Maybe, maybe not but that design cannot measure that question. You also cannot measure the polarization question like that imo to be honest, because on what planet is, say, Fox news outside of the algorithmic public sphere? It orients itself to compete with Breitbart on Facebook.
-
-
Replying to @zeynep @chris_bail
I share the frustration the data we need is locked up in the big companies, and nobody is funding the kind of longitudinal design we need. But the keys aren't under the light. We can try, like this study, but the world isn't separated neatly for our surveys to work like this.
1 reply 1 retweet 38 likes -
Replying to @zeynep
I agree on data access, but would also point out that most of the evidence in support of the algorithmic hypothesis is highly anecdotal and non-causal. The best analysis I've yet seen is
@BrendanNyhan's recent work w/web tracking data that also finds little support in my reading.2 replies 0 retweets 9 likes -
And-- though also not the experimental evidence you and I might like see-- I think Kevin Munger's observational analysis of YouTube also finds little evidence:https://journals.sagepub.com/doi/abs/10.1177/1940161220964767 …
1 reply 0 retweets 5 likes -
And even though we should be highly skeptical of work authored by people inside social media companies, it is perhaps worth noting that Bakshy et al (in Science) found little evidence of this many years ago as well.
1 reply 0 retweets 1 like -
Replying to @chris_bail @BrendanNyhan
On the other hand, we have a stream of news reports of (quashed) research from inside the companies that do allege extensive contribution *and* crucially, I'd like to repeat that Fox News is part of the algorithmic public sphere (which makes measurement harder).
3 replies 0 retweets 15 likes -
Replying to @zeynep @BrendanNyhan
I am familiar with the cases you describe, I believe, but do we actually know that they radicalized people, or just that people joined more radical groups? Also, shouldn't we be concerned that media accounts of such cases are highly selective themselves?
1 reply 0 retweets 1 like -
I'm sure we both agree that this is why we need independent audits of algorithms, but in my reading of the (still preliminary) literature, we should not be surprised if they are much less powerful than we might think.
1 reply 0 retweets 2 likes -
Replying to @chris_bail @BrendanNyhan
A small effect per person at a time in a feedback cycle (person-algo-person-algo-person) over time in a transitional time (when all the big players are reacting to it by adjusting output) can quickly add up to a substantial societal transition, though.
2 replies 0 retweets 2 likes -
Replying to @zeynep @BrendanNyhan
As the author of a book about how fringe ideas become mainstream via emergent processes I agree ;) But it's not clear to me that this is driven by algorithms anymore than other forms of collective behavior
1 reply 0 retweets 5 likes
Hahaha yes. Congratulations, btw! But how can it not be part of the cycle? Incentives matter, and if these algorithms had no effect, the companies who measure everything to death wouldn't use them. It's not *just* driven by algorithms: yes, that is obviously and trivially true.
-
-
Replying to @zeynep @BrendanNyhan
For me, it's not trivial given the amount of public discussion dedicated to this and other issues that research indicates have minimal effects (not just algorithms but misinformation and echo chambers too).
1 reply 0 retweets 1 like -
Replying to @chris_bail @BrendanNyhan
And my frustration is that we’re not looking at the full feedback cycle, which can easily be societally transformative but hard to measure like this, and our research isn’t equipped to find those effects.
1 reply 0 retweets 3 likes - Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.