I, for one, welcome our new search engine overlords.
September 3, 2008 9:46 AM Subscribe
CorporateOverlordFilter: Does Google already know the winner of the next election?
I'm posting this question for a friend without a MeFi account:
Google is the preeminent profile data mining organization in history. They know what we blog, email, search, quote, buy and read. Google can track us by state, precinct and address. They're experts in finding patterns and inferring intent (the better to target us with ads). They can track billions of keywords per minute, and I assume they can distill thought trends and track them as they ebb and flow. Google may well have algorithms to extrapolate political decisions from information flow. (There are already published correlations between polling data and search data.) Our “private” searches and gmail contents are probably more candid and accurately predictive than answers to pollsters, even if the contents aren’t overtly political, and the data sample is orders of magnitude larger than traditional polling. The correlation between likely voters and internet users is probably high. So does Google already know who will win the election? If not now, how close to election day will we need to get before Google could determine the winner with 96 percent accuracy? And is it counterproductive to our political system for that information to exist?
I'm posting this question for a friend without a MeFi account:
Google is the preeminent profile data mining organization in history. They know what we blog, email, search, quote, buy and read. Google can track us by state, precinct and address. They're experts in finding patterns and inferring intent (the better to target us with ads). They can track billions of keywords per minute, and I assume they can distill thought trends and track them as they ebb and flow. Google may well have algorithms to extrapolate political decisions from information flow. (There are already published correlations between polling data and search data.) Our “private” searches and gmail contents are probably more candid and accurately predictive than answers to pollsters, even if the contents aren’t overtly political, and the data sample is orders of magnitude larger than traditional polling. The correlation between likely voters and internet users is probably high. So does Google already know who will win the election? If not now, how close to election day will we need to get before Google could determine the winner with 96 percent accuracy? And is it counterproductive to our political system for that information to exist?
I can't answer all these questions but I can say this: No amount of information can predict a career ruining gaffe such as the "Dean Scream."
Also, I think one should take into account the demographic that uses Google and how that demographic will largely be voting. Lets put it this way: My dad is 49, watches Bill O'Reilly religiously, and has absolutely no access to the internet. And that's the way he likes it. For every hipster fawning about Obama on his/her blog, there is a curmudgeonly old pro-lifer sitting at home wondering why their grandkids haven't called lately.
posted by Brodiggitty at 10:04 AM on September 3, 2008 [3 favorites]
Also, I think one should take into account the demographic that uses Google and how that demographic will largely be voting. Lets put it this way: My dad is 49, watches Bill O'Reilly religiously, and has absolutely no access to the internet. And that's the way he likes it. For every hipster fawning about Obama on his/her blog, there is a curmudgeonly old pro-lifer sitting at home wondering why their grandkids haven't called lately.
posted by Brodiggitty at 10:04 AM on September 3, 2008 [3 favorites]
Let me just add to what Airhen said: the elderly are 2-3 times less likely to use the internet than the average American, but something like twice as likely to vote (at least as against the youngest, vote-rocking eligibles).
posted by Clyde Mnestra at 10:04 AM on September 3, 2008
posted by Clyde Mnestra at 10:04 AM on September 3, 2008
Most people online don't use gmail. Msn, aol et al are surprisingly popular if you are the kind of person who hangs out at a site like metafilter.
That said: yes.
posted by shownomercy at 10:04 AM on September 3, 2008
That said: yes.
posted by shownomercy at 10:04 AM on September 3, 2008
I think it's pretty damn hard to generalize from the set of all google users to the set of all voters.
In addition, the election will likely turn on votes from a few voters in a few states. GEOIP technology is still pretty lame, so I don't see how you'd correlate these categories in these key areas. Remember the electoral college?
posted by jenkinsEar at 10:09 AM on September 3, 2008
In addition, the election will likely turn on votes from a few voters in a few states. GEOIP technology is still pretty lame, so I don't see how you'd correlate these categories in these key areas. Remember the electoral college?
posted by jenkinsEar at 10:09 AM on September 3, 2008
You're giving Google way too much credit here.
