Skip to content
Apart Research
Apart Research
Contact us
Sprint projectAug 17, 2026Lübeck, Germany

Different Models, Different Nuisances: Counterfactual Auditing of AI Preference Elicitation

Kishore Kumar Mariappan · Team PREF-CAL

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

Read the report

Report: Different Models, Different Nuisances: Counterfactual Auditing of AI Preference Elicitation

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

PREF-CAL is a prospective counterfactual audit for AI preference elicitation. Rather than interpreting repeated choices as preferences immediately, it first tests whether the apparent semantic direction survives counterfactual changes to irrelevant interface features. In a frozen GPT-OSS-120B run, semantic/physical-swap consistency was 7/48 after three complete factorial blocks; even perfect consistency on all remaining contrasts could reach only 119/160 = 74.375%, below the preregistered 75% qualification threshold. The deterministic futility rule therefore stopped the experiment before downstream transport. An interrupted Llama-3.3-70B matched slice exhibited a different nuisance tendency, suggesting that measurement artifacts can be system-specific. The contribution is an executable qualification rule that permits explicit non-identification rather than attaching preference semantics to an unvalidated response regularity.

Preference elicitation can produce highly selective-looking behavior even when arbitrary interface features control the response; PREF-CAL makes measurement validity and principled abstention prerequisites for welfare-relevant interpretation.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

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

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

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.

  1. I enjoyed a lot this project, and I appreciate that you released as code the frozen designs, raw log, etc. Pretty much all that is needed to reproduce your findings. I also really like the discipline of trying to break the measurement before interpreting its output as a preference. Finding that “the instrument failed” is a super useful result, especially when the same prompt format fails in different ways across models.

    Great job!

Cite this project

@misc{mariappan2026different,
  title = {{Different Models, Different Nuisances: Counterfactual Auditing of AI Preference Elicitation}},
  author = {Kishore Kumar Mariappan},
  year = {2026},
  month = aug,
  note = {Submitted to Digital Minds Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/different-models-different-nuisances-counterfactual-auditing-of-ai-preference-elicitation-ky3v}},
  url = {https://apartresearch.com/sprints/projects/different-models-different-nuisances-counterfactual-auditing-of-ai-preference-elicitation-ky3v}
}

More from Digital Minds Research Sprint

All 233 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026