Are crowdsourced algorithms better than ‘old-fashioned’ decision making?

“We should change the title of our book to ‘Business Advice We Shouldn’t Believe Any More’,” announced Andrew McAfee at a recent Intel event in New York City. The purpose of the gathering was to show, via a variety of spokespeople and experts, how data and AI are the future and, of course, prove Intel’s readiness for tomorrow.

McAfee, who is a principal research scientist at MIT, is focused on how technologies are changing business, the economy, and society. His latest book (written with Erik Brynjolfsson) is titled Machine, Platform, Crowd: Harnessing Our Digital Future, was released in hardback this June and looks at the big promise of crowdsourced AI.

“Businesses fail because they continue to read the business playbook,” he explained to the assembled group of B2B leaders. Most businesses still make their decisions based on the highest paid person’s opinion—the HiPPO. But these days the HiPPO “has a mortal enemy in the room: the geek” because this person doesn’t just use subjective judgement but hard numbers.

Machines appear to trump highly paid human experts

McAfee points to a paper, published in the year 2000 [“Clinical Versus Mechanical Prediction: A Meta-Analysis”] that, based on sample of 136, suggests that the majority of the time, experts are worse than useless. This is, of course, “very 2017” thinking for a number of reasons not least because of the new found ascendency of data.

In fact, the figures McAfee selects reveal that 48% of the time the HiPPOs added no value to decision making, 46% they actively made things worse and damningly concluded that they only offered better outcomes 6% of the time. “Why aren’t we making HiPPOs an endangered species?” he asked.

This is a very interesting question but it is probably worth a caveat because while data can be absolutely brilliant at making certain decisions—like in a medical setting where things really are black and white, life or death—not everything is so clearly quantifiable. Take a business setting, not all decisions are clear cut, the main priority may be to drive more revenue but this is set against more unquantifiable goals like maintaining quality and building reputation. Maybe HiPPOs still have a place in some circumstances?

Despite this it is hard to argue with the impact that data-driven decision making has on old fashioned wisdom. McAfee cites the much used example of how an artificial opponent created by Google Deep Mind beat the human Go champion [YouTube video] and, most significantly, totally rethought the human strategies established over three millennia.

It’s a controversial truism but the many will always beat the few

Throughout history the crowd has generally had a bad reputation. It has mainly been characterised as an alarming mob of idiots with a few sensible people at the top able to get things done properly. McAfee spins this idea on its head. “The crowd is surprisingly wise,” he said describing the incredible life-saving, non-expert responses Harvard and NIH receive when they run an open data challenge to solve health challenges.

Then there is trading. Quantitative hedge funds—which employ professional number crunchers to identify and run trading strategies—are becoming big business. Bloomberg described the Medallion Fund as a “moneymaking machine like no other”. But how about if that pool of number crunchers is widened to anyone prepared to have a go?

Quantopian claims to do just that. It says it provides a “level playing field” by giving “talented people everywhere” the chance to write investment algorithms. McAfee is impressed. At the time of writing the book, he said, this crowd hedge fund had run 19 competitions of which only four winners were quant traders, one was in finance, while the remaining 14 were complete outsiders with absolutely nothing to do with the industry. “It is too early to know if this experiment will work yet,” he added “but wow, is it interesting.”

The voice of the crowd builds synergies in industries

The world of data-driven decisions has revealed previously unthinkable synergies between disparate industries, concluded McAfee. “I always thought there was no substitute for knowing an industry inside out,” but this is no longer the case.

He points to three industries that appear to have very little overlap – fitness, transport and phones – and shows how, via ClassPass, Uber and mobile app stores, they have more in common than you might think. This because once you open exercise classes, cars or mobile applications up to the crowd you get a network effect—which allows for economies of scale, better control of the ecosystem and the chance for people to say what they real want, which often isn’t what the experts think they want.

“Open source is one of the first ways we saw the power of the crowd,” said McAfee. And we all know how has gone. Once so hated Microsoft’s Steve Ballmer called Linux “a cancer”. Now every enterprise is clamouring to tell us how open it is and Linux is getting included in Windows 10.