAH submission, that's noted in the audit log (not a problem, just visible).
• Same-institution submissions that go to moderation should ideally route to a non-UAH moderator first; if one is unavailable, the UAH moderator notes the institutional disclosure.

Special case: commercial relationships

If you've received compensation (consulting, advisory, equity, paid speaking) from the submitter or their organization in the last 24 months, you must recuse. This includes:

• Active consulting agreements
• Equity in the submitter's company
• Paid advisory positions
• Honoraria for talks at the submitter's organization

Casual collaboration (uncompensated talks, joint open-source contributions, etc.) is disclosure-only.

When others have a conflict

If you spot what looks like a violation:

• A moderator with an undisclosed conflict — file a flag with reason other, naming the moderation action and the relationship.
• An endorser who's vouching for someone they shouldn't — same flag mechanism.
• A submitter who's not disclosing institutional or funding conflicts that should appear in their compliance block — flag the listing.

UAH admins review these within 7 days. Confirmed undisclosed conflicts are logged publicly. Repeated violations result in revocation of moderator or endorser status.

What this policy doesn't try to do

• It doesn't ban relationships. Scientific communities are small; some overlap is unavoidable.
• It doesn't try to define every edge case. The principle — disclose, recuse when in doubt — handles most situations.
• It doesn't replace the Code of Conduct or the Moderation Policy. Those cover how decisions are made; this covers who is qualified to make them.

When this policy applies

To every moderation action, endorsement, and appeal review. Public reads (browsing, citing, flagging) are unaffected — anyone can do those regardless of relationships.