It's become fashionable to decry "safe spaces" and echo chambers and to call instead for cognitive diversity. Instinctively, I agree: cognitive diversity is one counterweight to the tightly bounded knowledge and rationality that afflict us all.

In this context, then, I'm pleased to see a recent paper (pdf) by Ole Jann and Christoph Schottmuller which defends echo chambers.

To see their point, consider what happens when a rightist says a government can't raise much money merely by taxing the rich. Leftists reply: "you're just a shill for the rich. Who fnds you?" A useful message is therefore ignored. Imagine then we are segregated into echo chambers of left and right, and a leftist says the same thing to a leftish audience. Because s/he cannot be dismissed as a shill, the same message is taken seriously. Worthwhile information thus enters leftist thinking.

There are many examples of this. If CNBC says something disobliging about Trump, it's "fake news", but if (ex hypothesi!) Fox were to do so, it'd have credence with Trump voters. And whereas Jacob Rees Mogg's diatribes against the EU are ignored by Remainers, criticism of the EU (some of which is warranted) would be taken seriously if uttered by Remainers.

In these ways, echo chambers can actually promote worthwhile discourse by filtering out comments that'd not be taken seriously because they are discredited by association with our opponents, whilst giving proper credence to similar comments coming from like-minded people.

Jann and Schottmuller say:

Segregation into small, homogeneous groups can be a rational choice that maximizes the amount of information available to an individual. In fact, homophilic segregation can be efficient and even Pareto-optimal for society

There's another way this can be true. If we know we're speaking among friends, we can be more candid. "Safe spaces" can free like-minded people to express sentiments they would otherwise repress if they feared they'd be exploited by their opponents. The model here is perhaps the Chatham House rule, which allows speech to be reported outside as long as it isn't credited to a particular person. This frees people to speak more freely than they otherwise would. "Safe spaces may provide opportunities to communicate that would otherwise not exist" say Jann and Schottmuller.

Now, there are caveats here. All this takes for granted that there is sharp polarization. It would be better if there weren't and that we could speak freely and credibly across divisions. Given that we cannot, however, echo chambers might be a way to improve communication and to get messages across that would otherwise be ignored.

Also, this requires that there be some degree of diversity within the chambers. If people were to endlessly split and create new echo chambers after every slight disagreement - as the People's Front of Judea and some Trotskyites have done - then information is lost anyway.

The issue here is crucially important. In a polarized world, how can we best promote credible and high quality political communication? I'm not sure if Jann and Schottmuller are right that echo chambers really are the answer - although I welcome their challenging of my priors. But theirs is a much more useful intervention than partisan mythologizing about threats to free speech.