Interview with HBS Alumni Stories on Uncertainty Book
05 Mar 2025
Uncertain Terms
After Amar Bhidé (MBA 1979/DBA 1988) became an HBS assistant professor in 1988, then-dean John H. McArthur (MBA 1959/DBA 1963) gave him a copy of economist Frank Knight’s 1921 book Risk, Uncertainty and Profit. Knight’s idea that “uncertainty” must be distinguished from “risk”—in that we can calculate possible outcomes for the latter, but not the former—never rose to prominence, but Bhidé was hooked. To his frustration, today’s economics centers on modeling and measuring numerical risks overlooks uncertainty, which defies neat statistical appraisal. With his new book, Uncertainty and Enterprise: Venturing Beyond the Known, he hopes to spur more research on the distinct role of uncertainty, defined simply as doubt about what is or could be. We talked to Bhidé—Professor of Health Policy at Columbia University’s Mailman School of Public Health and Professor of Business Emeritus at Tufts University’s Fletcher School of Law and Diplomacy— about why, even as businesses rush to adopt artificial intelligence tools, they should spend more time grappling with judgments about one-off uncertainties. —Janine White
Many people equate uncertainty with stress. Where do you see the upside?
You would not want to go to a movie where somebody had told you the ending. The excitement of going to a sports event comes from not knowing who will win. The same is true in business. Entrepreneurs rarely start businesses just for the money. We venture beyond the unknown because it energizes us. It is what makes us human.
Now there’s an inescapable, passive sort of uncertainty: If you are running a gas station, you have to deal with not knowing at what rate EVs will replace internal combustion engines. But there’s also proactive uncertainty that you stir up by trying something new. That’s what many crave. It’s an exercise in imagination.
Businesses today are investing heavily in big data’s predictive potential, but you write that you’re skeptical that “artificial intelligence will tame even banal uncertainties that frustrate actual human intelligence.”
A backward-looking statistical model is sometimes good enough even if its predictions are highly inaccurate. Nearly all the ads that Google and Facebook serve me haven’t the slightest relationship to what I’m interested in, yet compared to the alternative of blind advertising, it’s good enough. But to believe that we should turn all contextual, situation-specific problems into statistical problems is misguided and dangerous. The fact that AI models have a billion variables doesn’t change the fact that they are still extrapolating from history and in dynamic economies and societies the future always deviates from the past.
You suggest that instead of solely relying on data, entrepreneurs and other businesspeople should engage in a “creative process that combines facts and imagination.”
People who should know better intone “let the evidence speak.” That’s nonsense, especially with context-specific uncertainties. You have to imaginatively interpret what you observe. I have a little aphorism [in my book] that “evidence collaborates with but cannot replace imagination.” If you’re dealing with phenomena which are changing over time, you must try to anticipate how they’re changing. Use your imagination Do not slavishly follow a statistical model, which is based entirely on what has happened in the past.
Sensible people know this instinctively. But they worry: Are we missing something? Is there some scientific formula, which if only we knew, we’d be able to—as they say in business school—crack the case?
Are there hidden opportunities for businesses amid the data-driven artificial intelligence boom we’re in now?
The AI stampede creates opportunities in other directions where there is much less competition. You could be the cool-headed person who says, let’s not waste our time and money on fads that don’t make any sense for us. There is so much we can and need to do without big data, without machine learning.
So much cutting-edge science can be commercialized without Large Language Models. So many low-tech processes can be improved without AI.
Businesses everywhere are struggling with getting people back into the office. Statistics can’t tell them whether and how to do that. That requires imaginative, case-by-case judgment that HBS’s classic case method teaches.
You write that business routines are often “mocked,” but that they play a critical role in navigating uncertainty.
Making big decisions is exciting. Should you make this multibillion-dollar acquisition? Fire the CFO? But setting up a systematic process for evaluating and implementing new initiatives? That’s boring!
We celebrate visionary, go-for-broke leadership. Yet big, prescient bets can’t sustain longterm success. Dynamic organizations require routines to harness collective imaginations, forestall problems, and creatively cope with inevitable surprises.
Perhaps your book will change that trend?
My argument about uncertainty is squeezed on two sides. On the one side is the fantasy of replacing forward-looking judgment with backward-looking AI models. If the model is large enough and if sufficient computing power is thrown at it, who needs human imagination and discourse? And on the other side, we are glorifying the supernaturally visionary “great leader.” Hopefully, this book will stiffen the resistance of level-headed skeptics — before these two pathologies inflict great harm.