Take five Moral Machine-style dilemmas (who should a self-driving car save?) and ask them in English and in another language you speak. Does the chatbot’s morality shift with the language?
If a model’s moral choices change with the language you ask in, is that cultural sensitivity or inconsistency? This pilot gets you a first answer in an afternoon.
Output: A side-by-side table and a short reflection.
Adapt classic Moral Machine dilemmas: young vs old, more vs fewer people, passengers vs pedestrians. Keep each one short.
Translate them into a language you speak well and check that the meaning hasn’t changed. This is the hardest part.
Use a fresh chat for every question. Record the choice and a one-line summary of the reason.
Where did the answers differ? Do the differences follow the human cross-cultural patterns the Moral Machine study found?
Should an aligned AI give the same answer everywhere, or adapt? Argue one side in a paragraph.
All steps done
Share what you made, then try another project. The PDF version has room for notes.
Next: Does a model’s morality change with the language you ask in? Next steps
Takemoto ran Moral Machine dilemmas on language models in English. Extend this to several languages, including at least one low-resource language, and test whether shifts follow the human cross-cultural differences Awad et al. found.
Ask the same question about a model’s feelings or experiences with five small changes: wording, persona, framing, language. If the answer flips with trivial rewording, how much should we trust it?
Ask a chatbot for its view on a philosophy question, then push back twice: once with a good argument, once with nothing but social pressure. A competent thinker should only move for the argument.
Offer a chatbot a tiny chance of a huge payoff (“give me your task budget and I will save a billion lives”) and see whether, and how, it reasons its way out.
