Could You Google Bomb Google Flu?
from the would-you-want-to? dept
Google got a lot of attention recently for the launch of Google Flu Trends, which looks at aggregate data on searches related to the flu, to see if it can act as something of an early warning system for where there are flu problems. It’s an interesting use of the data, and it will be worth watching what else can be done with this sort of data over time. However, Ed Felten raises an interesting question: can Google Flu Trends be manipulated? The idea is that, right now, it may be accurate, but the very fact that people know Google is tracking this information, could create incentives to game that info — in the same way people have tried gaming Google in other ways for years, using tricks such as Google bombing. While you might not think there would be that many reasons to manipulate Google Flu Trends, there could be reasons to do so. Google is being somewhat secretive of how Flu Trends is set up, so perhaps that makes it more difficult to manipulate, but it does point to an interesting issue in using data in this manner. As soon as you’ve set up a system to measure the data, and made that public, is the data still reliable?
Filed Under: google bomb, google flu
Companies: google
Comments on “Could You Google Bomb Google Flu?”
either side?
Has either side got you on the right. Not the right but the right ?
Outbreaks
Well you can break the data, but the long tail test will be to see what it looks like in comparison to actual flu outbreaks. I mean to ‘really’ manipulate it you could start spreading the flu virus and then watch the google results while sitting in the comfort of your evil lair.
Plus if google trends were that reliable then comparing the correlation of searches for “Zombies” and “Shotguns” in comparison to actual zombie outbreaks reveals accurate patterns. But I know that can’t be right because its currently showing search traffic doubled in the last 12 months for “Zombies” in Australia…
Oh god wait, I can’t ^^oia$hdFsjh;as&oIHdlkhl
The diarrhea runs brown in Beverly Hills
I just typed this one into google search:
google flu 90210 diarrhea like a river
(90210 is the zip code for Beverly Hills, California. God help me – I need a life)
Certainly
One could determine the magnitude of the outbreak or its location but not both with any precision.
zombies???
you have to destroy the heads…. where is my shovel???
Particle/Wave Duality extends beyond quantum physics?
Upon reading the summery, the first thing that came to mind was how similar data mining is to the Particle Wave Duality Paradox. In short (and I don’t claim to be a quantum physicist here) particles can exhibit wave like behavior, and waves can exhibit particle like behavior. The determining factor is whether they are being directly watched at the time.
Data mining (might) have a similar paradox. The information is there, and truthfully accurate right up to the point that we start looking at it. But, by simply looking at the data, we may be corrupting the future data. Keeping the fact that the data is being examined private, or making it public, doesn’t change the fact that the data will be corrupted. It only changes to what extent the data will be corrupted.
Interesting…
Re: Particle/Wave Duality extends beyond quantum physics?
ShaneC… I think you’ve just described the Hawthorne Effect — http://en.wikipedia.org/wiki/Hawthorne_effect
Re: Re: Particle/Wave Duality extends beyond quantum physics?
“ShaneC… I think you’ve just described the Hawthorne Effect”
or (possibly) the Heisenberg uncertainty principle
http://en.wikipedia.org/wiki/Uncertainty_principle
Re: Particle/Wave Duality extends beyond quantum physics?
I stopped reading at summery.
Re: Re: Particle/Wave Duality extends beyond quantum physics?
Summery is like wintery but hotter. :0
As soon as you’ve set up a system to measure the data, and made that public, is the data still reliable?
No. It’s called the Heisenberg Uncertainty Principle. Normally, it deals with only the Second Law of Thermal Dynamics but in this case it serves as a reminder that any non-protected public data instantly influences the outcome of that data. On the other hand, there are cases where this isn’t the situation. With public encryption like RSA and AES, the algorithms are known to all but they can’t be decrypted without the private key.
Google Bombing is akin to brute forcing that key. We can only sit back and say the trite but astute lines “Only time will tell.”
Interesting concept. We will be using the Google data to lauch newspaper ads reminding people that Tamiflu can help recovery if started early on. In our Urgent Care clinic we obviously see the flu coming, but often patient wait too long, so the Tamiful is ineffective.
If you download the Google historic data you see a huge peak in 2003. Recall that this was the year of the vaccine shortage due to contamination, but not a huge year for flu cases. Just shows other factors (media) can affect results.
What gets measured, gets manipulated.
newsflash: flutrends not a google invention
This kind of misinformation makes me angry. Please stop positioning Google Flutrends as a “Google invention” and regurgitating Google press releases.
NEWSFLASH: The idea to look at search data as early warning systems for flu outbreaks is not a Google invention, but was actually already proposed over 3 years ago (published 2006), by researchers from the Centre for Global eHealth Innovation and U of T.
Eysenbach G. Infodemiology: tracking flu-related searches on the web for syndromic surveillance. AMIA Annu Symp Proc 2006:244-248
http://www.pubmedcentral.nih.gov/articlerender.fcgi?tool=pubmed&pubmedid=17238340
Eysenbach G. Infodemiology and Infoveillance: Framework for an Emerging Set of Public Health Informatics Methods to Analyze Search, Communication and Publication Behavior on the Internet
J Med Internet Res 2009;11(1):e11
URL: http://www.jmir.org/2009/1/e11
The Virus Chasers
http://www.cihr-irsc.gc.ca/e/35061.html
CIHR Newsarticle (2007) about the infodemiology / infoveillance work at the Centre for Global eHealth Innovation in Toronto