I think the author is wrong in claiming that modern attention-optimizing recommendation algorithms are better than more primitive, poorer recommendation algorithms. Appearing to be more engaging/addictive does not imply more value. It's a measurement problem.
for modern day businesses it sadly does. But that misalignment of how to define "quality" is a part of why we're in this real time divide of whether social media is good/bad to begin with.
At what point do newspaper clipping arranged together become the work of the arranger and not the individual newspapers. If I take one paragraph from the NYT and one paragraph from the WSJ am I the author or are the NYT and WSJ the author? If I take 16 words in a row from each and alternate, am I the author? If I alternate sentences am I the author?
At some point, there is a higher order "creation" of context between individually associated videos played together in a sequence. If I arrange one minute clips into an hour long video, I can say something the original authors never intended. If I, algorithmically, start following up videos with rebuttals, but only rebuttals that support my viewpoint, I am ADDING context by making suggestions. Sure people can click next, but in my ransom note example above, people can speed read and skip words as well. Current suggestion algorithms may not be purposely "trying to say something" but they effectively BECOME speakers almost accidently.
Ignoring that a well crafted sequences of videos can create new meaning leaves us with a disingenuous interpretation of what suggestion algorithms either are doing or can do. I'm not saying that google is purposely radicalizing children into lets say white nationalists, buuut there may be something akin to negligence going on, if they can always point to a black box algorithm, one with a mind of its own, as the culprit. Winter v. GP Putnam giving them some kind of amnesty from their own "suggestions" rubs me the wrong way. Designing systems to give people "more of what they want" rubs me the wrong way because it narrows horizons not broadens them. That lets me segue into again linking to my favorite internet article ever (which the bbc has somehow broken the link to so here is the real link, and an archive https://www.bbc.co.uk/blogs/adamcurtis/entries/78691781-c9b7... https://archive.ph/RoBjr ) Im not sure I have an answer, but current recommendation engines are the opposite of it.
Read 230.
https://www.law.cornell.edu/uscode/text/47/230
"No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider."
"No provider or user of an interactive computer service shall be held liable on account of any action voluntarily taken in good faith to restrict access to or availability of material that the provider or user considers to be obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable, whether or not such material is constitutionally protected.
230 says you can moderate however you like and what you choose to leave up doesnt become your own speech through endorsement osmosis.
I agree with 230 to a point, but at some extreme it can be used to misrepresent speech as "someone elses." Similar to how the authors of newspapers wouldnt be the speaker of a ransom note because they contributed one letter or word, and it woild be absurd to claim otherwise.
If the arranger is the speaker, restrictions on free speech apply to their newly created context. Accountability applies.
There are better ways of formulating the question that avoid this paradox, such as "what are the necessary and sufficient conditions for editorial intervention to dominate an artifact?"
230 is about who is NOT liable. Platforms are NOT liable for what they don't moderate, just because they moderated other things. It protects them from imperfect moderation being used to claim endorsement.
i conject they had some kind of page move and archival that is broke.