A few years ago, an Oxford University study concluded that computers will likely replace people in 45% of American jobs, ranging from truck drivers to construction workers to office support staff.
Daniel Kahneman speculates that CEOs might one day join that endangered species list.
Kahneman, a Nobel Prize winner for his work in behavioral economics and author of the managerial must-read "Thinking Fast and Slow," spoke on Thursday at Wharton's People Analytics Conference, in Philadelphia. Kahneman was interviewed onstage by Daniel Pink, author of several bestsellers on work and motivation.
Perhaps our foremost expert on the wacky world of human judgment, Kahneman was a natural keynoter for an event about better decision-making through numbers. In a face-off between folks and formulas, he said, put your money on the machine. "In general when you pit professional judgments - even expert judgment - against formulas, algorithms, simple rules that combine information, in general the rules beat the experts," said Kahneman. "There is little or no evidence of cases in which expert judgment does better than intelligently constructed formulas."
Algorithms win because, unlike humans, they are not susceptible to noise. Noise is the kind of random, unpredictable stuff that causes people grappling with questions to arrive at very different solutions. When asked how much variation they expect between two people determining the same interest rate, for example, or evaluating the same insurance claim, leaders typically estimate between 5% and 10%. In fact the real variance is between 40% and 60%. "By the way, when you tell the team leaders that there is 50% variability, they want to take an algorithm," said Kahneman. "That really concentrates the mind."
The algorithmic edge applies even to very sophisticated human thinking. The conference took place less than a month after a computer program created by Google whumped a world champion at Go, a game uniquely suited to human intuition and ingenuity. Google's researchers programmed into the system 30 million moves by masters. The software generated new moves by playing repeatedly against itself. "We develop intuition with the data we collect in a lifetime," said Kahneman, who called the match among the most important events of the year. "AI will be able to do better. How we live with that - I don't know."
As Go goes, so, eventually, will go business judgment. Kahneman recently attended an AI conference where Google chief Eric Schmidt described the multi-talented digital assistant he imagined having in the future. He called it Not-Eric. Kahneman was impressed by Schmidt's vision, but wondered what would happen if Google or someone else developed a Not-Eric with business judgment. "One of the effects of creating algorithms that do complex things like business judgments or medical diagnoses is that the godlike feature of human judgment is eliminated," said Kahneman. "Because anybody with that algorithm can out-guess Eric Schmidt. How can this work within an organization? How does it affect the power structure?"
Kahneman argued one reason that decision-support systems never fulfilled their promise--except in certain industries like oil, where they are used to choose drilling sites--is that they are "very threatening to the leaders. Leaders, if there is one thing they don't like, it's to be out-guessed."
Pink wondered whether we are on a trajectory where eventually Not-Eric fires Eric.
Kahneman demurred: "My expectation is that Eric will fire Non-Eric."