What's frustrating is that we don't have many ways to find out what's true.
Youtube's algo is an arbiter of taste whether they're happy about it or not. I'm not sure it's possible to be apolitical here.
Curation is terrible too, when it comes to kids content. I won't let my kids use YouTube kids anymore, it's awful. It used to have mostly appropriate kids content, but now people have gamed it.
Meanwhile, an individual injecting bias without the hack's true purpose being obvious to co-workers would be difficult. The more subtle and circumstantial the hack, the more likely it will be broken/ignored by someone else's changes.
A hack is still possible, but I'd want evidence of intentional bias before entertaining the claim.
https://en.m.wikipedia.org/wiki/Algorithmic_bias
It's quite real, and it is not illegal (in the US), so asserting this is a "conspiracy theory" is silly.
Software engineers are human beings, so of course they have inherent bias.
In practice it means 'a theory about a conspiracy' with the implication that conspiracies aren't real and therefore the a theory involving a conspiracy is wacky bunk. But sometimes the official narrative is that there was in fact a conspiracy, and in those cases the term 'conspiracy theory' is not typically used. A prime example is the 9/11 Commission Report which describes in great detail a conspiracy to commit mass murder. When it comes to 9/11, everybody agrees there was a conspiracy. I've yet to even hear rumor of anybody who believes that day was a series of unlikely coincidences. Whether you think it was bin Laden or Dick Cheney, it's obvious somebody was doing some conspiring.
I think the difference between the logical and practical meanings of the term is rooted in a specific case in which popular theories were theories about conspiracies, but for which the official narrative is that the guilty party didn't conspire with anybody. Obviously I'm talking about the assassination of John F. Kennedy.
— Dr. Walter Gibbs, Tron, 1982
Occam's Razor isn't some sort of physical law that governs human behavior.
https://simple.wikipedia.org/wiki/Occam%27s_razor
> Occam's razor (or Ockham's razor) is a principle from philosophy. Suppose there exist two explanations for an occurrence. In this case the one that requires the least speculation is usually correct. Another way of saying it is that the more assumptions you have to make, the more unlikely an explanation. Occam's razor applies especially in the philosophy of science, but also more generally.
> That's a popular conspiracy theory, but it doesn't hold up to Occam's Razor.
This comment is open to interpretation. A reasonable interpretation is that something must hold up to Occam's Razor, or else it is not true, especially with the inclusion of "That's a popular conspiracy theory".
In fact, the "conspiracy theory" charge is also flawed. The speculative claim "The algorithms are a proxy for some youtube engineers' tastes" may actually be true. We do not know.
Statement #2:
> The simpler explanation is that content from all sides gets buried to some extent, but it appears biased to people because they only notice when their own content is buried.
This persuades the reader, perhaps based on a presumed knowledge of the phenomenon of confirmation bias, that any such beliefs are likely an example of such confirmation bias. Again, the truth is not known.
If you don't like your comments being interpreted, you could tighten them up to minimize the need or possibility for interpretation.
Bias has been introduced.
The fact that "both sides" are affected doesn't mean there's no bias, it means there's > no bias. It's literally impossible to censor something without SOME bias, because somewhere, someone doesn't find the thing you censor objectionable.
https://www.judiciary.senate.gov/imo/media/doc/Epstein%20Tes...
I wouldn't be surprised if there's an anti-alt right bias (which frankly I don't feel so bad about because many of those folks are toxic).
Also, to be pedantic that's a misapplication of Occam's razor. The simplest explanation to explain nature is likely true. Nature likes simple laws. People and organizations, however, can be endlessly complex.
> In 2012, Epstein publicly disputed with Google Search over a security warning placed on links to his website. [...] Several weeks later, Epstein admitted his website had been hacked, but criticized Google for tarnishing his name and not helping him find the infection.
The other party has its skeletons-in-closets, but no analog to the above-described document exists. So, people using popular search engines to research their options ahead of election day find this horrifying information about party A and nothing comparable for party B, and they end up voting for party B mostly on this basis.
