Here, though, it's being used to provide fake signals to a true machine-learning algorithm – Google's clicktrail-influenced ranking. So more of a 'complement' than an 'opposite'.
I’m not outraged about the specifics, i’m intrigued about the astroturfing. If this worked, it could and probably is being applied to all kinds of things...
If you search for the requested terms, most of the top links are spam blogs and other junk sites that describe him in glowing terms. It appears that they've successfully pushed most but not all of the critical articles down to the second or third page.
I think so, at least to some degree. Someone's certainly bet money on the theory that it works.
When I first saw this tweet, a search for "Dr. Paul Drago" (in an incognito window) consistently returned a glowing blog post first and the domestic violence article second. Hours later, those results are consistently reversed. Could be random, could be that people who saw this tweet are intentionally clicking the news article he's trying to bury.
Coming back to this from a few days later, the search "Dr. Paul Drago" doesn't turn up anything negative in the first dozen pages. Unfortunately for him the search bar is offering up "dr. paul drago domestic violence" and "dr. paul drago botched" as auto-complete options.
It's just "blackhat" SEO. Reminds me of a scheme in China where a company sold fake medicines and astro-turfed their way into appearing as a major supplier. So long as our view of the internet is governed by opaque algorithms with no oversight, gaming them will be profitable.
Security by obscurity ain't security, though. If they weren't opaque, at least we as a society would be better equipped to keep up with the algorithmic exploitation arms race.
I think, unfortunately, that is naive. While security through obscurity ain't security, there is something to be said for obfuscation of a system that people are trying to game.
That's the point, though: if a system is so fragile that anyone with knowledge of its inner workings can game it or otherwise exploit it, then it is not and was never secure (nor can it ever be while it continues to be opaque).
I say this enough to be a broken record, but transparency is a dependency of trust.
Well yeah, obviously, given that it ain't transparent.
> You are effectively saying poker would be better if everyone's cards were face up.
You are effectively saying that an ideal system is one that we'd have to treat like a poker game.
Even assuming the premise here holds true (that a transparent system will be more easily gamed by more people), that'd ultimately be better than the opaque case. The more people who are able to game a system, the less one individual can effectively game it for one's own individual benefit at the expense of everyone else in that system.
I thought of a better way to express this that might make sense to you.
In security, total transparency isn't effective. You want as much transparency as possible, but you need secrets for the system to work (usually passwords/certs/passphrases).
Now, there isn't a password/certs/passphrase in this context, so the secrecy is instead in the model.
Is this astroturfing? I thought the term referred to fake grass roots movements (Astro Turf being the brand name for that fake grass you sometimes see).
As a previous commenter said, this is more like grey/blackhat SEO.
Googling around I can't find anything that agrees with this but I see an overwhelming number of sites referring to the origin as related to the grass roots meaning with the popular origin from Lloyd Bentsen in 1985.
Not that it couldn't have just been overrun since that time I just couldn't find anything to support your claim.
IME it's not uncommon for successful surgeons to be fairly cold about the human body. Maybe that's not quite the same thing as "not valuing people", but ISTM it's on the same spectrum.
It's arguable that domestic violence is less cold and more passionate than the characteristics you're describing. I understand there are cold sadists, but speaking to the archetype.
Additionally, in the case of a bigot, the outcomes involving subjects of prejudice may be less meaningful to them even at a subconscious level. Coldness and bigotry can coexist, and certainly produce emergent results.
Also, that the best outcome of using surgeon who has a history of domestic violence is that they're so disassociated from the world, that beating a loved one doesn't matter is a really strange argument. Perhaps the argument is there isn't a better way, but damn.
I used to not be sure about robot surgeons, now I'm pretty convinced.
In my experience working with surgeons, while being a bad person is not a good predictor of skill, neither is being a good person. The most absolutely talented and innovative people in medicine are a pretty good mix of kind, empathetic, thoughtful people and BPD / narcissistic / sociopathic people.
Are you kidding? Someone who can't control their anger isn't someone you want poking around your insides with a scalpel. If such a person makes an error during an operation, do you really think they'll be transparent about it?
Should a person be banned from working because they physically assaulted someone. Seems to be that lots of people think so, despite that fact that many more people than are convicted commit these sorts of crimes. Didn't take much more than to type the name into Google News and click on the link on the page.
If I told you that the 3M Company had 2192 lawsuits[1] of all types filed against them in the the last 4 years, what does that tell you? Are they a terrible company getting sued at lot or a incredibly nice company that barely ever gets sued for a company that large?
Plastic surgeons are among the most sued professions. To know whether 11 lawsuits are significant, you'd have to know the rate at which plastic surgeons are sued, how many surgeries this guy did, what period those 11 surgeries represent, and possibly even weight it with the difficulty of the surgeries (because if he's taking on high risk surgeries that other doctors decline, he'll have more failures). Maybe 11 lawsuits in a 4-year period is an excellent track record for someone doing thousands of high-risk surgeries in an ultra-litigious medical specialty.
The first link says the North Carolina Medical Board just pulled his license. And a "nationally recognized and board certified plastic surgeon" said the doctor had no training to perform plastic surgery (only general surgery as all doctors receive).
So I'm not sure why you're trying to make some artificial point in isolation about lawsuits being meaningless, other doctors clearly think badly of his practice of medicine, to the point that they've stopped him doing so.
Doing that just ~10 times (with independant users/profiles) will probably eliminate those results when the ML model is next trained. Assuming this Dr isn't a celebrity, he probably doesn't get more than a few searches per month, so 10 in the same day will be a massive signal.
