not necessarily - the data could be used to understand overall usage patterns and preferences in the target population as a whole - much in the same way as traditional surveys and focus groups. If indeed it is put into ML which then tries to match the data with the actual participants, then I agree that that is misrepresentation and probably criminal.