Earlier this month, a family member texted me a news alert that wildfire smoke was going to billow its way from Canada down to where I lived. The air quality, she said, would be “worse than red.” But I didn’t feel the slightest bit concerned. And I went to bed without giving it a second thought. Early the next morning, however, I woke up in a fit of coughs. Smoke had funneled through an open window into my home, settling in a toxic haze. 

Why, I wondered, had I not taken the threat more seriously? Reflecting on the situation, I realized that I had based much of my response on assumptions: that the media was overblowing the situation, that companies would have closed their offices if the air quality was truly hazardous, that governing bodies would have taken sweeping measures to ensure our safety. Even if such assumptions might seem reasonable, they were nonetheless wrong. And that cost me.

This week, we explore the potential consequences of making incorrect assumptions in the context of artificial intelligence and corporate mandates.

Knowing what matters

In recent years, AI has become incredibly good at certain medical tasks, like detecting heart disorders on an EKG. It’s gotten to the point that even seasoned surgeons like Kellogg’s James Weinstein are starting to believe that, “in some respects, artificial intelligence may know more medicine than any individual physician” with its capacity to synthesize millions of scientific documents in seconds.

But when it comes to high-stakes medical decisions, AI still has blind spots. Weinstein and cowriter Dr. Ogan Gurel tell the story of a case where an AI tool flagged a patient’s heart-rhythm abnormality and suggested an invasive procedure.

The AI’s recommendation was based on the best medical outcome for the average patient, assuming that was the patient’s priority. But the patient—a retired teacher who wanted to avoid a long recovery—preferred the option that would allow him to stay healthy enough to travel to see his grandchildren.

The physicians didn’t make the same assumption. They listened to the patient instead. And they decided medication and monitoring fit the patient’s goals much better than surgery.

“The AI wasn’t wrong,” writes Weinstein, who is also a clinical professor and an adjunct professor in healthcare. “It just didn’t know what mattered.” 

Beyond the hospital setting, when AI says a particular option is best, Weinstein suggests first asking, “Best for whom?”; and then, “What does this system not know about me?”; and finally, “What happens if I wait or choose differently?” That line of questioning could help people avoid falling into the trap of hidden assumptions when using AI tools.

“AI is very good at telling you what usually works for people like you,” he writes. “It is far less capable of understanding what you are trying to protect, avoid, or prioritize.”

Read more in the Los Angeles Times.

Trickle-down effect?

Starting in the mid-2000s, several European countries passed laws that set a minimum quota for women on company boards. In 2005, for example, a Norwegian law required corporate boards to be at least 40 percent women. Representation of women on U.S. boards also grew rapidly in the 21st century, due in part to pressure from large institutional investors like Vanguard, BlackRock, and State Street. 

But initial optimism that these requirements would translate into more women leaders at all levels did not materialize, says David Matsa, a professor of finance at the Kellogg School. 

In an analysis of European public companies from 1999 to 2023, Matsa and his colleague found that gender quotas for corporate boards led to a 20 percent increase in the number of women board members. But these laws had a minimal effect on the number of women senior executives or senior managers.  

“The broader goal of those policies was presumably not just to change those particular companies at the board level but to spur changes throughout those organizations that would cascade throughout the economy,” Matsa says. “And that doesn’t seem to have happened.” 

Read more in Kellogg Insight.

See you next week,

Abraham Kim, senior research editor
Kellogg Insight