Slashdot Mirror
When Google Got Flu Wrong
ananyo writes "When influenza hit early and hard in the United States this year, it quietly claimed an unacknowledged victim: one of the cutting-edge techniques being used to monitor the outbreak. A comparison with traditional surveillance data showed that Google Flu Trends, which estimates prevalence from flu-related Internet searches, had drastically overestimated peak flu levels. The glitch is no more than a temporary setback for a promising strategy, experts say, and Google is sure to refine its algorithms. But with flu-tracking techniques based on mining of web data and on social media taking off, Nature looks at how these potentially cheaper, faster methods measure up against traditional epidemiological surveillance networks." Crowdsourcing is often useful, but it seems to have limits.
Computer modeling is a powerful technology that should not be underestimated.
However, it should also not be overestimated.
When the "real world" has millions of convergent factors responsible for an event, computer models can sometimes capture a few thousand. Based on those, a simulation is created that suggests a certain outcome. But it may be using less than 1% of the necessary data.
This is like making architectural models out of child's blocks and then being surprised when the building falls down after it is eventually made. There are issues of scale in addition to data that can reveal periodistic or epicyclic patterns that cannot be modeled in a linear method.
They should subtract out a factor based on how much the flu is being talking about in the media.
In short, a system that learns from abnormal circumstances will no longer work as well under normal circumstances. This year's flu outbreak didn't follow previous models, so Google's application of those models was inaccurate... but we'll blame Google for it anyway, and cast shame upon them for being so terribly wrong.
Of course, the article is much better, delving into other systems that also predict and monitor flu outbreaks, and why they were or were not correct. TFA is really about the difference between traditional reporting sources (as from doctors' offices) and newer data-mining approaches (harvesting from searches and Twitter).
Screw you, Slashdot.
You do not have a moral or legal right to do absolutely anything you want.
This is probably because people will update their social media sites with claims of having the flu. If they actually had the flu odds are they would not have the strength to even do that.
The real flu is pretty terrible and people often think they have it when they have a minor cold.
This is borderline conspiracy think. Scientists of all stripes want their predictions to be testable, with minimal error bars and as accurate as possible.
Why stop there? Just arrest people for non-conforming behavior.
Why stop there? Just arrest everyone.
(Disclaimer: My 401k is all in for-profit prison systems.)
If Slashdot were chemistry it would look like this:Cadaverine
Actually, that's kinda the goal. When it comes to the expenditure of time and money, if you don't come in with a Chicken Little, people are just going to ignore you. With the Chicken Little, you get people to fall in line and the effects of major epidemics or problems are mitigated.
Slashdot-friendly example: Today, people will say that the Y2K issue was completely blown out of proportion. Airplanes didn't fall out of the sky, bank accounts were there on Jan 1, 2000, and everything was just fine. Of course, that ignores the teams of coders working in even-then-archaic coding languages to adapt old software to work beyond their expected lifespan. Who knows what Y2K would have been had we just done nothing, but we're all better off with the purse-string-holders getting concerned.
It's only a problem when it causes people to panic (like yelling "fire" in a crowded theater, then defending yourself with "Well, it got them to think about fire safety, didn't it?"). If it just causes Cleatus Dipshit to wash his hands more and cover his goddamn mouth when he sneezes, I'm okay with it. If it causes people to sell their houses and empty their bank accounts to buy underground bunkers and canned goods, then we have a problem.
Of course, there is also the issue of fraud when it comes to public grant money. I don't like the idea of a scientists who are knowingly exaggerating their findings taking grant money away from those who aren't.
What political party do you join when you don't like Bible-thumpers *or* hippies?
No. You only hear in the media about epidemic and pandemic estimates of the upper range. The prediction "we'll have 30,000 deaths in 2013 due to the normal flu" wouldn't make any headlines, because every year, about 30,000 die after getting sick with the flu. But most predictions of epidemics and pandemics are exactly like this -- it's just the expected behaviour. There is a big difference between the average estimates coming from the scientists and the single highest estimates reported in the media. And of course, "everything is normal" is no news, thus it doesn't get reported that often. Information is the inverse of probability, and reports about highly improbable events have higher information content than reports about average events. Highly improbable events happen and contradict our expectations, and thus it is important to report them. Normal events happen, but we were expecting them anyway, thus there is no point in reporting them. Your "ALWAYS" is probably more due to confirmation bias on your side than anything else.