Skip to content
Apart Research
Apart Research
Contact us
Sprint projectAug 17, 2026San Francisco

The Mirage of Model Introspection: Elicitation Effects on Self-Report Reliability

Elsie Piao

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

Read the report

Report: The Mirage of Model Introspection: Elicitation Effects on Self-Report Reliability

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

When models report internal experiences, does the model “look inside” first before answering? With Gemma 3 27B, we start with the same self-referential processing technique as Berg et al. (2025), and add on top a concept injection (Lindsey, 2026) to test whether the content of Berg-style self-reports can be modulated by manipulated model internal states. We then tested whether Berg-style self-referential processing can elicit greater ability to attend to concept injection, both in detection and in recognition.

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. Due to severe time constraints, this review may contain mistakes or oversights. For the same reason, it focuses on the paper’s key idea, not the detailed execution: The paper combines two techniques from prior research in an interesting way. The results are interesting; it is generally useful to see how model responses are affected by different ways of prompting. The paper is also written quite clearly. At he same time, since the paper combines two prior techniques in a somewhat predictable way, I dont think it has the highest degree of innovation.

  2. Defining newmeasures on self-report reliability

  3. Nice project! Would be good to see this extended to other models; it would also be interesting to try to test some of the alternative hypotheses you mentioned in the discussion section (e.g., whether the models become just more predisposed to answering affirmatively after the self-referential prefill -- this would be pretty cheap to test. )

Cite this project

@misc{piao2026mirage,
  title = {{The Mirage of Model Introspection: Elicitation Effects on Self-Report Reliability}},
  author = {Elsie Piao},
  year = {2026},
  month = aug,
  note = {Submitted to Digital Minds Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/the-mirage-of-model-introspection-elicitation-effects-on-selfreport-reliability-x9y3}},
  url = {https://apartresearch.com/sprints/projects/the-mirage-of-model-introspection-elicitation-effects-on-selfreport-reliability-x9y3}
}

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