Great pull request here which improves the algorithm: https://github.com/twitter/the-algorithm/pull/17
Why do companies even bother to put source up on github? To put up a front that their open source? What a joke.
Twitter's Recommendation Algorithm
571–580 of 1001 posts
Re: Twitter's Recommendation Algorithm
#572Expect to see A LOT more spam on Twitter after this release. It's like giving SEO spammers access to google search ranking algorithm.
Stuff like this always has consequences, it doesn’t mean it’s a net negative for society. It means you need to adapt and actually fix the problems, while also benefiting more from the accountability.
That’s always been a risk of open source and not being hyper-centralized.
Re: Twitter's Recommendation Algorithm
#573Earlier quoted context omitted.
That lines up with reporting from Casey Newton a few days ago where a handful of VIPs e.g. Musk, LeBron James, AOC were being used as weather vanes to understand what the algorithm was doing. It definitely isn't just metrics. Any algorithm change that negatively affected Musk was clearly not going live.
Do you think the code looked like that prior to Elon's purchase? I suspect that there was another name there before. Separately, which of these groups do you think that they use as a control?
There is absolutely no reason to believe there was another Single user getting this treatment before. The Elon-case was just copy & pasted as an ego-stroking hack.
Re: Twitter's Recommendation Algorithm
#574So as expected, there is exactly nothing that favors posters from one side of the political spectrum. I don't expect that this article will do anything to calm down those who are convinced otherwise though. Well written article, from an engineer's perspective.
[dead]
Re: Twitter's Recommendation Algorithm
#575For ranking the candidates these predictions are combined into a score by weighting them: "recap.engagement.is_favorited": 0.5 "recap.engagement.is_good_clicked_convo_desc_favorited_or_replied": 11* (the maximum prediction from these two "good click" features is used and weighted by 11, the other prediction is ignored). "recap.engagement.is_good_clicked_convo_desc_v2": 11* "recap.engagement.is_negative_feedback_v2":…
Having worked at similar companies on similar systems usually A/B experiments and smaller probability of an action bigger weight it must have to matter much overall. The constants are generally done through some ab tests to get them into reasonable overall behavior but they are a pain to tune and very unlikely optimal in any real sense as it’s often too difficult to do extensive search of them. Like often I’ll see new target have a couple different weights tried on an ab and then maybe second set of experiments after rough magnitude is determined.
Re: Twitter's Recommendation Algorithm
#576Earlier quoted context omitted.
Elon is addressing this in the Twitter Space right now. "It definitely shouldn't be dividing people into Republican and Democrats; that makes no sense[...] you've identified something we should be getting rid of right away."
As if Elon has a clue what that feature is or is not being used for.
He didn’t say he knew what it did? It’s a good enough response to say that it shouldn’t be doing that period.
Re: Twitter's Recommendation Algorithm
#577Re: Twitter's Recommendation Algorithm
#578Earlier quoted context omitted.
So many questions. How are users tagged D or R? Is that a manual process or automated somehow? What is the effect of these tags? Can I find out if my Twitter account is in one of those buckets?
And how are they choosing to balance them, per capita, or just both sides should get 50%? It seems pretty clear they are making editorial decisions here. Does that break their section 230 protections?
Please read something. Preferably section 230, which is short. Alternatively something entirely different, as long as that keeps you occupied.
Re: Twitter's Recommendation Algorithm
#579Interesting: // we only keep unfollows in the past 90 days due to the huge size of this dataset, // and to prevent permanent "shadow-banning" in the event of accidental unfollows. // we treat unfollows as less critical than above 4 negative signals, since it deals more with // interest than health typically, which might change over time. val unfollows: SCollection[InteractionGraphRawInput] = GraphUtil .getSocialGraph…
How long does the NSA record them?
Re: Twitter's Recommendation Algorithm
#580Earlier quoted context omitted.
So many questions. How are users tagged D or R? Is that a manual process or automated somehow? What is the effect of these tags? Can I find out if my Twitter account is in one of those buckets?
And how are they choosing to balance them, per capita, or just both sides should get 50%? It seems pretty clear they are making editorial decisions here. Does that break their section 230 protections?
>It seems pretty clear they are making editorial decisions here. Does that break their section 230 protections?
No.
https://www.techdirt.com/2020/06/23/hello-youve-been-referre...