So, those search engines, despite providing results that faithfully reflect reality without a priori bias toward either party, will (hopefully) greatly influence the outcome of the election. Is that bad? Should they have censored this document in results so that nobody comes away from a web search with different opinions than they had going in? Is finding the truth an unacceptable thing for "big tech" to facilitate?
It's like when politicians like Steve King complain that search results for their names make them look bad. To me, if you conduct yourself as a racist, bigoted lunatic on the public political stage, that's what a web search result about you should say. I recognize that statements like "racism is bad" are subjective value judgments, but that's rolled up in all of this too: if as a society we don't think racism and other bigotries are bad, then this sort of search result won't be seen as negative. So there's some admission here that Steve King and his ilk know their words and actions are viewed negatively by a large segment of the population and so they would like them to be suppressed. By not suppressing them, are major search engines exhibiting a bias?
No, not really. I don't think so.
P.S. The burden of proof falls on the purveyors of conspiracy theories, not the other way round.
I guess when it comes to politics the Youtube algorithm just shows whatever the user doesn't want to see.
Do people really believe youtube engineers are encoding their values into the algorithm, intentionally or unintentionally?
Intentionally, I just can't imagine. What might this look like? How does it get through code review? Is it broad enough to apply to the entire ranking algo, or does it only act on tiny pockets that the offending engineers specifically care about?
In the unintentional case, certainly they're twisting specific knobs like engagement time, ad revenue, vids watched per session, etc... but is that encoding a specific person's taste, or a broader company's thirst for revenue? Do they even care about taste, or just metrics?
In other words, I find it suspect that youtube's policy is "More Jordan Peterson conflict vids", and more likely that it's "float high activity posts to the top", creating an echo chamber.
To be clear, I think there's some blame that rests on youtube eng (and let's be realistic, the broader product teams bossing eng around).
But it's not for encoding their tastes, it's for indiscriminately allowing the tastes of their markets to echo in a feedback loop that becomes a screech of extreme content. Not always anti-social, or even politically directional, just extreme flavors of whatever it is.
It would not take much to nudge the system in that direction. The resulting payoffs could be in the billions.
Not sure if anyone has refuted this person's points but they are pretty alarming no matter what side of isle you sit.
https://www.youtube.com/watch?v=fDqBeXJ8Zx8 (What 40,000 Videos Tell Us About The Trending Tab)
An awful lot of discussion around YouTube and human curation/moderation elides just how much content they're dealing with; the most recent figures I've seen -- which are over a year old at this point -- show that about 500 hours worth of video is uploaded to YouTube every minute. If we figure an average video is about 15 minutes (a 2017 statistic I found suggested that "the average length of a first page YouTube video is 14 minutes, 50 seconds"), that's 2000 videos a minute -- over 30 a second.
Is it impossible that YouTube employees have secret levers to raise and lower the profile of individual videos and channels? Of course not. But it's absurd to think that such manipulation is common, let alone corporate policy.
To me the problem isn't that flat earth videos exist on Youtube, but the implied "legitimacy" of the recommendations, along with the typically addictive UI elements like autoplay. Flat earth / Nibiru / etc insanity has always been on the internet, but when you saw it, you knew that you were in some obscure corner of the internet that was not normal. If anything, YouTube's insistence on mixing brand-safe, front-page recommendations along with content-adjacent recs lends credibility to the (many fewer now) "rabbit hole" recs. If you're on "Tony Blair is a reptile.mp4" and all you see are things like "Tony Blair is ACTUALLY a hologram.mp4" with 2k views, it's like "whoa, this is weird". But if you get 3 reptilian recs and 5 TED talk recs that have millions of views, maybe it seems more normal? I'm not sure.
Source: an analysis of 40,000 trending videos [7 months of trending tab videos]: https://www.youtube.com/watch?v=fDqBeXJ8Zx8
It's dangerous to become the arbiter of taste, I can understand Google's aversion to more human involvement beyond just the problems with scale.