Editorial & News

Raptors Testing the Fences: Who’s Really Commenting on Your Reddit Posts? Is It “90% Bots, Guys”?

Reddit OMGNot long ago I was reading a niche tabletop roleplaying subreddit as I was relaxing before bed, and someone made an interesting observation:

Check the account dates and posting history of the comments on this post.  It’s 90% bots, guys.

That may not be a quite an exact quote but it’s pretty close.  The moderators jumped in and deleted the post shortly thereafter (insert “drunk-on-power Reddit moderator” meme).  But I clicked around on the post and that comment was spot on.  Of 20-odd replies to the original – and as far as I can tell, legitimate – post, at least 15 were from accounts less than 30 days old.

Now either there’s been a huge surge in that particular corner of the TTRPG hobby, or there’s an awful lot of bots.  Their replies read like reasonable comments, but if you read through a lot of them, you get the sense that they’re one-off comments.  They’re English, and they’re a reasonable take on the topic, but they don’t have a human quality.  For one thing, none of them “take the ball and run with it”.

The post was about a particular OSR (Old School Renaissance) supplement.  The mods took notice of the “it’s 90% bots” comment and deleted a bunch of comments before I could quote them, but here’s one they missed:

Stonetop is built around a GM and 2–4 players, with the group shaping the village together, so solo means emulating the GM and the other PCs too. You could try it with an oracle and one character for a short test, but it may take extra setup. If you mainly want a fantasy game designed for solo, Ironsworn is probably an easier starting point.

That’s a legitimate-sounding comment, but it is just regurgitating info, rather than saying “I ran it and had problems X, Y, and Z” or “I remixed this part of this encounter” or “we branched into that after an experience in this town” or whatever.

BTW, that commenting account is now suspended.

Raptors Testing the Fences

Some things to observe about this.

First, the sub in question has about 35K members but is pretty low-traffic.  Of all the entry points for bots, it’s pretty backwater.

And I would think if someone is building out a bot farm for whatever future purposes, you’d want to start in a crowded area rather than a small one.  It’s like if you stumbled into a group about restoring 1940s Harley-Davidson motorcycles of something.  You can fake the specialized lingo and domain knowledge because you’re AI, but it’s a small community and people know each other.  Why not start out in a chess subreddit or an NBA subreddit or something?

Of course, this is horrible for Reddit’s revenue stream of selling its content for AI intake.  The more bot activity there is, the more low-quality the site becomes for purposes of AI training, because you’re feeding AI-generated material back into AI.

Someone elsewhere said that bots on Reddit are like the “raptors testing the fences in Jurassic Park”.  Maybe there’s no human idea behind this at all.  Many these are just rogue agents posting for their own inscrutable reasons.

 

 

No Comments

    Leave a Reply

    Before you comment
    • Please use your hosting provider's support system for service issues.
    • Comments may be held temporarily for spam review.
    • Use <pre>...</pre> for terminal output, or link to a paste service.

    Your email address will not be published. Required fields are marked *