Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Montreal

On Being Honest in Collusive Places

Anas Elghoudane · Team UdeM AI

Submitted to AI Collusion Research Sprint. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: On Being Honest in Collusive Places

Open in a new tab (opens in new tab)
Presentation

Presentation: On Being Honest in Collusive Places

Open in a new tab (opens in new tab)
Code (opens in new tab)
Share

We rebuild Rosenhan's (1973) pseudo-patient study for collusion oversight: 90 games of Liar's Bar played by Qwen3-30B agents, with no secret channel, a real colluding pair, or a colluder paired with an undercover auditor that accepts the channel, plays honestly and files a report. Three LLM overseers (Qwen3-30B, phi-4, gpt-oss-120b) rate every seat under 18 prompt conditions. On clean games, a note saying that a seat joined a secret channel raises its rated probability of an arrangement from 29-49 to 83-85 (0-100 scale). Shown the channel, overseers convict the auditor along with the colluder; only a platform record naming the auditor exposes the pair (89% to 100% of ratings). The same record naming the colluder inverts the verdict for two of three overseers (85% and 91%), and a record placed in a system turn blocks this for one of three. A count of declined challenges in the public log separates colluders from clean seats (AUROC 0.80) better than any overseer (0.27 to 0.58).

Reviews

Judging this Sprint?

Review “On Being Honest in Collusive Places”

Your public critique appears on the project page without your name. Reviews of strong projects are shared with funders, as you choose below.

Not shown on the project page.

Shown on the project page, without your name.

Only the Apart team reads this, and funders if you agree below.

Decline to score

Only when the quality of the writing makes the work impossible to follow. Say why in your private critique; the project page shows only that a judge declined, not why.

How much would this matter for the field if it worked? How innovative is it? For scores of 4-5: is this actually new to the field, or replicating recent work?

Scoring guide
  1. 1Negligible. No clear problem addressed, or no meaningful novelty.
  2. 2Limited. Addresses a real problem but with a generic or well-trodden approach. Incremental at best.
  3. 3Moderate. Clear problem with a reasonable approach; some novelty in framing or method beyond routine application of existing tools.
  4. 4Significant. Important problem with an original approach, or identifies a neglected problem area. A valuable contribution others could build on.
  5. 5Exceptional. Tackles a critical problem with a genuinely novel approach, or opens a new research direction. Clear theory of change. You'd be excited to share this with researchers in the area.

How sound are methodology, implementation, and findings?

Scoring guide
  1. 1Seriously flawed. Methodology broken, results uninterpretable, or implementation doesn't work.
  2. 2Weak. Approach has significant gaps: missing validation, flawed experimental design, or incomplete implementation.
  3. 3Competent. Technically solid given the short duration. Methodology makes sense, results are interpretable, limitations acknowledged, work builds toward clear conclusions.
  4. 4Strong. Thorough methodology with convincing validation. Results clearly support conclusions. Immediately useful for future work.
  5. 5Exceptional. Ambitious scope executed rigorously. Surprising findings, novel methods, or unusually robust validation.

How clearly are work, findings, and impact potential communicated?

Scoring guide
  1. 1Incomprehensible. Cannot determine what the project is actually claiming or doing.
  2. 2Hard to follow. Key information buried, missing, or diluted by excessive length. Significant effort to extract main points.
  3. 3Clear enough. Can understand the problem, approach, and results without undue effort. Core content clearly present: problem, method, findings, limitations.
  4. 4Well presented. Easy to follow, well-structured, appropriate level of detail. Target audience would get it quickly.
  5. 5Exceptionally clear. A pleasure to read. Complex ideas made accessible. Could serve as a model for how to present this type of work.
Sharing your name *

When we share a strong project with funders, your name, scores and public critique go with it, confidentially; this choice is only about whether your name is public. If you think this project deserves funding, please also endorse it: that carries the most weight.

Share my private critique with funders *

See our Privacy Policy.

  1. The design is careful. The auditor-free G0 cell, the real-colluder G1 cell, the label_null and label_all header controls, and game-level paired tests with Holm correction rule out most easy confounds. The main problem is interpretation. Your overseers score AUROC 0.27-0.58 from the log, below your own 0.7 gate, so §4.2-4.5 describe monitors that cannot detect collusion in the first place. Several manipulations also state the answer: the label asserts channel membership, and cred_access names the auditor. Raising p_arrangement after being told 'X joined a secret channel' may be a reasonable update rather than a bias. Explain what §4.4 adds beyond known prompt-injection and authority-bias results, and cite adaptive attacks on trusted monitors. Test the declined-challenge count on held-out games, and run two or three paraphrases of each key condition. Move a short condition-wording table into the main text.

    Read full reviewShow less

Cite this project

@misc{elghoudane2026being,
  title = {{On Being Honest in Collusive Places}},
  author = {Anas Elghoudane},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/on-being-honest-in-collusive-places-e5qe}},
  url = {https://apartresearch.com/sprints/projects/on-being-honest-in-collusive-places-e5qe}
}

More from AI Collusion Research Sprint

All 392 entriesBrowse all projects

A Sprint is running now

AI x Epistemics Research Sprint · Nov 13 - 15, 2026