One possible option that never gets discussed is to nuke the amplification methods. If we stop recommending content automatically this ceases to be a problem.
But most of the time people like getting content recommended. It's what consumers want.
Gigantic unmoderated platforms existing like this that promote random snippets of speech to drive user engagement and ad-revenue is a thing that shouldn't exist. The problem we still haven't solved is how to specifically kill off platforms of this type without killing forums and discussion boards in general. I think there is a distinction there but I'm not certain precisely what defines it - but if anyone figures it out please let us all know!
IMO the moment you start highlighting things that people didn't explicitly ask for, it's an endorsement.
Automated recommendations of a human-curated set of content - e.g. Netflix recommendations for its suite of programming - are much less objectionable, because they can't amplify anything the organization has not intentionally decided to present. It's the combination of UGC and ML recommendations that presents problems.
Do they, or do they just boost some KPI that suits proxy for actual utility?
Anecdotally even in non-tech circles most of my friends complain about how bad recommended content has gotten, or roll their eyes at whatever "personalized" ad for garbage they've been recommended.
Do consumers want it, or is it merely taking advantage of some more subconscious human behavior patterns. And if the latter, is this something that is bad for humankind?
A threshold that calls the First Amendment in the US into question.
For example foreign states that pays armies of internet trolls to in effect choose president in the US -- is that what the 1st amendment wants to happen
I think "information" can kill more people than cocaine, is more dangerous
Indeed, I mean exactly what I said: the questions you raise call the First Amendment into question.
You're saying that maybe the 1st amendment could be edited a bit so that it was more clear that such things aren't okay?
I'm thinking people interpret "free speech" in sometimes quite different ways
The appropriate situation for those that want cocaine is something similar to the rules around purchasing/possessing/using alcohol.
From an economic standpoint (increased tax revenue, reduced spending on "enforcement" and incarceration, increased economic output because fewer people are in prison, etc.) and a societal standpoint (more resources available to the 2-5% of folks who end up with dependency problems, reduced property crime, not harming communities with significant numbers of residents being pulled out of the community and incarcerated, etc.)
As such, there's no good reason for any mind altering substances to be illegal. Rather, they should be regulated and taxed appropriately.
The only reason people keep using services like Twitter is because their network keeps them there.
Yes, people say they don't want recommendations, because 95% of them are irrelevant.
But then the 5% (or 2% or 0.5%) turn out to be super-relevant, and they find new people to follow that they love, and learn about things they love, and the experience in the end turns out to be a net positive.
Their actions show that it's valuable in the end. Otherwise the feature wouldn't exist at all. Recommendations aren't advertisements, sites don't make money off them -- sites use them because people genuinely find things that lead them to use the sites more.
People aren't stupid if you don't build your UI/UX around the assumption that they are. They also like transparent, understandable algorithms. Chronological feed of (only) the people one follows is as transparent as it gets. A chronological feed with some recommendations mixed in is more opaque and confusing. An algorithmic feed is an epitome of opaqueness. Opaqueness naturally drives users away because it doesn't exactly instill confidence that their posts will reach their followers.
Another example: do you understand what the "see less often" button in Twitter does? No one does. No one likes cryptic algorithmic bullshit forced on them with no way to disable it.
Choice is very important.
> versus what they actually do.
Do manipulations work? Of course they do. Are people happy when they are manipulated? Of course they are not.
> Recommendations aren't advertisements, sites don't make money off them
They absolutely do. Recommendations aren't there because Twitter wants to be helpful — they'd be more user-respecting as I said above if that was the case. They're there because they drive engagement metrics up, and those in turn translate into someone's KPI.