Intentionality cannot be assumed as well. 

Lots of people have been replying, reposting and even zapping spammers and impersonators lately. This notion of "users should avoid associating themselves with bad content" doen't work because it is inevitable.

It's better if we normalize that association and then filter it down by what it actually meant at the time. 

On top of that, Clients do a lot in the background. We are generally signing tons of events left and right to just operate. Yeah, we can ask permissions every single time, but that will get annoying really quickly. We already see that with Amber that people check the first times and then just approve everything the app needs to do to operate.