Hacker News Reader

Hacker News

Comment by bufferoverflow | Hacker News Reader

Parent Full thread

bufferoverflow

> insinuates the initial suspension was politically motivated, which seems hard to believe

How is that hard to believe? We know for a fact Google does political censorship.

https://www.projectveritas.com/2019/06/24/insider-blows-whis...

https://www.projectveritas.com/2019/06/25/breaking-new-googl...

https://www.youtube.com/watch?v=ricI5t66cj8

(note these are fresh links just from the last 2 days)

View on HN

I work at Google on ranking algorithms for Search. I'm aware of everything that controls how web pages are ranked. If I ever encountered a system that was designed to alter the results for political reasons, I would resign.

Just as an example: a controversial political figure's popular website is buried on page 5 of Google results (underneath a mountain of criticism) when you search for his name, despite being in the top 3 results of page 1 for all of Qwant, DDG, Bing, Mojeek, and Yandex. Obviously, the website is being downranked for being full of "conspiracy theories and fake news" (I wouldn't completely disagree that it is). But who decided all this? Is this not "political"?

What if it's downranked because most of the people who search for him aren't looking for his site, but for commentary about him?

I mean, maybe this sounds implausible to you, but I think you should be conscious of whether you're discarding it as a possibility and why.

I'd be happy to pass it along to debug if you have a specific query in mind. In the cases of this I've seen in the past, there's been a specific actionable bug that we were able to fix.

(Lots of other search engines use each other as backends, so don't be too surprised when they return consistent results.)

I think DDG and Qwant basically use Bing as a backend, potentially augmenting with their own crawlers. Yandex and Mojeek use all their own stuff, as far as I can tell. The query is "alex jones". At least put the garbage website at the bottom of page 1 where most people who aren't already looking for it still won't click it, page 5 is just lacking in subtlety.

Thanks, I'll pass it along.

Infowars being buried so deeply in the "Alex Jones" results is highly suspicious. How can Google's results be so ineffective compared to other search engines?

The "Pelosi drunk" results are also markedly different from DDG. I just now tried this one because the results were being manually short-circuited the day after the video went viral, both by DDG and Google (but not Bing). It appears that DDG has reverted to delivering actual results for that query while Google's appear curated.

During the election some fuss was made over auto-completion for "Hillary" vs. "Trump" searches ... apparently embarrassing for Google because these were soon "fixed".

I could go on and on.

Do you at least admit that Google restricts searches for the Christchurch shooting video? Maybe your definition of "political reasons" is very narrow.

I don't actually have a problem with burying Alex Jones. The only reason I have enough interest to engage in this discussion is that Google continues to deny these things.

Google does not curate web results, ever. The results for [pelosi drunk] are different certainly, but in what sense are DDG's better?

Google has a policy about "sensitive and disparaging terms associated with named individuals" in autocomplete https://support.google.com/websearch/answer/7368877?hl=en which is often why you sometimes may not see the queries in autocomplete that you expect. It also has a policy about "violent or gory predictions" so it wouldn't surprise me if you aren't seeing autocomplete suggesting the Christchurch video, if that's what you mean.

I'm sure search results are not curated in the sense that a stock answer is provided for a given query, but I would expect that filtering and ranking of search results involves lists that are curated -- for example, "offensive" words or "disparaging terms", etc..

Since you work on search at Google, maybe you could explain how typing "hillary clinton cri" stopped suggesting the common search term "hillary clinton crimes" and instead suggested "hillary clinton criminal reform".

https://trends.google.com/trends/explore?date=today%205-y&ge...

Apparently "crimes" a disparaging term when associated with "hillary clinton", but not when associated with, say, "jared kushner"?

Isn't it explained by the policy I just linked? "Jared kushner cr" yields [jared kushner credentials] as a suggestion for me.

"cri", not "cr". I have two different devices that give different results; one lists "jared kushner crimes".

Since it isn't a matter of a single word (crimes) being flagged as disparaging, this blacklist can apparently be finely tuned to specific prediction results. (Maybe it's not a simple blacklist, but an ML model. Even scarier.)

