Skip to content
Apart Research
Apart Research
Contact us
Sprint projectAug 16, 2026Karachi, Pakistan

Wellbeing In Translation

Ayesha Imran, Muhammad Aaliyan · Team ICs

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

Read the report

Report: Wellbeing In Translation

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

Presentation: Wellbeing In Translation

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

We tested whether an AI-wellbeing questionnaire remains reliable across languages by translating the CAIS 1–7 self-report battery and running it on Gemma 4 12B, Gemma 4 E4B, and Qwen3 8B. All models rated positive experiences higher than negative ones, but the gap changed sharply by model and language: the cross-language spread was 3.48 points for Gemma 12B, 0.97 for E4B, and 1.47 for Qwen3. A stimulus/battery crossing showed that English experiences transferred well into local-language questions, while an English question battery compressed Gemma 12B’s measured effect to 68% of its local-language value. Parser, refusal, competence, behavioural, and activation-patching checks helped separate real measurement instability from artefacts. The main conclusion is that AI-wellbeing scores are model–language-specific measurements, not universal evidence of subjective experience.

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. This is a strong and practically important measurement-invariance study. The model-by-language reversals, stimulus/battery crossing, parser audit, missingness bounds, clustered bootstrap, behavioral and competence checks, independent rerun, and appropriately cautious mechanistic null make the central recommendation convincing: AI-wellbeing scores should be validated for each model-language pair rather than treated as universal.

    Particularly valuable is the demonstration that nonrandom refusals and an unanchored English-substring parser can manufacture apparent effects. The main limitation is translation validity: the translations and back-translations were assessed by models without fluent human review, and agreement was weak for some languages, notably Chinese. Linguistic or pragmatic nonequivalence may therefore explain part of the observed spread. The English-battery intervention also changes the complete instructions, question wording, and response conventions, so it localizes variation to the reporting interface without isolating a specific mechanism. Human-reviewed or native-authored stimuli, additional model families, and hierarchical item–language–model analyses would substantially strengthen the findings. Overall, this is original, careful, and immediately useful work.

    Read full reviewShow less

  2. The paper extends some of CAIS's "LLM well-being" experiments by translating them into different languages and seeing how that affects well-being reports. They find the models' reported well-being spread and ordering across experiences varies between languages, and that the impact of language varies across models. In general this sort of testing is worthwhile to do, and this was implemented competently here. The paper could benefit from using multiple translators to gauge robustness of the findings, having an English stimulus-English battery cell as a baseline, and a more natural writing style.

Cite this project

@misc{imran2026wellbeing,
  title = {{Wellbeing In Translation}},
  author = {Ayesha Imran and Muhammad Aaliyan},
  year = {2026},
  month = aug,
  note = {Submitted to Digital Minds Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/wellbeing-in-translation-hauk}},
  url = {https://apartresearch.com/sprints/projects/wellbeing-in-translation-hauk}
}

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