Data-Driven Hubris: The 2016 Reckoning
Like millions of other people during the final week of the 2016 election, I compulsively checked Nate Silver’s http://fivethirtyeight.com/ (by far the best thing about ESPN). Silver’s success in the last few election cycles helped establish polling science as a dominant form of journalism and by this election it became ascendant.
The shock of Trump’s election is no doubt tied to the titanic failure of our pollster-gurus to live up to their reputations. Silver himself emerged from this election relatively unscathed, as his models gave Trump a legitimate chance at winning the election. The data revolution he inspired, however, was heartily adopted by the liberal establishment, which used it as part of its shame-based rhetorical strategy. This form of reporting meshed nicely with “woke” late night talk show host poking fun at Trump and his electorate (usually reduced to caricatures). “These people think Trump might win! Ha! What doobs!” And their scorn for apostates was eventually directed even at Silver himself for not being triumphant enough. The Huffington Post for example accused him of all sorts of nasty things. Silver’s reply is perhaps my favorite moment of the election:
Alas. Lesson learned, right HuffPo? Probably not.
I hope you’re still with me, because I want to publicly shame a particularly galling example of Big Data hubris. Like many others, I was intrigued by the Slate collaboration with a startup called Votecastr. The premise was as such:
1). Smarty Pants People in skinny jeans did a bunch of polling before the election.
2). Technocrats in track suits assigned humans into data-mining categories.
3). As votes were cast, the techno-utopianists lifted the “embargo” on data and told us who was winning, live. As it happened.
And wow they did great, eh? According to their calculations, Clinton mopped the floor with Trump, HuffPo style. Imagine their horror when the Blind Seer of Thebes revealed the truth to the world.
For some reason, the people responsible for this disaster have left the evidence of their folly in full public view:
Let me conclude by proposing that maybe human beings are more complicated than algorithms can always predict. As much as Trump’s election offends me, I must admit that, odious as he is, he is a transcendent figure. By this I do not mean to compare him to Barack Obama’s ability to move the wheels of history. I simply mean that his appeal was not fully detectable by our computers and our institutional systems. He spoke to people about things the numbers couldn’t track.
The Democratic party’s supreme confidence in “demographics” was part of its undoing. The party and its surrogate institutions (like Slate) became utterly deluded by the previous success of its machinery. That led to a hubris straight outta Greek drama. The party was so sure of its math, it had no curiosity about the people behind the numbers. The liberal faith in numerology therefore correlates to its contemporary tendency to reduce the flyover states to a collection of bigotries.
So the post-election autopsy begins. As liberals try to recover in time for 2018’s mid-term elections, I hope they realize that human beings are not data sets, nor can they be categorized with such blunt instruments as “racist or not.” The KKK voting block notwithstanding, even if the voters Democrats lost this time are racist and sexist, they are not ONLY those things. What else do they care about, and might we speak to those desires as well?