Note to person behind this: It will have even more impact if each turker is instructed to click only one of the desired links. Clicking "back" to a google search result is a strong signal that the page you clicked didn't answer your query.
Title is incorrect. This is not on Mechanical Turk. Tasks like this are not allowed on MTurk as per TOS. The original tweet author says so on the thread.
>heads up that this is NOT from @amazonmturk! black hat SEO/shady reputation management techniques like this are strictly against amzn's policy. this is from a competing micro task site
The title makes it seem like intentional click-bait, given this.
In the past, Google claimed to not use clickstream data because of how easy it is to manipulate. Have they changed their tune or are they just targeting Bing?
I did it a few times. When I google Dr. Paul Drago. I do seem some of the articles. That may be because I've click a few of the links and have done the search a few times. Also The https://www.lipstickalley.com/threads/plastic-surgeon-paul-d... link also appears on the second page now.
It's probably worth noting this guy has a very unique name, therefore negative news stories about him are much more likely to go to the front page of results, since few other people with the same name would get any publicity.
69 comments
[ 3.1 ms ] story [ 136 ms ] threadLet's be outraged about the correct problem, please ...
When I first saw this tweet, a search for "Dr. Paul Drago" (in an incognito window) consistently returned a glowing blog post first and the domestic violence article second. Hours later, those results are consistently reversed. Could be random, could be that people who saw this tweet are intentionally clicking the news article he's trying to bury.
(as long as people can get away with it,) gaming them will be profitable.
I say this enough to be a broken record, but transparency is a dependency of trust.
You are effectively saying poker would be better if everyone's cards were face up.
Well yeah, obviously, given that it ain't transparent.
> You are effectively saying poker would be better if everyone's cards were face up.
You are effectively saying that an ideal system is one that we'd have to treat like a poker game.
Even assuming the premise here holds true (that a transparent system will be more easily gamed by more people), that'd ultimately be better than the opaque case. The more people who are able to game a system, the less one individual can effectively game it for one's own individual benefit at the expense of everyone else in that system.
In security, total transparency isn't effective. You want as much transparency as possible, but you need secrets for the system to work (usually passwords/certs/passphrases).
Now, there isn't a password/certs/passphrase in this context, so the secrecy is instead in the model.
As a previous commenter said, this is more like grey/blackhat SEO.
So creating fake or useless media attention to distract or cover up something you don't want people to see.
Not that it couldn't have just been overrun since that time I just couldn't find anything to support your claim.
FWIW, whoever he hired to do this does seem to also be doing stuff that more traditionally falls under the "astroturfing" umbrella:
https://www.quora.com/What-are-some-facts-about-Dr-Paul-Drag...
Additionally, in the case of a bigot, the outcomes involving subjects of prejudice may be less meaningful to them even at a subconscious level. Coldness and bigotry can coexist, and certainly produce emergent results.
Also, that the best outcome of using surgeon who has a history of domestic violence is that they're so disassociated from the world, that beating a loved one doesn't matter is a really strange argument. Perhaps the argument is there isn't a better way, but damn.
I used to not be sure about robot surgeons, now I'm pretty convinced.
If the surgeon is prone to violence fits as you claim he might be, wouldn't it show on his track record as a surgeon?
It's a good thing no one said that then!
Should a person be banned from working because they physically assaulted someone. Seems to be that lots of people think so, despite that fact that many more people than are convicted commit these sorts of crimes. Didn't take much more than to type the name into Google News and click on the link on the page.
- "Accused doctor wasn't certified in plastic surgery"[0]
- "11 women say plastic surgeon botched jobs"[1]
- "Plastic surgeon - Paul Drago - Plastic surgery from hell"[2]
I'd say the accusations here are more than just DV.
[0] https://www.wcnc.com/article/news/local/accused-doctor-wasnt...
[1] https://www.courthousenews.com/11-women-say-plastic-surgeon-...
[2] https://www.lipstickalley.com/threads/plastic-surgeon-paul-d...
If I told you that the 3M Company had 2192 lawsuits[1] of all types filed against them in the the last 4 years, what does that tell you? Are they a terrible company getting sued at lot or a incredibly nice company that barely ever gets sued for a company that large?
Plastic surgeons are among the most sued professions. To know whether 11 lawsuits are significant, you'd have to know the rate at which plastic surgeons are sued, how many surgeries this guy did, what period those 11 surgeries represent, and possibly even weight it with the difficulty of the surgeries (because if he's taking on high risk surgeries that other doctors decline, he'll have more failures). Maybe 11 lawsuits in a 4-year period is an excellent track record for someone doing thousands of high-risk surgeries in an ultra-litigious medical specialty.
[1] Completely made up that number.
So I'm not sure why you're trying to make some artificial point in isolation about lawsuits being meaningless, other doctors clearly think badly of his practice of medicine, to the point that they've stopped him doing so.
https://eu.greenvilleonline.com/story/news/crime/2018/11/26/...
Notice the EU in the link so o the results will vary per region.
Doing that just ~10 times (with independant users/profiles) will probably eliminate those results when the ML model is next trained. Assuming this Dr isn't a celebrity, he probably doesn't get more than a few searches per month, so 10 in the same day will be a massive signal.
Note to person behind this: It will have even more impact if each turker is instructed to click only one of the desired links. Clicking "back" to a google search result is a strong signal that the page you clicked didn't answer your query.
>heads up that this is NOT from @amazonmturk! black hat SEO/shady reputation management techniques like this are strictly against amzn's policy. this is from a competing micro task site
The title makes it seem like intentional click-bait, given this.
In the context of crowdsourced labor, it absolutely implies Amazon.
I made sure to click on them.
His record appears clean in the first page.