Ars Technica: I was frankly surprised at the ineffectiveness of the various intervention strategies you tested. But it does seem to explain the Bluesky conundrum. Bluesky has no algorithm, for example, yet the same dynamics still seem to emerge. I think Bluesky's founders genuinely want to avoid those dysfunctional issues, but they might not succeed, based on this paper. Why are such interventions so ineffective?
Petter Törnberg: We've been discussing whether these things are due to the platforms doing evil things with algorithms or whether we as users are choosing that we want a bad environment. What we're saying is that it doesn't have to be either of those. This is often the unintended outcomes from interactions based on underlying rules. It's not necessarily because the platforms are evil; it's not necessarily because people want to be in toxic, horrible environments. It just follows from the structure that we're providing.
We tested six different interventions. Google has been trying to make social media less toxic and recently released a newsfeed algorithm based on the content of the text. So that's one example. We're also trying to do more subtle interventions because often you can find a certain way of nudging the system so it switches over to healthier dynamics. Some of them have moderate or slightly positive effects on one of the attributes, but then they often have negative effects on another attribute, or they have no impact whatsoever.
I should say also that these are very extreme interventions in the sense that, if you depended on making money on your platform, you probably don't want to implement them because it probably makes it really boring to use. It's like showing the least influential users, the least retweeted messages on the platform. Even so, it doesn't really make a difference in changing the basic outcomes. What we take from that is that the mechanism producing these problematic outcomes is really robust and hard to resolve given the basic structure of these platforms.
Ars Technica: So how might one go about building a successful social network that doesn't have these problems?
Petter Törnberg: There are several directions where you could imagine going, but there's also the constraint of what is popular use. Think back to the early Internet, like ICQ. ICQ had this feature where you could just connect to a random person. I loved it when I was a kid. I would talk to random people all over the world. I was 12 in the countryside on a small island in Sweden, and I was talking to someone from Arizona, living a different life. I don't know how successful that would be these days, the Internet having become a lot less innocent than it was.