Slashdot Mirror
What Would Google Decide?
Gary Stock, the guy who invented the Googlewhack, tried a bit of Google election predicting last night. Using a methodology that is entirely indefensible, and which he does not try to defend, Stock asked Google to call the results on Michigan's five referendum questions. The result: Google's answers to two questions were spot-on, two questions were answered correctly but underrepresented the 'yes' vote, and one question was reversed. An 80% accuracy rate has got to beat any number of the pollsters and pundits who have been shouting at us since last August, no?
People who care enough to blog, care enough to vote. The real answer to this is that he's found a way to construct a poll made up entirely of likely voters, and NO non-voters. I suspect that this 80% accuracy rate will get better as time goes on (and the older Luddites die). I also suspect his accuracy rate would have been somewhat less say, had he used this technique in the 2004 Presidential election (because federal country-wide elections would beat the signal to noise ratio).
SJW: a person who perceives an injustice, and while correcting it, commits a greater injustice.
Shouldn't elections be determined when the ballots are cast, not when they're counted? Or maybe when voters decide how they're going to vote?
This surprises me not at all. Usually the vote follows wherever the majority of marketing money is spent. Marketing works, that's why people spend so much money on it. Google results probably mirror the amount of marketing pretty well, although if one group ignored online marketing or concentrated on one other form, it might be off.
I'm actually very discouraged by the Michigan elections, not because I think all the votes went the wrong way, but because 99% of the people I talked to about an issue could repeat what they heard in a television ad, but had obviously not thought the issue through at all beyond that. I was 100% right in my predictions of the election, based simply on the TV ads.
You're assuming everyone votes on the day of the election. Some states offer "early voting" up to a couple weeks before the voting date, and all of them offer absentee ballots.
Personally, I think voting shouldn't be a single day but should last for eleven days; for example: from the first Friday in November until the end of the Monday eleven days hence, with "the count so far" results being published at noon on days 4 (Monday), 7 (Thursday), and 10 (Sunday), with final (as much as possible, anyway) number announced at noon the day following the close of the polls (Tuesday, day 12). This would provide a whole lot more incentive for folks who haven't voted yet to get off their lazy buts when the preliminary numbers are published and their candidate/issue is behind. I bet we'll get 80+ percent turnout on most elections this way.