I feel that both the articles you critique and your own work ignore the issue of self-selection. Maybe colleges turn young adults into liberals. Maybe liberals self-select to go to college...
Interesting! No I hadn't seen that article, though I've seen other things critiquing the echo chamber narrative. It's definitely right that, in general, persuasive effects are hard to predict, especially with small-effect changes like control of internet exposure. That said, there is a mountain of evidence in favor of the (not-too-surprising) claim that exposure to arguments in favor of a position, on average, tends to make people more convinnced of that position. (Coppock, Persuasion in Parallel (https://press.uchicago.edu/ucp/books/book/chicago/P/bo181475008.html) is a good book in the political case.)
I agree there's certainly a level of self-selection when it comes to going to college, which is hard to control for. (Though a variety of studies try, eg if I remember correctly this one: https://www.tandfonline.com/doi/abs/10.1080/2194587X.2016.1195750). But I think we should have a very low prior that it's ALL self-selection, for two reasons:
1) First, just the above evidence for the persuasive effects of arguments. There's no question that people see more liberal-leaning arguments when in college; some will have a back-reaction by becoming more conservative, but the average will be moved by those arguments.
2) Second, there are *tremendous* socioeconomic pressures to go to college, so it would be extremely surprising if there were that strong selection effects amongst those who had a choice.
Perhaps more to the point, though, notice that the argument I made in this blog post didn't presuppose that Nathan was going to become more liberal. It presupposed that he could expect to EITHER become more liberal OR become more conservative—either way, his opinion in those biconditionals will budge. And the data on that is pretty solid: increasingly, most people who are politically informed do fairly-consistently lean one way.
Your response is definitely to the point. Which does not mean that I am fully persuaded.
If I wanted to counter the self-selection argument I would probably point to a difference between FB and the college. We are probably much more resistant to changing our mind in an adversarial environment (=FB, where many people are in a "warrior" mode) than in an environment where we want to feel among friends and hence have a strong incentive to reduce our cognitive dissonace from non-majoritarian views (the college).
Still, the above points to a broader idea that the speed of absorbtion and internalization of information is partly endogenous and depends on people's incentives. This is what, in my view, is missing in the Bayesian literature, where exposure to information pretty much mechanically changes our views via Bayesian updating.
Our views likely serve many purposes, including signalling and reducing cognitive dissonance. They may also sometimes be pure rationalizations (I like shooting a gun, hence I believe that easy access to guns increases public safety). They do not just passively reflect the random information stream we have been exposed to during our lifetimes.
Thanks Rafal! A lot of what you say seems right to me—I think it's one of the biggest missing parts of standard Bayesian models that they are in a sense 'passive' when it comes to processing information: whatever information comes in, you condition on it; the only ways prior beliefs or attitudes can affect it is by affecting your prior conditional probabilities. A lot of what I like about the generalizations of Bayesian models I tend to use (eg in the linked paper on polarization) is that they allow more flexibility here, including that people can have some control over how they process information and therefore how they (on average) react to it. For example, the model in section of that paper (on whether to scrutinize an argument, or not) captures some of what you mention wrt whether people are in an adversarial or open-minded frame of mind. I'm hoping to explore variations and elaborations of it in future work!
hi Kevin, have you seen this article?
https://www.nature.com/articles/s41586-023-06297-w
I feel that both the articles you critique and your own work ignore the issue of self-selection. Maybe colleges turn young adults into liberals. Maybe liberals self-select to go to college...
Interesting! No I hadn't seen that article, though I've seen other things critiquing the echo chamber narrative. It's definitely right that, in general, persuasive effects are hard to predict, especially with small-effect changes like control of internet exposure. That said, there is a mountain of evidence in favor of the (not-too-surprising) claim that exposure to arguments in favor of a position, on average, tends to make people more convinnced of that position. (Coppock, Persuasion in Parallel (https://press.uchicago.edu/ucp/books/book/chicago/P/bo181475008.html) is a good book in the political case.)
I agree there's certainly a level of self-selection when it comes to going to college, which is hard to control for. (Though a variety of studies try, eg if I remember correctly this one: https://www.tandfonline.com/doi/abs/10.1080/2194587X.2016.1195750). But I think we should have a very low prior that it's ALL self-selection, for two reasons:
1) First, just the above evidence for the persuasive effects of arguments. There's no question that people see more liberal-leaning arguments when in college; some will have a back-reaction by becoming more conservative, but the average will be moved by those arguments.
2) Second, there are *tremendous* socioeconomic pressures to go to college, so it would be extremely surprising if there were that strong selection effects amongst those who had a choice.
Perhaps more to the point, though, notice that the argument I made in this blog post didn't presuppose that Nathan was going to become more liberal. It presupposed that he could expect to EITHER become more liberal OR become more conservative—either way, his opinion in those biconditionals will budge. And the data on that is pretty solid: increasingly, most people who are politically informed do fairly-consistently lean one way.
Is that responsive to your worry?
Hi Kevin, thanks.
Your response is definitely to the point. Which does not mean that I am fully persuaded.
If I wanted to counter the self-selection argument I would probably point to a difference between FB and the college. We are probably much more resistant to changing our mind in an adversarial environment (=FB, where many people are in a "warrior" mode) than in an environment where we want to feel among friends and hence have a strong incentive to reduce our cognitive dissonace from non-majoritarian views (the college).
Still, the above points to a broader idea that the speed of absorbtion and internalization of information is partly endogenous and depends on people's incentives. This is what, in my view, is missing in the Bayesian literature, where exposure to information pretty much mechanically changes our views via Bayesian updating.
Our views likely serve many purposes, including signalling and reducing cognitive dissonance. They may also sometimes be pure rationalizations (I like shooting a gun, hence I believe that easy access to guns increases public safety). They do not just passively reflect the random information stream we have been exposed to during our lifetimes.
Thanks Rafal! A lot of what you say seems right to me—I think it's one of the biggest missing parts of standard Bayesian models that they are in a sense 'passive' when it comes to processing information: whatever information comes in, you condition on it; the only ways prior beliefs or attitudes can affect it is by affecting your prior conditional probabilities. A lot of what I like about the generalizations of Bayesian models I tend to use (eg in the linked paper on polarization) is that they allow more flexibility here, including that people can have some control over how they process information and therefore how they (on average) react to it. For example, the model in section of that paper (on whether to scrutinize an argument, or not) captures some of what you mention wrt whether people are in an adversarial or open-minded frame of mind. I'm hoping to explore variations and elaborations of it in future work!