It'll take decades, but… what approaches are worth trying?
Here's an example: Get X (10?) million people to watch 10 hours/year of strangers who they would normally not encounter or agree with, and to understand them as real people.
To do that, produce and televise+stream a long-form TV show, like a version of Braver Angels' Red/Blue Workshops[1] that's actually enjoyable to watch. Imagine a well-produced show with deep participant profiles - a cross between a reality TV show and a binge-watchable Netflix/HBO series.
It would humanize the participants first, then after viewers care about them, their lives, and their families, the actors would gradually explain their backgrounds and opinions - some of which a viewer will disagree with. This uses TV to scale up the "Contact hypothesis"[2]: viewers would "meet" and hear from people they may not interact with regularly. Just like all TV, the producers would have an agenda. (Sarah Silverman's "I Love You, America"[3] is the closest I've seen to this, and it's not all that close.)
I'm only giving this as an example. What approaches do you think would be worth testing?
Assume you have decades plus unlimited philanthropic and/or public funds. For the sake of argument, imagine Mark Zuckerberg decided that a divided America is bad for Facebook's valuation and seeded this project with a $1 billion donation. It took over 20 years to get here and it'll take decades to undo it.
[1]: Braver Angels: https://braverangels.org/what-we-do/red-blue-workshops/
[2]: Contact hypothesis AKA intergroup contact theory: https://en.wikipedia.org/wiki/Contact_hypothesis
[3]: I Love You, America: https://www.youtube.com/watch?v=xmQpf-B94mc
(This question came out of a thread comment yesterday: https://news.ycombinator.com/item?id=25021438)