[SC] Using AI for Social Predictions

[Phoebe]

, maybe not full Science Corner (here is a better one for that, about the wee fast rotor that hopefully will shed light on the movement of very tiny things: https://www.zmescience.com/science/phys ... ning-0423/) but more Social Science corner. My thumbnail version of what is already a thumbnail of other studies is that some social scientists have developed an AI modeling system that can decently-well predict various interesting things about people's beliefs and conflicts. Some of the examples in the article pertain to religious belief or secularism, and violence or hostilities motivated by religious differences and xenophobia. The article is also concerned about the ethics of doing the modeling in the first place, given that it might suggest certain unwelcome truths or solutions to a problem.

I have questions, you know - it's nice that they can predict some things with their model, but much depends on how people understand those things in the first place and thus set up the modeling conditions. This would need to be explored in further detail. I will pick on the secularization example both because it is featured in the article and because it makes little sense to me. Two studies were cited, one which made predictions about things like religious attendance and belief in god, and the other which tried to explain why the U.S. and Europe have secularized differently, at different rates. The latter argued that four key conditions support secularization (basic resource needs met, personal freedom to choose your beliefs, pluralism or acceptance of diversity, and education in science and humanities). Variance in these areas, they conclude, explains the slower secularization in the U.S.

Yet I wonder if those 4 conditions are even the right ones, or how exactly they differ between Europe and the U.S. (e.g. I'm sure we can find examples of widespread hostility toward pluralism in both places - how is this being measured and compared?). Despite the overall difference in religiosity in the U.S., we're still growing our cohort of atheists at a rapid clip, and it's not clear to me that it's because of any sudden change in those four key areas. What if it's just that people are more likely to be exposed to the ideas of atheism, not have as intensive a training in religious alternatives due to declining religious attendance overall (which could be for many different reasons), and more likely to believe that nice people can live moral lives without basing their morality on religion or God? I'm pretty sure one of the big differences between atheists and believers is whether they think you can be moral without religion, and I'm also pretty sure it is far more likely for Europeans to believe that is true than for Americans. Is the AI model supposed to explain all of this purely as a function of education?

Maybe so - one of the researchers appealed to differences in education like private schools. Maybe he's thinking of the proliferation of religious private schools, particularly ones that don't teach science in a normal way. On the other hand, private school students in general perform better on measures of science education, which could be happening for many different reasons. More private school students are white compared to population averages, and atheists are more likely to be white men. If this AI model about beliefs and the four key conditions is right, how does it explain differences in those conditions that would make it more likely for white men to become atheists? Maybe they're supposed to be getting better educations? But then, are they also more pluralistic and accepting of diversity? Hm.

Anyway, it's interesting that social science research could use AI to more accurately model even things like shifts in people's beliefs, and make more accurate predictions about behavior. I'm not convinced its a success based on the brief summary in that article, but it is important to consider what conditions would make the modeling more accurate, and whether such models should influence public policy.