In the age of AI, better questions are the new investment edge

More than a decade ago, a group of amateur forecasters competing in the Good  Judgment Project—a U.S. intelligence‑sponsored forecasting tournament launched in 2011—achieved a remarkable feat: they proved better at predicting geopolitical events than professional intelligence analysts with access to classified information.

Their story was captured in Superforecasting, the bestselling book by Philip Tetlock and Dan Gardner, which is enjoying renewed interest amid the rise of prediction markets. Now, Gardner believes AI is on course to rival the superforecasters. Speaking during a recent visit to our Baltimore office, Gardner said that, if current trends continue, AI could match the best human superforecasters as soon as 2027 or 2028.

So, if machines are becoming as good as the very best humans at predicting what happens next, where does human advantage come from? Gardner’s answer is deceptively simple: better questions.

After all, a forecast is ultimately an answer to a question. As AI causes the supply of high‑quality answers to explode, Gardner argues, their relative value falls—and the value of good questions rises.

For investors, this matters. If everyone can interrogate the same filings and datasets with increasingly powerful tools, processing information efficiently becomes less of a differentiator. The edge may lie further upstream: knowing what is worth investigating in the first place.

What makes a superforecaster?

According to Gardner, the original superforecasters were not distinguished by genius‑level IQ, specialist knowledge or sophisticated mathematics. They were smart—typically in the top quintile for intelligence—but what really set them apart was how they thought. They were analytical and organized, hungry for new knowledge, comfortable expressing uncertainty in numerical probabilities and adept at breaking complex problems into smaller, more manageable ones. Above all, they lacked hubris and accepted that they would sometimes be wrong.

That last quality may be particularly important for investors. Once we form a view, Gardner argues, we naturally want it to be right, and contradictory evidence can feel like a challenge not just to our forecast but to our expertise. The danger is that we become anchored to our original view and explain away information that should make us rethink it.

Gardner points to John Maynard Keynes as an example of a different mindset. Famously confident in his own intellect, Keynes was also remarkably willing to acknowledge mistakes. As an investor, he suffered major setbacks, examined what had gone wrong and adapted his approach. For Gardner, that is intellectual humility in action: not an absence of conviction, but a recognition that even strongly held views must be tested against new evidence.

The superforecasters behaved similarly. They updated their forecasts far more often than other participants as new information arrived—but typically in small increments. React too little and you cling to a view the evidence no longer supports; react too much and every new datapoint sends you in a different direction. Good judgment lies somewhere between the two.

Curiosity needs discipline

A willingness to change your mind is only part of the superforecaster mindset. Gardner also highlights an appetite for learning. Asked to forecast a presidential election in Ghana, most people might think, “Who cares?” A superforecaster was more likely to think: “Great—here’s an opportunity to learn about Ghana!” Gardner points to Warren Buffett and Charlie Munger as investment exemplars of the same trait: voracious readers whose curiosity extended far beyond markets.

But curiosity works best when paired with a disciplined process. Gardner cited physicist Enrico Fermi, who tackled seemingly unanswerable problems by breaking them into smaller questions and exposing the assumptions buried within them. Investors can do the same: rather than simply asking what AI will mean for markets, they can break that big question into smaller ones. What would need to happen for a particular outcome to materialize? What evidence would support—or challenge—that view? The aim is not certainty, but to expose blind spots and gradually build a better picture.

Probabilistic thinking adds another layer of discipline. Words such as “likely” or “unlikely” can conceal large differences of opinion; putting a number on a forecast forces precision and makes it easier to revisit that judgment as the evidence changes. The point is not to pretend uncertainty can be eliminated, but to be explicit about how much confidence a view deserves.

Where the edge moves next

For decades, investment advantage has partly depended on better information, greater expertise and better analysis. None of those things will cease to matter. But AI may democratize some of them, making high‑quality answers easier to obtain. If so, differentiation may move upstream—from answering questions better to asking better ones in the first place.

The investors who stand out may be those willing to ask questions that test their own views—and humble enough to change their minds when the answers surprise them. In a world awash with answers, that combination of curiosity and humility may provide an increasingly valuable investment edge.