Not only the most incendiary topics but the most incendiary people as well! So the crazy angry neighbor who AFK has two friends gets 500 followers on Twitter, and lots of "likes" on various partisan sites. That rush of approval causes the neighbor to invest even more of their time online where they feel important. This wouldn't be possible without network effects, so I agree, the algorithm can't be discounted or disambiguated from the phenomenon as a whole.

Yeah, we shouldn't discount the algorithms put in place by people; but we should also keep in mind that those algorithms only operate based on inputs from people. If the outputs the algorithms are giving you are not desirable, then it is solely the fault of the people who are feeding that algorithm in the way they are, etc and so forth. Algorithms are not sentient nor have agency over the results of their actions.

We however, do. This is why the human element side of things is my target. Even if the algorithms change, not much will if we don't.

Algorithms are designed, evaluated and impoved by organizations, so are not a point of blame at all. They’re not accidentally putting divisive stuff to the top, the analytics, devs and management all surely see the consequences and decide that the algos are aligned with the business .

Many Organizations generally are within some sort of competitive landscape - companies in the market, non profits with donations, city and state govts for tax base. So they all will optimize for effectiveness towards some metric to some degree, restricted by whatever fear of backlash or consequences they may have.

No one is going to approve a ‘say anything as long as it bumps ratings immediately’ algorithm, but almost everyone is going to try to do at least what the top competitor seems to be doing.

Which is algorithmic/ML based engagement optimization, with some kind of ‘but not the most controversial if it’s literally nazis’ type filter for things with too much political blowback (aka calls for insurrection), or legal issues (child porn).

If you look, it’s literally almost everywhere now. And earlier they didn’t even have the ‘not literally nazi’s’ filter.

But it's still people at the end of it. The "algorithm" didn't come up with the topic, didn't write the article, didn't perform the actions. People did. People bought the ads. People promoted the content. And ultimately, people also wrote the "algorithm".

Some hyperbole, but considering the real world consequences right now maybe not as much as we’d like to believe - People are also the ones who use chemical weapons to gas their opponents in particularly nasty wars.

It doesn’t mean we shouldn’t all agree to not do that, or even try to enforce that people don’t do that.

Incendiary topics draw more attention, thus more likes / reshares. Feed algorithms can amplify this effect, but it's always there, unfortunately:

https://www.youtube.com/watch?v=rE3j_RHkqJc

Likes and reshares are often consumed via feed like interfaces though?

I get it definitely happens with folks texting friends links or sharing it via email, but the friction with those is higher and it gives time for folks to think more than the feed interface.

It’s highly efficient psychological pressure/engagement and maximizing the focus towards the engaging content that is the real problem IMO. It makes something that someone could think about (potentially) and choose to stop into something they become hyper engaged with constantly and can’t turn off.

> Nothing gets people going like someone saying something divisive

Shiri's Scissor: https://slatestarcodex.com/2018/10/30/sort-by-controversial/

Thank you for this.

Not to quote taco girl too overly much; but how about both?

Algorithms require input to give you output, right? So even if you are correct that badly created algoritms are to blame, we still have the people who fall for the algorithms outputs because they are feeding it the inputs.

To put it simply, it all goes back to the calculator/computer quote in regard to their users. They are only as smart as the person using them.

So, from that point of view; there are a lot of really dumb users on the internet.

Not quite.

If 99 out of 100 users input not-BS, and 1 user inputs BS, but the algorithm picks the 1 users BS input and feeds it to the other 99 first so they get angry and yell at each other over the 1 BS input together - that is a historically unique situation.

If it does it at scale for cheap, that is also a historically unique situation.

If it does it without a human being directly involved in the choosing and pushing process (and feeling bad potentially!), that is also a historically unique situation.

If someone sees all 100 users posts together, it’s easy to see the BS is BS. If the BS post goes to the top of the feed and is the first thing they see, and the normal posts either don’t show up, or only a few of them show up? That is a completely different experience for the consumer of the media.

I feel like all of this just means that teaching proper critical thinking in schools again would solve most of this. Along with maybe an educational update for those who missed it.

While it would help a little, you’ll always have a minimum of 25% (probably more like 50%+) of folks who lack the degree of emotional self control necessary for any sort of rational learning to override the emotional response.

They might feel worse about it later because they know they shouldn’t have reacted that way, but in my experience that doesn’t help as much as you’d want.

Doesn’t take much to get a mob going unfortunately.