Computer Says No
Three days ago, after Donald Trump had progressed from generalised bile to actually insulting voters individually, his odds of becoming President of the United States fell to 3.6%. At least, that was the interpretation of immediate polling reactions by ‘the Model’ on the political statistics website FiveThirtyEight, based on the assumption that the election was held today.
FiveThirtyEight founder Nate Silver revolutionised politics in 2008 when he used aggregated polling and demographic data to correctly predict the electoral outcome of every single senate race during the American election. In the eight years since then, FiveThirtyEight has become the Oracle of Delphi for American politics junkies. It’s not always right (though because it only deals in probabilities, it’s technically never wrong), but it’s been highly successful, and for many of us the outputs of the Model have taken on the aura of gospel. FiveThirtyEight takes no moral stances, the Model doesn’t tell Americans that they shouldn’t elect Donald Trump. It just says that they’re probably not going to.
3.6% odds of becoming President are better than most of us can hope for, but it’s an unprecedented poor showing for a major party nominee. The media smelled blood, and articles started suggesting that Trump might as well give up now and save himself a few million dollars in advertising. Just today Trump was asked by CNBC reporters how he could possibly close his polling gap, to which he awkwardly replied:
“Just keep doing the same thing I’m doing right now. At the end it’s either going to work or I’m going to, you know, I’m going to have a very, very nice long vacation.”
In fact, Trump’s FiveThirtyEight odds have bounced back, and look relatively better when weighted for economic and historical data. But there’s the Model and then there’s our perception of the Model, and for most people anything under a 10% probability might as well be discounted completely. It’s because of perception that polling numbers don’t just reflect our preferences, they can shape them and become self-reinforcing. No self-respecting Trumpian wants to tie themselves to a Loser, and 3.6% has Loser written all over it.
While I enjoy watching Trump get his just desserts, there are things about all this that I find uncomfortable. Whether or not FiveThirtyEight is on the mark, how should a person respond to being told that they have almost no chance to succeed at their goals? It’s a question we’re going to have start asking ourselves, because we are all in the grip of Big Data now, and Big Data is honest like fluorescents in a bathroom.
It’s a strange thing that at a time where people feel increasingly uncertain, actual uncertainty is being purged from our lives. It’s not just political polling. Big Data is effecting a quiet revolution, where the fog of half-arsed estimates, hope, paranoia or boundless self-confidence in which we traditionally operate is being swept away and replaced with cold numbers, if we choose to look. And perhaps, we can’t handle the truth.
As a writer, I use database sites like Duotrope and the Submission Grinder. They’re a useful tool for tracking which magazines I’ve submitted stories to, but they also offer statistical feedback based on the thousands of data points fed in by users. So now I have the dubious pleasure of knowing that my chance of having a piece of fiction accepted at Clarkesworld is 0.08%. At literary darling magazine Tin House, it’s a 0.22% acceptance rate, and I can expect to wait 235 days to hear the bad news. That’s cream of the crop, of course. If I turn to Aurealis, an Australian sf magazine which pays only 10% what Clarkesworld does, I’d be in the money a whole 8.89% of the time.
Luckily for me, I set my heart on becoming a writer back in the Fog Years, where on my most pessimistic days I couldn’t have grasped just how many other versions of me were out there trying to achieve the same thing. And since then I’ve had enough little victories that I can buttress my self-confidence with the knowledge that I have sometimes succeeded, against the odds. As most people sometimes do, some of the time.
But what do figures like this say to anyone just starting to follow their ambitions? That whatever their dreams of greatness, they’re probably going to fail. And as technology advances, probability prophecies are spreading into all aspects of our lives. I can buy a DIY genetic testing kit and find out that I have an 80% chance of developing diabetes. Government ‘nudge’ units are predicting my choices and manipulating them for the better. Our futures can feel more and more out of our control.
Previous cultures understood this feeling — they called it Fate. And if you resisted it, the Ancient Greeks might add, then that was Hubris, and you would be lucky if you got off no worse than being turned into a spider. Accepting your fate was a coping strategy for a tough world.
But the world isn’t supposed to be quite as tough any more. My generation was brought up to believe that we could make our dreams come true. ‘Dream Large’ was even the motto of my university. But it’s increasingly hard to balance that optimism with the self-knowledge that the information age has gifted us.
It’s not that self-knowledge isn’t useful. Big Data tells me which movies I’ll like and it’s usually right. It can advise me on health check-ups, tell me the cheapest time to buy a gym membership, or which crops to plant and when. As long as I’m happy to stay well inside the bell curve, it can help me in all sorts of ways. But if I’m a talented young athlete wanting to go for Gold in an Olympics, then from now on some version of the Model will be there to tell me just how very unlikely that is.
None of this means humanity is any less amazing than we were before. Statistics can never say never, but they can be a major buzzkill. Tell someone that their goal is tough, and they’ll pump themselves up to meet the challenge. But the more specifically you can predict their failure, the harder it is to keep a brave face. Reality TV told us we could all be stars, but a future Guidance Counsellor algorithm might give a hopeful pop-singer the more likely news — 99.9% of failure. So why bother?
And that’s a problem, because our society thrives on blind optimism. The people who pour their savings into new businesses, or set sail towards an empty horizon. We take risks because we’re brave, or we’re dreamers, but let’s be honest — we also take risks because we routinely underestimate our chance of failure. The more clearly we can see that possible failure, the more hesitant we will become. In the end it might only be the super-rich or delusional who keep Dreaming Large.
Which brings us back to Donald Trump.
3.6% is not 0%, but it changes the way you approach the game. Probably half a dozen Presidential candidates have sleepwalked their way through similar odds in the past, but Trump doesn’t have the luxury of ignorance. And when all precedent suggests that you will probably fail, then it’s logical to reach for the unprecedented. So while it would be nice to think that Trump represents the high-watermark of delusional politics in America, maybe our society is setting itself up for a situation where more losing politicians will see an incentive to try for destabilising, high-risk strategies that could defy the Models that chain them.
And what about the rest of us? If the Model says our favoured candidate has a 3% chance of winning, will we even bother to vote? Sure, lots of us will go through the motions because we think it’s the right thing to do (though less of us than thirty years ago, which may also have something to do with our growing understanding of how statistically weak a single vote is). But sooner or later, if an election seems to be a foregone conclusion then people are going to stop bothering to show up. At which point I suppose the Model just adjusts for lowered participation rates, and the game continues.
I love science fiction, and “a good science fiction story should be able to predict not the automobile but the traffic jam.” So here’s a piece of prediction of my own — that beyond apathy, a generation told that the Model foresees their every move might turn to contrariness as a philosophy to live by. Lie to pollsters. Boycott the census. Act randomly.
There are signs this is already happening. “People in this country have had enough of experts,” said UK Brexiter Michael Gove to much derision, but defying the experts is a way of reasserting free will in a society that seems more and more deterministic.
By historical standards, the degree of polling we use is very weird. The will of the people is supposed to be discovered in elections, not mapped out three months beforehand. Since Donald Trump he is whipping up distrust of the whole electoral process, he could do worse than conjure resentment against Big Data itself. Computer predictions only reflect what we tell them, but the more pervasive they become, the more they can feel like constraints.
I don’t think Donald Trump is going to be President. But who knows what will happen down the road. Another ten years of the Model whispering in our ear, and we might find that we want to defy our fate. Even if, like the Greeks, it’s ultimately to our detriment.