I see: [jared kushner criticism] [jared kushner cries in closet] [jared kushner's father crime]

and so forth. Are you sure you aren't seeing your own search history as the suggestion? Another possibility is that you're seeing "... crimes" instead of the whole query? That can happen when the system isn't able to suggest a full query and starts suggesting individual word completions.

Do you really think that it simply excludes predictions with the word "crimes"?

Yes, I was seeing "... crimes" when typing for "jared kushner cri". But with other names it shows a complete prediction:

"julian assange cri" --> "julian assange crimes"

"jeffrey epstein cr" --> "jeffrey epstein crimes"

I don't know all the details of how that system works, I don't work on it, but surely you'd agree that they're both more appropriate suggestions, as both of them have been indicted. I don't get "trump crimes" for "trump cri" either.

I agree they are more appropriate. Surely you'd agree there's a lot of room for bias to manifest in a system like this.

Yes, and I definitely understand the argument that Google shouldn't be doing this, but note that we aren't talking about evidence of bias anymore. When someone sees confusing behavior on a political query, the tendency is to assume that political bias is involved, when it's just a very ordinary failing that search engines have all the time.

I don't know the basis of the decision to enact that policy around autocomplete, but I do know Google has been sued many times due to the queries in autocomplete associated with people's names. https://searchengineland.com/google-faces-autocomplete-lawsu... https://www.pcmag.com/news/326136/google-could-face-lawsuit-... http://www.dmlp.org/blog/2013/filing-lawsuits-united-states-...

I think he's talking about actual links to the Christchurch video. They exist, and I'm pretty sure that's what most people who search terms such as "Christchurch video stream" are looking for. I don't think I can believe that Google doesn't censor or re-rank certain "flagged" search results, that just doesn't the smell test.

There are a few situations where Google will remove urls listed here https://support.google.com/websearch/answer/2744324 as well as for spam.

> Google does not curate web results, ever.

Where is the transparency or independent third party oversight to back up this statement?

Or do we just have to take it on faith?

That's a very strong claim you're making there. Do you not consider mixins from News and YouTube to be "web pages"?

I've also worked at Google. Things that are clearly the case:

1. There are blacklists of certain political query completions on web search, like "hillary clinton emails".

2. There is a controversial query blacklist on YouTube with political content. Go look at the file in google3 if you really think the screenshots are faked:

https://www.projectveritas.com/2019/06/26/blacklisted-leaked...

3. Google News is quite clearly blacklisting Breitbart from ranking: it's indexed but in many years of using Google News I've never once seen it be selected. The Guardian, on the other hand, routinely makes it into nearly every selected story.

The writing is clearly on the wall for you. Whether organic web index results are reranked politically or not, and for what it's worth I believe you for that specific subset of the system, other teams in the company that didn't inherit the same culture are doing politically motivated things with the full knowledge and acceptance of the executives. It's only a matter of time before web search falls to that new culture too.

And when it happens you probably won't even realise it. Some team you've never heard of will quietly adjust a machine learning model in subtle ways you'll never detect, to "debias" it in order to yield "better outcomes".

1. There are not blacklists of "political" query completions, as I explained extensively in this other thread. The policy is to not show "sensitive and disparaging terms associated with names" in autocomplete. This applies uniformly, and there's nothing political about it. https://news.ycombinator.com/item?id=20530414

2. I can't speak to YouTube's policies, but there isn't any evidence there that this was designed to favor either side, rather than just to avoid the issue entirely.

3. The guidelines that Google uses to assess the quality of websites are public. You can read them here: https://static.googleusercontent.com/media/www.google.com/en...

Does your claim include the keyword "abortion"? https://www.theguardian.com/world/2018/feb/13/abortions-near...

There are complaints on both sides of this issue: http://christianresearchnetwork.org/2019/01/16/the-smoking-g...

You sir, are a liar:

https://www.blog.google/products/search/our-latest-quality-i...

