We can seemingly draw speculative guesses out of thin air -- literally. From all the noise that surrounds us, we recognize potential patterns and infer significance. Scientists call them hypotheses. Artists call them artistic inspirations. Entrepreneurs call them innovations.
Whatever the label, we’re not exactly sure what happens. Mihaly Csikszentmihaly (which, in case you’re wondering, is pronounced Me-high Cheek-sent-me-high) explored where these hunches come from in his fascinating book, "Creativity: Flow and the Psychology of Discovery and Invention." But despite the collective curiosity about the source of human creativity, the jury remains out. The mechanism that turns these very human gears and sparks the required connections between our synapses remains a mystery.
We’re good at making hunches. But we suck at qualifying those hunches. The reason is that we rush a hunch straight into becoming a belief. And that’s where things go off the rails. A hunch is a guess about what might be true. A belief is what we deem to be true. We go straight from what is one of many possible scenarios to the only scenario we execute against. The entire scientific method was created to counteract this very human tendency, forcing rational analysis of the hunches we churn out.
Philip Tetlock’s work on expertise in prediction shows how fragile this tendency to go from hunch to belief can make us. After all, a prediction is nothing more than a hunch of what might be. He referred to Isaiah Berlin’s 1950 essay, “The Hedgehog and the Fox.” In the essay, Berlin quotes the ancient Greek poet Archilochus:"A fox knows many things, but a hedgehog one important thing.”
Taking some poetic license, you could said that a hedgehog is more prone to moving straight from hunch to belief, where a fox tends to evaluate her hunches against multiple sources. Tetlock found that when it came to the accuracy of predictions, it was better to be a fox than a hedgehog. In some cases, much better.
But Tetlock also found that when it comes down to “crunching hunches," machines tend to beat men hands down. It’s because humans have been programmed for thousands of generations to trust our hunches and no matter how much we fight it, we are born to treat our hunches as fact. Machines bear no such baggage.
This is an example of Moravec’s Paradox: the things that seem simple for humans are amazingly complex for machines. And vice versa. As artificial intelligence pioneer Marvin Minsky once recognized, it’s the things we do unconsciously that represent the biggest challenges for artificial intelligence: “In general, we’re least aware of what our minds do best.” Machines may never be as good as humans at creating a hunch – or, at least, we’re certainly not there yet. But machines have already outstripped humans in the ability to empirically analyze and validate multiple options.
Fellow Online Spin columnist Kaila Colbin posited this idea in her last column, “When Watson Comes for Your Job, Give it to Him.” As she points out, IBM’s Watson can kick any human ass when it comes to reviewing case law – or plowing through the details required for an accurate medical diagnosis – or assisting students prepare for an upcoming exam.
But Watson isn’t very good at coming up with hunches. It’s because hunches aren’t rational. They’re inspirational. And machines aren’t fluent in inspiration. Not yet, anyway.
Maybe that’s why, even in something as logical as chess, the current champion isn’t a machine, or a human. It’s a combination of both. As American economist and author (Average is Over) Tyler Cowen explained in a blog post, a “striking percentage of the best or most accurate chess games of all time have been played by man-machine pairs.” Cowen shows four ways a man-machine team can outperform -- and they all have to do with leveraging the respective strengths of each. Humans use intuition to create hunches, and then harness the power of the machine to analyze relevant options.
Hunches have served humans very well. They will continue to do so. The trick is to decouple those hunches from the belief-making mechanism that has historically accompanied it. That’s where we should let machines take over.