OpenAI publishes studies that classify aggregate ChatGPT conversations, but users cannot ask their own aggregate questions. I asked: ‘How many ChatGPT users report operating one or more diesel-powered machines?’ ChatGPT could provide neither a count nor a trend.
Please let users submit questions like this and receive the number of explicit reports, sample size, time window, comparison trend, classification method, and uncertainty. Suppress results when a query is narrow enough to risk identifying someone.
This would let people benefit from and scrutinize patterns built from the observations they collectively contribute.