Make your algorithms downplay controversial stuff (which is often controversial because it uses dishonest language tailored to sow discord and minimize common ground interpretations).
Make your algorithms promote level headed, honest, non-tribal language and verifiable truths so that areas of global consensus are elevated.
Democracy will work again and we can get rid of the despots.
Controversial stuff is often verifiable truth that is politically inconvenient for some interested party.
> Make your algorithms promote level headed, honest, non-tribal language and verifiable truths
Algorithmically detecting content that is either “honest” or “verifiable truth” is a pair problems go beyond mere hard AI problems.
Read "In Defense of Troublemakers" for an overview of where research on dissent is at.
It's pretty scary, because there just isn't anything you can do to turn the political machinery in everyone's heads off and it's default behavior is to shutdown opinions it doesn't agree with seeking consensus not truth.
It's doubly scary if you read books on our cognitive shortcomings. We are not capable of computing the truth values of most truth claims. It's so expensive to know anything that most of our mental models are ultra-low resolution. We are a guess-and-check species who rationalize and politicize everything. To top it all off we barely ever think we are wrong.
It's not a mystery why the world is so chaotic and crazy. It's not a mystery why authoritarianism has run rampant throughout history.
I wish we were far more educated on the subject of how we get things, so terribly wrong.
I'm coming strongly to similar conclusions - people are poor thinkers and totally unaware of this fact. They're not self-critical, but they are very critical of anyone who disagrees with them.
I summed it up in a quickie comment recently that has stuck with me. "Nobody thinks they're a racist. When they say racist things, they don't think they're being racist. They think they're being reasonable."
Maybe social networks are not the right place for these types of controversial discussions as everything gets distorted by the worst elements of each tribes.
I admit that verifiable truth is hard but they must be able to see what stuff only gets "liked" in a tribe-wise manner and they could downplay that.
FB has a fiduciary responsibility to make their platform show content that it's users with engage with the most, for the longest amount of time. If that happens to be controversial content, it's going to be difficult to convince them to change their algos to inherently downplay controversial content.
Not sure it's possible to make them change like that without government intervention. And even then, what change can one government impose vs a a massive global company like FB? (I actually don't know this, but would be interesting in learning from some examples if anyone has any)