Trying to mess with the Washington Post’s AI

The Washington Post recently instituted an annoying new comments system, for comments on articles. It used to just show all the comments, and give readers a chance to thumb-up the ones they liked. The reader could then choose to see comments by time of posting, or by number of thumbs. Easy enough.

The new system often offers an annoying (in my opinion) writing prompt, for comments. (I find this annoying because I’m able to come up with a comment on my own; I don’t need a prompt.) The new system also allows readers to choose from four reactions, all positive; “clarifying,” “new,” “thoughtful,” and one other I don’t recall. I don’t find that annoying, really, but it’s no improvement over the previous system. Finally, the new system offers an AI-generated summary of the comments, after about 25 are posted. This summary includes a disclaimer saying that you should still read the comments, because it might be inaccurate . . .

I didn’t ask for the AI; I find it distracting and annoying; I don’t find it any better than a human, and I’d prefer that a human have the job; et cetera; so I tried to crush it recently.

I read a lot of soccer articles on the Post. The frequent commenters on those articles are a relatively small community who in many cases have been commenting for years, and in some cases even know each other outside the comments section. So I thought they’d be a good group to try to enlist to manipulate the AI summary.

(They’re also a group which is more friendly than most commenters, in my opinion, although that’s changing a bit with articles about the District’s pro men’s team, DC United, because DCU’s owners are not investing in it, and readers are getting bitter.)

Okay, so I picked a recent article which was about a roster choice for the U.S. men’s national team. It was about a certain player — Matko Miljevic — who was invited to join the next camp. (The article was written by Steven Goff, the longtime Post soccer writer who deserves better than my manipulation, sorry Mr. Goff, love you man, but AI must be fought.)

In an early comment to that article, I . . . or maybe I shouldn’t admit it was me. Okay then, in that article, SOME COMMENTER asked all the other commenters to write something about Jeff Agoos, instead of Matko Miljevic.

Jeff Agoos is a former pro soccer player who played for many years for the US national team and also for DC United. I picked him because he is very well-known in soccer circles and I THINK has a very good reputation as a good player and decent guy. (As George Orwell wrote about Gandhi, “what a pleasant smell he has left behind him.”) I’m also pretty sure he still lives in this area.

My fellow commenters obliged. Most of the comments completely ignore the thrust of the article, and instead just praise Jeff Agoos.

So that was cool! Unfortunately, there were only 18 comments total, and the AI summary doesn’t kick in until there are more, maybe 25 or 20.

Here’s the article llink btw: https://www.washingtonpost.com/sports/2025/01/06/usmnt-roster-matko-miljevic/

BUT on a subsequent article — and this happened more organically — the AI was indeed taken for a ride because it parroted an idea which was raised as a joke by commenters; this is a frequent joke made by this community, the idea that DC United will definitely win the league championship in the coming season. This is clearly a joke because the article is all about how DCU is putting together a budget roster. The AI summary:

Well done lads.