3 ms·
Perhaps the model simply isn't rich enough. Right now, we model posts as either belonging to the good bucket or the bad bucket, and do various things (upvoting,
by sfink 5y ago
Perhaps the model simply isn't rich enough. Right now, we model posts as either belonging to the good bucket or the bad bucket, and do various things (upvoting, downvoting, AI, ...) to guess the "right" bucket. Maybe it shouldn't be surprising that a system based on polarization results in polarization?
Which is not to say that just adding complexity is the right fix. Sure, you can add dislike + disagree + ijusthateyourguts buttons, or force votes to be justified (funny/incorrect/disagree/...), or whatever. But it's not obvious what's worth the overhead (UX-wise), and if it ends up boiling down to just finding more nuanced reasons to toss things into one bucket or another (or even a single score), I'm not sure if we're any further ahead.
Perhaps clustering is a better model than polarized buckets? As in, this comment belongs to the set of things that left-leaning people seem to support. This other comment belongs to the conspiracy theory #1 cluster. (The system would likely not know or care what the clusters mean.) If you choose to engage with the idiots who love cluster A, you can do so, and if you want their noise out of your feed then you can do that too.
You'd still need to decide what input signals will enable this -- you could do it with just upvotes and correlating users' activity across a site, but you would need to evaluate that in terms of what best supports the model (clusters instead of buckets).
And clusters aren't the end-all be-all either. A racist, misogynistic comment that also says something that group X strongly agrees with isn't going to be fully described by being in the "jerk comment" cluster. But we have to deal with a low-dimensional approximation in order for any mechanism to be worthwhile.
- rspeele 5y agoPerhaps I misunderstand your idea, but this clustering sounds like exactly the thing to drive polarization. Put me in my comfortable bubble where I only see content and comments from like-minded people, and I'll be increasingly unable to understand where people outside the bubble are coming from. The only thing I see about those people is ragebait about how awful and stupid they're being this week, written by other people I agree with. Even if I start on the edge of a cluster or in the overlap between two clusters, the feed will likely drive me towards one or the other. That's how I feel with YouTube's aggressive recommendation system -- I watch a funny Bill Burr comedy clip, and YouTube knows there is some overlap between "Bill Burr fans" and "men's rights advocates", and now I get a bunch of recommendations for Jordan Peterson stuff and if I click one of those then its next recommendations want to take me down the MRA/incel rabbit hole. It's a radicalization pipeline where I touch some innocent surface-level stuff and YouTube goes, "hmm, well, of the people who liked that, our most highly-engaged users who click through videos all day long REALLY like THIS" and wants me to become that person.
- MrStonedOne 5y ago> I watch a funny Bill Burr comedy clip, and YouTube knows there is some overlap between "Bill Burr fans" and "men's rights advocates", and now I get a bunch of recommendations for Jordan Peterson stuff and if I click one of those then its next recommendations want to take me down the MRA/incel rabbit hole. I do not think it is productive to a healthly discussion to link wanting to end discrimination against men with incels or Jordan Peterson
- rspeele 5y agoWell, that might be an example of how my perception is swayed by these recommendations. There are probably plenty of people out there making YouTube videos purely advocating for equal treatment in things like the draft, or allocation of child custody in divorce, or whatever. But if I were to watch a few of their videos I suspect I would be swamped with recommendations for other videos which are more in line with what I was referring to. Stuff made by and for men who are bitter and resentful of women and spend a lot of time online seething about it. I mentally lump it all together because it all appears together. That's admittedly unfair.
- MrStonedOne 5y ago>I mentally lump it all together because it all appears together. That's admittedly unfair. and another example of how such algorithms can influence collective thought. I'm sure there are people who do the same with other equality measures as more divisive content of the same vein like "you're a fucking white male" is gonna be more likely to show up until the viewer links it with social justice entirely.
- fpgaminer 5y agoSomewhat related to this, I recently watched a video essay on VR Chat and one of the takeaways was how VR Chat is much more like the earlier internet (think around the dawn of Web 2.0 and before). There's no recommendation systems or an algorithm partitioning people into isolated spaces. It's just a bunch of random strangers browsing through uncurated rooms. And, as the author argues, that lends itself to a much healthier environment where people are exposed to and more understanding of those outside of what would otherwise be their bubble. I definitely feel some nostalgia for those days. To be clear, it's not like VR Chat or even the early internet were completely bubbleless. But without an algorithm dividing people, it was just the path of least resistance to ... explore. Click on links and see where they go. Would you end up on some neo-nazi forum sometimes? Sure. But you were just as likely to end up on a Pokemon fan page. It was far more random and diverse. As you point out, that's different today with these algorithms driving people directly towards radical content.