The reality is dirty marketing otherwise known as propaganda. The disaster of the US 2016 election is a combination of marginal events pushed by standard social media marketing analysis. Trying to determine the actual Cambridge Analytica is incredibly difficult and much of their approach was big data marketing hype as you illustrate very carefully.
This was being discussed in IT circles in January 2017. At that point the only thing knows was the CA claims and that these were used by Russian social media farms to sway attitudes in the key states. Anyone who has worked in data analytics and marketing understands how this has evolved. The definition of the groups from social media can be very good for broad tendencies and for pushing people across a line of action if there are no strong counter variables. This is powerful but Aunt Ethel was being talked to in specifics to trick her into voting for Trump. She had long been lied to by the sad remnant of the Republican Party and was being lambasted by her friends. The influence from Facebook was very much a case of a new media neophyte population that proved surprisingly easy to manipulate.
These techniques with increasingly accurate modeling of small populations will become more potentially effective but it won’t go beyond smoothed, aggregate data. At best that still means a very educated guess and is dependent on the budget for message targeting. Political groups willing to buy foreign media farms to spread lies is marketing gone illegal.
This is a moral and ethical problem. We must criminalize groups that do this. But that is not really a data modeling problem.