First of an occasional series where I’m going to copy over some of my old posts from LiveJournal. This is from 1st of May 2013.
I tried to fit this into a couple of tweets a minute ago and that didn’t work so…
I went to see Nate Silver of FiveThityEight fame tonight at an IQ2 event. It was good, albeit a little odd since his speaking section was much shorter than the time given over to questions. Still, in a major change to usual IQ2 events, all the questions were sane and provoked interesting answers.
One of the questions, following the “Target story“, was about the potential use of purchasing data by government to track things like disease outbreaks. If a lot of people in a certain area are suddenly buying flu medicine what does that mean for the CDC? Of course, today all that data is held behind corporate firewalls and opening that up to government use, even anonymously, is a slippery slope to CISPA or worse. Now, if that data was instead held by individual users in a VRM ecosystem and they chose to allow the government access to it (“notify the government anonymously when I buy these kinds of medicines”) then that would be a whole different matter.
Another question was about polling bias. As alobear and I have agreed for years there is a natural bias in polling for the kind of people who are willing to do polls. As Mr Silver said tonight, less than 40% of American’s now have landline phones, so if your polling company is still relying on robocalling only landlines you’re mostly going to get annoyed old people (which may be fine if you’re looking for Republican voters, but I digress). The questioner asked if any polling companies can be trusted to provide truly accurate data. Mr Silver said, as alobear has previously recounted to me, that they try to control for a lot of things like age / gender / race / socioeconomic status / etc but in the long run it still comes down to people who are prepared to fill in an online form (mostly younger people) or go through a telephone interview.
Now, here’s the interesting bit I thought up on the way home. If there’s a tight correlation between what we buy / what we watch / where we go / the people in our social graph / etc and how we are likely to vote (and I’m sure there is, the ecochamber is real) then what if that could be used, via VRM, instead of polling? Since people have been shown to lie to pollsters (c.f. Bradley Effect) perhaps this would be even more accurate?
Of course, the next step on from there is naturally why have elections at all? If we can look at people’s everyday lifestyle habits and from that pick out their likely voting choices then it would be feasible to simply elect politicians by looking at the data of people in their constituency. At a further level of granularity, if it were possible to devise with some accuracy people’s likely opinions of different political planks then the system would automatically elect someone standing with commitments to the planks most agreed with by people in their area – perhaps leading to more just two people standing in each constituency in the US. At the very finest level it might even be possible to enact policy by simply studying how people act – doing away with politics completely (if someone would like to write a paper on this I’m happy with an “inspired by” credit 😉 ).
Naturally I’m not taking any of this seriously. It starts off by relying on a lot of things (universal VRM, people opening their data, excellent prediction methods, etc) and ends by removing the universal right to vote in a democracy. It does however make one think about how interesting things are going to get as there is more and more data on each of us “out there” for people to use.