I was just reading this. The Medium article does leave some parts from the paper untouched though:

We find that negatively-evaluated users are more likely to down-vote others in the week following an evaluation, than in the week before it. In contrast, we observe no significant effect for the positively evaluated users.

Interesting that they didn't find an effect for the positively evaluated users.

We define a social networks around each post, a voting network, as illustrated in Figure 9. For a given post a, we generate a graph G = (V,E) with V being the set of users who voted on a. An edge (B,C) exists between voters B and C if B voted on C 30 days prior to when the post a was created. Edges are signed: positive for upvotes, negative for downvotes. We examine voting networks for posts which obtained at least 10 votes, and have at least one upvote and one downvote.

So on the left those who upvoted it and on the right those who downvoted it, and their underlying relations. Interesting graphs! mk or forwardslash, if you're bored...

edit: sorry for the double notifications, something went wrong on my side.