As with all elections, the vast majority of people are not very interested in the election; they do not go out and seek information on it. There's enough information on TV, from friends and family, from loud random strangers on the street - that making a decision doesn't involve a lot of research. These people are not going to spend their free time on the internet looking at political blogs, participating in political discussions, etc.
The internet is not even close at representing the US population, just look at any online poll. White males living in the coasts, above the poverty line, are likely to be overrepresented. Lots of users don't even know what Google is or what they use as their search engine; they type search terms into the address bar. Maybe it's Google, but for IE users (still the majority) it's Live Search. gMail isn't even close to being the most popular web-based email.
They elderly are unlikely to use the internet as much as others, but they're more likely to vote.
I'm not sure even the mighty Google algorithms can determine my political preferences based on, say, my Pokemon hobby.
posted by meowzilla at 10:15 AM on September 3, 2008
As with all elections, the vast majority of people are not very interested in the election; they do not go out and seek information on it. There's enough information on TV, from friends and family, from loud random strangers on the street - that making a decision doesn't involve a lot of research. These people are not going to spend their free time on the internet looking at political blogs, participating in political discussions, etc.
The internet is not even close at representing the US population, just look at any online poll. White males living in the coasts, above the poverty line, are likely to be overrepresented. Lots of users don't even know what Google is or what they use as their search engine; they type search terms into the address bar. Maybe it's Google, but for IE users (still the majority) it's Live Search. gMail isn't even close to being the most popular web-based email.
They elderly are unlikely to use the internet as much as others, but they're more likely to vote.
I'm not sure even the mighty Google algorithms can determine my political preferences based on, say, my Pokemon hobby.
posted by meowzilla at 10:15 AM on September 3, 2008
I don't doubt that Google could probably take a fair stab at estimating the popularity of candidates among strongly partisan voters. They could look at blogs, comments on blogs, even IP numbers behind comments on blogs in some cases, and probably come up with an interesting predictive voting map.
It's the uncommitted/not-highly-committed voters that are the problem. There are a lot of people who apparently make up their mind on election day, or change it. (At least, that's the conventional wisdom explaining why so many polls predicting Kerry would win in '04 were wrong. Ahem. But I digress.)
And there are people like my folks, who are interested in politics, have made up their minds, and are on the Internet, but are never going to blog about politics, or comment in a political blog.
posted by adamrice at 10:39 AM on September 3, 2008
It's the uncommitted/not-highly-committed voters that are the problem. There are a lot of people who apparently make up their mind on election day, or change it. (At least, that's the conventional wisdom explaining why so many polls predicting Kerry would win in '04 were wrong. Ahem. But I digress.)
And there are people like my folks, who are interested in politics, have made up their minds, and are on the Internet, but are never going to blog about politics, or comment in a political blog.
posted by adamrice at 10:39 AM on September 3, 2008
Google can't compensate for demographic distribution because it has no information about the people providing the samples, whereas pollsters should know the demographics of their samples and can compute how to compensate for their non-uniform selection.
And to add another data point, there are lots of voters who already decided which party to vote for long before they knew either of the candidates. These people have no need to participate in discussions because their mind is made up. When was the last time you were actually on the fence about the two political parties?
posted by meowzilla at 10:47 AM on September 3, 2008
And to add another data point, there are lots of voters who already decided which party to vote for long before they knew either of the candidates. These people have no need to participate in discussions because their mind is made up. When was the last time you were actually on the fence about the two political parties?
posted by meowzilla at 10:47 AM on September 3, 2008
And to add another data point, there are lots of voters who already decided which party to vote for long before they knew either of the candidates. These people have no need to participate in discussions because their mind is made up.
No need, maybe. But desire, plenty. Visit any political site and you will see a lot of the already committed spewing to friends and enemies alike.
