Skip to content
Apart Research
Apart Research
Contact us
Sprint projectSep 14, 2026Porsgrunn, Norway

What Should We Check Next? Testing Ambiguity-Preserving Evidence Selection for AI Incident Investigation

Linda Thorstensen

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

Read the report

Report: What Should We Check Next? Testing Ambiguity-Preserving Evidence Selection for AI Incident Investigation

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

AI incidents can leave several causal explanations compatible with the evidence. I built and tested a small investigator that keeps those explanations open until the evidence rules them out, then selects the next check that should separate them most usefully. On a frozen synthetic holdout, it improved first-check evidence selection over neutral baselines while preserving unresolved and model-gap states. I then applied the workflow to unresolved questions from the OpenAI/Hugging Face incident and turned them into concrete cross-organizational evidence requests.

Reviews

Judging this Sprint?

Review “What Should We Check Next? Testing Ambiguity-Preserving Evidence Selection for AI Incident Investigation”

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 AI safety 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 AI safety 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. I like that this focuses on a very specific incident-response problem.

    The main limitation is that the evaluation is still highly synthetic and structured. I wld like to see the same approach tested on independently written, noisy incident cases and compared against more mature diagnosis methods, then put in front of actual incident responders.

  2. This paper tackles a critical and underexplored bottleneck in AI incident response: the overwhelming volume of multi-source telemetry that frequently hardens into premature, unsupported causal explanations before evidence can actually separate competing hypotheses. The author introduces a rigorous, auditable investigator implementation (AP-MINIMAX) that explicitly preserves ambiguity, marks cases as UNRESOLVED or MODEL_GAP when closure is unjustified, and uses worst-case residual ambiguity to select the next discriminating check. Across a rigorously frozen 400-case holdout evaluation, AP-MINIMAX demonstrated statistically significant improvements in realized candidate reduction per unit cost over neutral baselines across all five seeds, while avoiding unsupported closures entirely. By grounding the methodology in a practical, source-grounded handoff for the July 2026 OpenAI/Hugging Face incident, this work provides a valuable, modular blueprint for bringing disciplined active-diagnosis principles to frontier AI safety and incident investigations.

    Read full reviewShow less

Cite this project

@misc{thorstensen2026should,
  title = {{What Should We Check Next? Testing Ambiguity-Preserving Evidence Selection for AI Incident Investigation}},
  author = {Linda Thorstensen},
  year = {2026},
  month = sep,
  note = {Submitted to AI Incident Response Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/what-should-we-check-next-testing-ambiguitypreserving-evidence-selection-for-ai-incident-investigation-5nv4}},
  url = {https://apartresearch.com/sprints/projects/what-should-we-check-next-testing-ambiguitypreserving-evidence-selection-for-ai-incident-investigation-5nv4}
}

More from AI Incident Response Sprint

All 231 entriesBrowse all projects

Build something like this at the next Sprint

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