I actually gave a talk about automating influence on social media with what effectively amounted to chatbots, as a part of the 2016 election. Many of these factors were used, but also a few other interesting elements this article didn’t mention:

1. Geographic locales have linguistic norms that silently help in-group association. Adjusting rhetoric to these terms make people more pliable.

2. Context-aware sentiment is more impactful than general sentiment analysis. Someone may not have been a Trump supporter, but they didn’t like Hillary. This became another fulcrum.

3. Social leaders had the biggest impact, but social leaders can be made. This was leveraging a thought experiment conducted by DARPA, published in a paper titled, “Containment Control for a Social Network with State-Dependent Connectivity“ (https://arxiv.org/pdf/1402.5644.pdf)

These and a few other factors lead to my prediction of polling trajectories delivering a win for Trump in 2016 (the lecture was in August).

Is your talk online anywhere? I'm in the middle of LikeWar and it's the first I've really dug into mass influence/propaganda. It's infinitely interesting

It is, but the sound is _bad_: https://m.youtube.com/watch?v=iIIemMC4hrk