Protest bots, or "bots of conviction" are canvassed by Mark Sample in a medium post and talk. link
A bot of conviction, or agit-bot, is coded to respond in precise ways to pre-chosen political triggers, or to retransmit historical or political data with timeliness.
The conviction and the expression are both human in design; the automation function achieves efficiency and scale, like any advertising. To this extent, bots of conviction are no more disconcerting to humans than the printing press or the call centre.
The ethical dimension to the creation and use of bots of conviction focuses on the problem of disclosure.
The paradox of automated protest is referenced by Sian Bayne in her discussion of the creation and deployment of a teacher bot to manage student enquiries in a MOOC. link
Postscript: research consistently shows that robots introduced to industrial processes do increase efficiency, and don't diminish human employment overall, but impact more heavily on unskilled workers. Bots are already used to monitor consumer forums online and are appearing in MOOCs; is this on the presumption that engaging in conversation is unskilled work in human terms?