That is by Google's VP of Search, Ben Gomes. That's your boss, supposedly. So, Human evaluators flag content in general ("low quality"); and particularly, remove or down-rate content which Google determines to be "fake news", as Gomes puts it.

and some of the results:

https://www.wsws.org/en/articles/2017/09/19/goog-s19.html

Consortium News, Global Research, Counterpunch, Truthdig, Russia Today, Truepublica, World Socialist Website and others - censored. Not entirely, but to a great extent.

Those ratings are used to evaluate the search system, not to change it directly.

Project Veritas has utterly shattered their credibility so many time, that they should not be used as a source even if what they say is true.

They edit videos, not even deceptively, but in a way that makes the people featured in the video make them look like they're saying something completely different. Not just clipping but also reordering the clips

James O'Keefe claiming something doesn't make it a fact. On the contrary, I'd be inclined to doubt the claim in that case.

You clearly didn't look at the links. O'Keefe is irrelevant.

With the amount he's been caught doing "selective editing" I'd say he's pretty relevant.

Project Veritas was created by a known liar and manipulator. Anyone falling for their schtick needs to learn how to be less gullible.

Veritas is tabloid propaganda. They have less than zero credibility.

You realize the rightwingers say the same about CNN, NYT, and other "reputable" sources? They are all dirty, on all sides. Thankfully, we have videos, so we don't have to rely on anybody's reputation.

Nope, reputation is definitely still important. As it turns out, you can edit videos to produce a desired effect/impression -- just saying "it's a video" doesn't really establish objective truth.

Since O'Keefe paid $100,000 to settle a lawsuit centered around edited video designed to make ACORN look bad, it's pretty reasonable to treat his videos with skepticism.

As I said, they all have horrible reputation to me - I've seen major lies in all major news sources. They all editorialize to fit their agenda. Left or right wing, all bad.

But since you're the one claiming these particular videos are somehow edited, present your evidence. The burden of proof is on you.

That's interesting. Where exactly did I say these particular videos were edited? I only have one comment in this thread, should be pretty easy to point it out.

In the event that you realize you read something into my words that wasn't there, I would politely invite you to ask whether or not that reflects on your priors regarding all major news sources.

This says more about the rightwingers that say that than it does CNN, NYT, WaPo, Bloomberg, etc. - there is a reason the rightwing in this country has embraced Trumpism, and it isn't because truth is a guiding principle of their movement.

The media analysis skills of folks on HN continue to astound. What makes you think project veritas is a credible source? Im serious.

The video recording. Are you gonna claim they are fake?

Yes, because he's literally done exactly that many times, and lost lawsuits over it.

Ok, prove that they are fake.

Do your own research, it's not like documentation of his bullshit is hard to find.

Project Veritas is in no way a reliable source on political issues. Both it and its founder have a long record of being exposed committing fraud.

Project Veritas has a long record of collecting large sums in judgement against those who make such libelous accusations without proof.

I couldn't find any lawsuits they've won, just them getting lawsuits against them dismissed.

And information about James O'Keefe pleading guilty to breaking into a Senator's office.

https://twitter.com/jamesokeefeiii/status/115837860224274022...

I think you have that backwards. Okeefe has paid 100k in damages to an ACORN employee, for libel.

One case, poorly phrased, is hardly a "long list".

More than zero, which is the number of libel cases that O'keefe as won and gotten a payout from. (this is opposed to a number which he hasn't lost, because plaintiffs didn't win, but he was the defendant. But these didn't give him any payout)

https://twitter.com/jamesokeefeiii/status/115837860224274022...

I understand you're trying to do character assassination. Thankfully, we have actual videos of google insiders.

I also trust my own eyes, I see which political side gets banned en masse by Google, Twitter, FB and other large Silicon Valley companies. Are you going to deny this too?

Can you provide a different citation than a publication known to generate fabricated claims bolstered by fabricated evidence? (I'm serious and taking you in good faith here for the record.)

Can you even name a publication that has not fabricated stuff? I'm not aware of any.

That's not my responsibity; I'm not making any claims.

Some people still cling to the 10-year old image of "good guy Google."

I used to be one of those people.

Reply on news.ycombinator.com