After their stunning failure to accurately predict the results of the presidential election, the press, pollsters, pundits, and political consultants are all due a ferociously and meticulously conducted autopsy. That will take time.
There are some early lessons to learn. But our rumination and recrimination, the hubris of self-pity, of the need for certainty in an uncertain time, will push us in unproductive directions, too. It would be easy to learn the wrong lessons. Let's focus on learning the right ones instead. For instance:
Wrong lesson: Don't believe the polls. The data is too skewed.
Right lesson: Don't trust only in polls. Data is valuable, but imperfect and contingent on our assumptions more than we want to believe.
Not even the Republican Party's internal targeting mathematics envisioned a universe in which white women with college degrees voted for Donald Trump at the rates they did, or that Democratic turnout in cities would be insufficient to overwhelm the turnout from rural and exurban white voters.
Think of it like this: If you're asked to describe the contents of a box, it helps to know its weight, volume, and size. Shake it a little. How does it sound?
This is what pollsters do. They use increasingly refined models to guess the box's dimensions. But if we're just totally wrong about the dimensions of the box, or if we hope the box contains a diamond when it in fact it contains a pumpkin, those errors will throw everyone off.
Apply that to the election. Pollsters' models reflect a consensus about the way the world works, or in this case, the "feel" of politics.
I've always felt that the divide between pundits and data journalists was artificial. Both operate off of educated guesses based on the consensus, their own mental shortcuts, their partisan predilections, and even their wishes. Some find more comfort in hard numbers than they should. Usually, we are too quick to extrapolate based on an anecdote.
Pollsters and pundits got the basics right. They understood that working-class whites would vote for Donald Trump by significant margins, that Hispanic and black voters would overwhelmingly favor Hillary Clinton, that the gender gap would be significant, and that Clinton would do well in cities and Trump would do better outside of them. The polls correctly assessed the magnitude of Clinton's support among millennial voters, and of Trump's support among older voters. Adjusting assumptions a few degrees in either direction, and pollsters would be crowing — rather than eating it.
The bottom line is this: Silly forecasts based on abstract models should be out. More contingent, measured analysis has a place at the table.
Wrong lesson: Technology disrupts the quest for the truth.
Right lesson: Technology is truth-agnostic.
For months, analysts knew for certain that technology enabled people to see only what they wanted to see and reinforce their own points of view. Facebook, in particular, transmitted false information more quickly than Macedonian tweetbots could disseminate it, and more widely than any Russian propagandist could ever dream.
Technology companies have become primary sources of misinformation. They don't generate the news, though, and they don't generate the preferences that predispose people to believe it. They exacerbate — even "weaponize" — the effect of these filter-bubbles, and we can, and should, urge companies with a sense of civic responsibility to reform.
But Facebook isn't the cause of our political polarization any more than Airbnb is the cause of the housing crunch in cities. Forcing people to hear outside these echo chambers is really hard because that's not what people are inclined to do by nature.