Skip to content
Apart Research
Apart Research
Contact us
Sprint projectJun 21, 2026Mumbai, India

Blindfolded Governance: Mapping Economic Dark Output from AI in Asia

Mahi H Shah · Team Dark Output in Asia

Submitted to Global South AI Safety Hackathon. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: Blindfolded Governance: Mapping Economic Dark Output from AI in Asia

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

AI is making informal and self-employed workers across Asia more productive, yet this productivity gain often leaves no trace in GDP, tax records, or labour statistics, a phenomenon recently termed "Dark Output." This work does not attempt to measure Dark Output directly, an exercise that remains infeasible given current data; instead, it constructs the Asia Dark Output Index (ADOI), a composite, pre-emptive proxy for where AI-driven economic activity is most likely to escape formal measurement. ADOI combines four publicly available pillars: informal employment, self-employment, AI task exposure, and digital infrastructure, normalised and weighted across twelve Asian economies.

India, Indonesia, Pakistan, and Cambodia emerge as highest-risk. The novel contribution of this work is the first cross-country index to combine informal-economy structure with AI task-exposure data specifically to anticipate measurement blind spots, rather than to assess AI readiness or labour displacement. South and Southeast Asia is the natural focus given the region's unusually high informal employment share alongside rapidly expanding digital infrastructure and AI exposure. The result is a tool for governance under uncertainty: identifying where policymakers are most likely to be acting on an incomplete picture of their own economies.

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 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.

  1. The submission explores an underexamined part of AI economics and is a strong fit for the hackathon. The writing is strong and the added figures are a nice touch.

    The methodology is transparently explained and mostly appropriate. The main thing holding back the index is its aggregation. The paper defines Dark Output as a conjunction (activity that is both AI-exposed and institutionally invisible), but the additive weighted sum is fully compensatory, so a country can rank high on invisibility alone even with near-zero AI uptake.

    Two consequences follow: Informal employment and self-employment (60% of the weight combined) measure a highly overlapping population in these economies, so the invisibility side is in effect double-counted. As AI task exposure correlates negatively with informality across the sample, the two sides partly cancel inside the sum, leaving the ranking close to a pure informality index with the AI pillars doing little work.

    Consider a nested structure: combine the invisibility proxies additively, given that they are substitutable measures of one construct. Then combine that invisibility factor multiplicatively with an AI-uptake factor, ideally gating exposure by connectivity. A geometric mean across the two dimensions would also honour the paper's own "necessary but not sufficient" logic, which the current additive form contradicts.

    Read full reviewShow less

  2. Strong framing. The notebook also reproduces cleanly.

    To improve:

    1. Add sensitivity analysis. Vary the weights, compare min-max and rank-based normalization, and show whether the top cluster still holds.

    2. Clarify data provenance. E.g. compare sources of data, state which you use and why, since this affects the scores.

    3. Address China exclusion directly. Leaving China out weakens the Asia framing.

  3. This is a thoughtful and original project that frames an important governance blind spot: AI-enabled productivity gains in informal and self-employed work may remain invisible to official economic measurement. The proposed index is a useful and creative proxy, and the report is especially strong in explaining why direct measurement is infeasible and why a pre-emptive governance tool could still matter. The execution is solid for a weekend hackathon, with transparent indicators, weighting, normalization, and a reproducible dataset, though the weighting choices remain mostly theoretical and would need validation or sensitivity analysis in future work. The link to AI safety is somewhat indirect compared with projects focused on model behavior, misuse, or institutional safeguards, but the contribution is still relevant to governance capacity and Global South policy design. Overall, the report is clear, well structured, and persuasive, with a concise narrative that avoids overclaiming.

    Read full reviewShow less

  4. Like the core move here: you are not asking how ready a country is for AI, you are asking where AI output slips past the official numbers. That reframing is the contribution, and you built the index cleanly from public data and said upfront there is no ground truth. The soft spot is the weighting. The 30/30/20/20 split carries the whole ranking and never gets tested, so I cannot tell if the order is real or an artifact. Fix that first: vary the weights, show how much the ranking moves. Then bring China into the sample, since its absence probably undersells the whole point, and find one external signal to validate against, even a rough shadow-economy estimate. Honest and well-scoped for a weekend. Build the sensitivity check and this gets a lot more persuasive.

Cite this project

@misc{shah2026blindfolded,
  title = {{Blindfolded Governance: Mapping Economic Dark Output from AI in Asia}},
  author = {Mahi H Shah},
  year = {2026},
  month = jun,
  note = {Submitted to Global South AI Safety Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/blindfolded-governance-mapping-economic-dark-output-from-ai-in-asia-tjx3}},
  url = {https://apartresearch.com/sprints/projects/blindfolded-governance-mapping-economic-dark-output-from-ai-in-asia-tjx3}
}

More from Global South AI Safety Hackathon

All 216 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026