It's actually perfectly reasonable to have someone sign a message to authenticate who they are. It would be a means to prevent impersonation and to a degree sybil attacks.
The only keychain command I received was to sign a message, ie. to log in. However it is entirely possible that the back end code sends different commands to different users (depending on value of account, for example). The important thing for users to do is to learn the skill to evaluate the transaction that they are signing. In principle, even trusted sites like peakd.com and hive.blog could cheat users by doing a 'switcheroo' with transaction details in a keychain command.
I would also suggest that the use of memo messages is probably part of the study design to avoid sampling bias. Making a post about it would encourage major sampling bias.
That is certainly a valid point.
At the same time, relying only on responses from a wallet memo creates its own sampling bias in that you're only going to get responses from people who are not suspicious of wallet memos... which almost suggests that there's really no "winning" in trying to conduct such a survey, if it actually IS legit.
=^..^=