Andrea Jones-Rooy

Could we actually nail this down?

And I think it's a beautiful example of really using scientific methods to start to tease out some of these serious issues, particularly in this case, this endogeneity between user choices and algorithm recommendations.

Right, so George, you're hitting the nail on the head in terms of like, what does this mean for what we do about what we see in the world that is concerning to a lot of us around political radicalization and extreme and harmful views, is exactly this, is the debate is often, do we intervene with the algorithms?

Do we try to regulate the algorithm?

And I think one should, and there's plenty of other reasons to consider a thoughtful regulation in addition to political extremism,

But there's also the case that, yeah, it does, from this evidence, it does say that, like, a lot of this, the intervention should be at the user level.

Like, if you... And this is something that I've seen these particular researchers talk about before, and this is the first time they've really kind of tried to answer it, which is, like, is it that...

The pathways are there.

If you want to go down the pathways, it's sort of like this is a bad example or bad comparison, but it's sort of like if you want to try drugs and you live in a big city, you can probably find it.

And then once you find it, it's probably that much easier to find your next hit and your next hit, your next hit.

I feel like this seems to be what they're finding as far as the algorithm that it's like the algorithm is like neutral ish bends towards conservative.

Yes, there are these ideological echo chambers, but it's not a given that anyone who goes into here is going to end up somewhere extreme.

And I would imagine that many people, most people do not end up in these extreme ideological echo chambers.

I like to think that my algorithm is not an ideological echo chambers, but I also don't consume political news.

So it is very much like there's sort of a tipping point, but it does suggest the intervention is on the user side of things.

And I have inspected other data on like,

you know, where does like racial bias come from, racism, sexism, what are some things you can do to curb that sort of behavior?

And some of the ways they track that are, are you consuming racist content?

And it is the case that because these algorithms do give you more of what you're already watching, that if you're already interested in racist content, you're going to get fed more racist content.

But if you're not interested in racist content, there's nothing in here that suggests you're going to get nudged towards it until you're, because otherwise we would all be radicalized if you kind of think about it at this extreme.