There's No Crying in Baseball
Whether AI Can Feel Isn't the Problem. Legitimacy Is.
There’s a lot of talk right now about which decisions we should hand over to AI. Should it approve the loan? Grade the essay?
The conversation usually gets stuck on the big dramatic questions—is this thing really intelligent? What will AI-driven job displacement do to the economy? To our social institutions? I find those genuinely interesting.
My limited take is that a lot of these AI debates get more navigable the moment we stop indexing them against big unknowns and start indexing them against trust and legitimacy.
By legitimacy I just mean whether we accept the authority of something as trustworthy and valid. Not necessarily whether we like it.
It’s an old and ordinary idea—we lean on it constantly without ever naming it—but I think it’s the thing doing most of the real work in the AI debates.
The Issue of Legitimacy
Take medicine. There are already well over a thousand AI tools cleared for medical use, and most doctors use some of them. In a lot of specific tasks the technology is good—sometimes better than the humans.
And yet the regulators’ whole posture, so far, has been that these systems support a decision a person still makes. They inform; they don’t decide. So why would that be, if the software is often more accurate? It’s not really about accuracy.
In one survey, nearly half of radiologists said they thought patients simply wouldn’t accept a report that came from an AI alone.
That’s not a question about whether the machine can read the scan. It’s a question about whether we’ll accept its verdict—which is to say, a question about legitimacy.
There’s even a bill floating around Congress about whether an AI could count as a “practitioner” allowed to prescribe medication.
The point is, handing a decision to a machine sometimes makes us trust the outcome more, and sometimes makes us trust it less—with the very same technology.
So the deciding factor was never the chip. It was whether the output had earned the right to be believed and acted on. That’s an old question. AI just makes questions about truth and trust less philosophical and more immediate.
Example: Major League Baseball
Major League Baseball is running this experiment this season with an automated system—the same tracking tech behind all its stats—to help settle balls and strikes during games.
