Why is it unethical to record what people say in public settings, like a cafe, transcribe it, and then throw away any identifying information? You could go even farther and only store the individual set of words spoken without any ordering, so it would be impossible to reconstruct the actual content of anyone's conversations.
Universities in certain countries have strict ethical standards that require consent and crediting informants. These standards were often inspired by working with indigenous informants, who were viewed as having been historically exploited in the collection of data, but they apply now to anyone eliciting data from anyone.
What is the practical harm that is actually being avoided here?
When people speak in public, they assume that other people around them can overhear what they're saying. This is just gathering anonymous statistics about word use. It's hard to imagine how there's any real downside for anyone here.
In a perfect world, it wouldn't be. Personally, if I knew an infallible person was recording my private conversation and anonymising it to an acceptable degree, I wouldn't care. For example, by removing personal identifiers and aggregating the data across a topic such as "conversations at cafes in central Paris in 2026".
But the second there is doubt about security across any link in the chain from raw recording to anonymisation and storage, I would no longer consent.
The word set is enough to identify some people based on speech patterns or interests. The inclusion of other words in the set — cocaine, divorce, abortion, merger, lymphoma — would make many people uncomfortable.
