Good point. I’m thinking out loud now… To my knowledge those consequences are for the most part only being tracked by the companies using Big Data, not being shared so the research community can research and test. So ultimately they are not being checked for casuality and statistical significance, but for return on investment (ROI). For example, the cost of tossing ads in front of people could be so low that it doesn’t matter that the prediction doesn’t reach statistical significance, it still gets enough clicks to make money. There could still be unidentified confounding variables that don’t matter because people who (made up example) own more than two cats are more likely to drink soy milk. Giving a person two cats doesn’t suddenly make people drink soy milk. But targeting ads and posts about soy milk to cat owners does get more clicks. The thing is, since the data and algorithms are proprietary, the people who could figure all this out and publish don’t see it. And the people who do see it are judging by other criteria. (And under NDAs).
Yes. Good point. Which brings us back to our “maybe big data is currently being over-hyped” hypothesis.
Yes, and that’s the ideal case. Some polling operations are partisan and skew their results to shift the polling averages, and the people being polled on one side or the other of an election have been gaming it too. They’ll purposely give dishonest answers or refuse to participate. Then after the election, that side will say, “See, the polls always underestimate our candidate!” I mean, humans are impossible to study scientifically if they know they are being studied and decide to mess with you.
Indeed, the polarization of politics into “this team or that team” has been driven in part by such forced-choice polling and the media’s relentless reporting on it.
Just came across “The Averaged American” by Sarah E Igo which talks about this at length: how our very conception of “the public” has been shaped and skewed by surveying.
Interesting point. It’s one of those basic assumptions that we rarely question. The reality is that there is always more that unites us than divides us, but it doesn’t make such a good story.
There’s something very strange about a system that runs the whole world under the assumptions that (a) we are irreducibly divided into two opposing camps with one winner and one loser, and (b) once one camp or other wins we will be whole again.
