Should you use a computer, not a person, to solve business issues?

Here is when a human expert is not always best.

In an increasingly connected world, argues investment strategist, academic and writer Michael Mauboussin, experts add less and less value to predictions. They are being outcompeted by computers on the one hand, and collective wisdom on the other. Back in 2008, Mauboussin dubbed this trend “the expert squeeze”.

Computers outdo experts on decisions with set rules and a limited set of outcomes – credit scoring, for instance, or simple medical diagnosis. And large groups of people also outdo experts wherever probabilities have to be assessed – from the poker table to the economy.

When to use experts

That doesn’t mean experts are past their use-by date. In Mauboussin’s analysis, the one area where experts still outperform others is on “rules-based” problems where the rules allow for a very wide range of outcomes. Think of chess, or the game of Go. In business, that category of problem tends to crop up in areas such as innovation, strategy development and troubleshooting.

Rather than throw an expert at every problem, “managers must carefully categorise the business problems they face,” writes Mauboussin. Computers should be used to crunch data on rules- based problems with limited outcomes, for example, identifying which customers are most profitable for a business. Probabilities such as sales forecasts, should be based on the crowd wisdom of diverse but knowledgeable individuals right across a company. Experts should focus on issues where they recombine existing building blocks in new and creative ways.

The best experts for such problems? They’re not super-specialists, but informed generalists who “tend to know a little about many aspects of their field and are not wedded to a single approach in solving complex problems,” writes Mauboussin.

US psychology professor Philip Tetlock came up with similar finding political pundits: they don’t know any more than the average intelligent person who reads the newspapers. The worst predictions came from those who saw the whole world through one or two guiding principles and used those to make spectacular predictions. “Boom and doom pundits are the most reliable over-claimers,” says Tetlock.

 

‘Should You Use A Computer, Not A Person, To Solve Business Issues? In The Black (September 2015): 10-11. Print