what I do is teach computers to do sentiment analysis on social media posts

He pays for Tweet rating tasks on Mechanical Turk. Thankfully, tasks requirements are invariably incoherent and applied with no rhyme or reason, so the data is useless.

we look for people we know are in their upstream -- people that they read posts from, but who themselves are less influential. (This uses the same social media graph built before.

All they're really doing is tracking publicly-available evidence of visibility to their targets. Their graph is almost "engagement", but it's not "influence".

The goal is to keep opinions we don't want fragmented and from coalescing in to a single voice for long enough

BTW, this part isn't shilling, it's jamming. Jamming could be used as tactical support for shills, but it isn't the act of shilling itself.

Does the reason you picked McDonalds over Burger King need to be anything more than just you talked about one more recently than the other?

You wonder why businesses hate their customers? They ask these marketing gurus to deliver the rotten apples they can reach.

technology enables normal people to distance other people in to being just data values, and thus enabling them to take actions they otherwise wouldn't.

Now he looks like a shill. But not for the organizations he pretends to be with.