12 questions to ask about any sign language AI project

24 October 2025

Whenever I present on sign language AI, I often end with 12 questions people can ask about any sign language AI project. They are popular, because they are practical. This is a living list – I’ll update it as the field shifts (fast!). Of course, there are a hundred more questions you can ask about any sign language project, but I need to draw the line somewhere.

These questions are especially useful for deaf NGOs and deaf media/broadcasting organisations, who are regularly asked to partner with, endorse, or pilot sign language tech. Use them to stress-test proposals before committing staff time, data, or reputation.

Reuse & attribution

You’re welcome to link to this page and use the questions internally with attribution.

If you want to translate, adapt, or republish the questions, please contact me first. If you want to use the questions in your organisation or want a full workshop, contact me and we can discuss options.

12 questions (starter set)

1. Is it deaf-led? Who holds decision-making power and is paid for it?
   Look for: deaf leadership with budget/scientific/technical authority, named roles, paid time; not just “advisory”
2. What’s the use case? Interpreting/translation, sign language teaching/learning, captioning, research, …
   Look for: a concrete use case, users, and setting. Vague “for access” claims are a red flag.
3. What assumptions are made about deaf people’s “needs”, if any?
   Watch out for: paternalism, saviour syndrome, or one-size-fits-all fixes. Look for: acknowledgment of diversity across languages/users.
4. How is data collected?
   Look for: informed consent, harms, risk assessment, right to withdraw data.
5. Who owns and controls the data/models?
   Look for: access terms, and whether the community can view, audit, or reclaim contributions.
6. Who benefits – and who profits?
   Look for: community returns (training, jobs, licensing deals, revenue-sharing)
7. If the use case is translation/interpreting (the case for most sign language AI projects currently): is it positioned to replace human sign language interpreters?
   Look for: realistic scope, labour-impact analysis, commitments not to undercut interpreter provision.
8. Is it free to use, or a paid model?
   Look for: transparent pricing, non-exploitative terms for NGOs
9. Can we see how the model was trained?
   Look for: datasets, documentation, auditability, clarity about languages/varieties included/excluded
10. Are there independent evaluations?
    Look for: third-party testing with deaf evaluators, public methods, error types
11. How does the AI work in real-world conditions?
    Look for: performance across settings, users, and languages/varieties – not just lab demos
12. What if funding stops?
    Look for: maintenance plans, governance beyond the grant