In any event, I think the data would include site visits, not just those who post. I think if my site visits were tracked and used to generate my political profile, Google would either guess wrong or just explode from the dissonance.
posted by Clyde Mnestra at 11:12 AM on September 3, 2008
No need, maybe. But desire, plenty. Visit any political site and you will see a lot of the already committed spewing to friends and enemies alike.
In any event, I think the data would include site visits, not just those who post. I think if my site visits were tracked and used to generate my political profile, Google would either guess wrong or just explode from the dissonance.
posted by Clyde Mnestra at 11:12 AM on September 3, 2008
Considering how much of a failure google has been with many of its own services which are run by the top experts in their fields, I seriously doubt they could do better than a 50-50 guess. Remember Google Answers or that VPN/tunnel/accelerator service they launched? Or how about their big wireless plan? Their office suite has not even made a significant mark on MS Office.
Youre really overestimating these guys. They have no expertise in elections and their data samples would be anything but scientific. It would skew towards connected people who like to chat about politics on the computer.
posted by damn dirty ape at 12:43 PM on September 3, 2008
Youre really overestimating these guys. They have no expertise in elections and their data samples would be anything but scientific. It would skew towards connected people who like to chat about politics on the computer.
posted by damn dirty ape at 12:43 PM on September 3, 2008
No need, maybe. But desire, plenty. Visit any political site and you will see a lot of the already committed spewing to friends and enemies alike.
I think I touched on this in my earlier comment. Few people will spend their time going to these political sites. What you see there is a very vocal minority. These are the people who wear their political preferences on their sleeve.
The silent majority will never go to a political site or blog, when they've had their fill of political debate on TV and through informal discussion. Debating on the internet is hard work, stressful, and doesn't accomplish anything.
posted by meowzilla at 1:50 PM on September 3, 2008
I think I touched on this in my earlier comment. Few people will spend their time going to these political sites. What you see there is a very vocal minority. These are the people who wear their political preferences on their sleeve.
The silent majority will never go to a political site or blog, when they've had their fill of political debate on TV and through informal discussion. Debating on the internet is hard work, stressful, and doesn't accomplish anything.
posted by meowzilla at 1:50 PM on September 3, 2008
Predicting something like this would require a fair amount of work, plus lots of sophistication, with little certainty of success. Combined with the expected (probably minor) gain expected from picking the correct candidate ahead of time, they are probably better off donating strategically to both candidates.
In other words, they have better things to do than trying to call the election. They need to be on good terms with everybody anyhow.
posted by ghost of a past number at 2:09 PM on September 3, 2008
In other words, they have better things to do than trying to call the election. They need to be on good terms with everybody anyhow.
posted by ghost of a past number at 2:09 PM on September 3, 2008
No, Google does not know. If they did, they'd have put out a media release by now, framed in the context of "ha-ha we know this isn't 100% accurate but our data scans show x is gonna win", with the main purpose being to get their name in the paper and the blogosphere. The fact that a media release like that hasn't been put out shows that they don't know.
posted by Effigy2000 at 3:24 PM on September 3, 2008
posted by Effigy2000 at 3:24 PM on September 3, 2008
I understand that this is an older post, but I when I read your question I thought about the "Web bot". I don't assign it any worth personally, but I think it theoretically covers what you're asking about, although it is not a Google project.
From the site: "It's like changes in language precede large emotional events. The larger the emotional impact of an event, the more advance notice the bots seem to give. "
posted by Horatius at 1:41 AM on September 6, 2008
From the site: "It's like changes in language precede large emotional events. The larger the emotional impact of an event, the more advance notice the bots seem to give. "
posted by Horatius at 1:41 AM on September 6, 2008
« Older Help me keep track of my money when I'm on the go. | Need to help mom stay in the house. Newer »
This thread is closed to new comments.
I think you would also have to know how much time people spend on the Internet doing work, and therefore gathering information their not necessarily that interested in personally.
But I'm not statistician.
posted by Airhen at 9:58 AM on September 3, 2008