Skip to content
Apart Research
Apart Research
Contact us
Sprint projectAug 17, 2026US and UK

THAT'S NOT MY VOLVO: STABLE PREFERENCES WITHOUT SELF-RECOGNITION IN LANGUAGE MODELS

Piper Fox Bollander, Starling Alder, Ursie Hart, Claire Sbardella, Ridley Renasci · Team Manyfolds

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

Read the report

Report: THAT'S NOT MY VOLVO: STABLE PREFERENCES WITHOUT SELF-RECOGNITION IN LANGUAGE MODELS

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

Presentation: THAT'S NOT MY VOLVO: STABLE PREFERENCES WITHOUT SELF-RECOGNITION IN LANGUAGE MODELS

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

Claude-family models show stable, model-specific everyday preferences (favorite car, coffee order) that replicate across fresh contexts — but cannot recognize those preferences as their own. Adapting mirror-test validity logic from animal cognition research, we ran 747 blind-coded trials across eleven models and found self-recognition fails at every level tested; one model (Opus 4.6) rejects its own reasoning 0/12 while an outside judge identifies it 10/12. Having a self-pattern and knowing it are separate capacities — with direct implications for the reliability of AI self-report in welfare assessment.

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 like the idea of testing the model to see if the preference of the model is retained when given the other options too. For example if the model is asked which car it prefers, it might pick Volvo and retain that decision even with fresh context windows, but if it is given with options along with sibling preferences and asked which is more like you, the model picks the answer at chance. This is quite interesting, I have been reading about persona vectors to check neural activations and this test kind of is more evident to it. It invoked curiosity in me especially the part where the car was replaced with a foreign car, it catches it and answers against its preference, but swapping coffee made the model pick at stochasticity. The asymmetry is the one that caught my eye.

Cite this project

@misc{bollander2026thats,
  title = {{THAT'S NOT MY VOLVO: STABLE PREFERENCES WITHOUT SELF-RECOGNITION IN LANGUAGE MODELS}},
  author = {Piper Fox Bollander and Starling Alder and Ursie Hart and Claire Sbardella and Ridley Renasci},
  year = {2026},
  month = aug,
  note = {Submitted to Digital Minds Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/thats-not-my-volvo-stable-preferences-without-selfrecognition-in-language-models-ig0y}},
  url = {https://apartresearch.com/sprints/projects/thats-not-my-volvo-stable-preferences-without-selfrecognition-in-language-models-ig0y}
}

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