A Risk of Cognitive Convenience

What ChatGPT and GPS have in common

In the last post, I wrote about how using ChatGPT before doing your own thinking might be rotting your brain—or at least preventing the kind of deep cognitive engagement that leads to actual learning. It’s like having someone else lift weights for you; the weights get moved, but you lose the benefit of the workout.

There’s an exploding body of research on how AI tools impact our thinking and ability to learn. My take on this research so far is that—as with the study featured in the last post—there is both good and bad news, and a lot depends on the precise details of how the tool is being used.

This is an emerging and extremely important issue. I’ve decided I’m going to track this research and make it a regular Range Widely topic so that we can keep up together. I’m also going to write about other research that I think is conceptually related—which brings me to today’s study.

The Taxi Driver vs. The Bus Driver

The study, published in The BMJ (formerly The British Medical Journal), examined death certificates from nearly 9 million people to analyze Alzheimer’s deaths across 443 different occupations.

The scientists had a clever hypothesis. We know from a famous piece of research that London taxi drivers—who have to memorize “The Knowledge” of 25,000 streets and 20,000 landmarks—have a larger hippocampus (a part of the brain responsible for memory and spatial navigation). We also know that the hippocampus is one of the first regions to atrophy in Alzheimer’s disease. So these researchers asked: Does a career spent engaging in intense, real-time wayfinding protect you from Alzheimer’s?

As it turned out, of all 443 different occupations, taxi drivers and ambulance drivers had the lowest risk of dying from Alzheimer’s after adjusting for demographic factors.

You can explore other occupations in the chart here.

The researchers specifically included other transportation jobs like bus drivers and aircraft pilots. These professionals direct vehicles, but mainly along predetermined routes that don’t require constant, real-time spatial problem solving. Those occupations did not show the same benefit.

“Undesirable Ease”

Cognitive psychologists use the term “desirable difficulty” to describe tactics that make learning or practice more difficult, and often more frustrating in the short-term, but more effective in the long run. Navigation may be an example of an occupational desirable difficulty—challenging drivers in a way that improves their learning, and, it seems, may have an important health benefit.

To harken back to the students in the MIT study I wrote about in the last post—who were better off using their brains before ChatGPT, not the reverse—I wonder if taxi drivers and ambulance drivers who start to rely on GPS before their own navigational skills will lose the Alzheimer’s benefit. As with the essay-writers in the MIT study, it could present the opposite of a desirable difficulty—an undesirable ease.

I posed that idea to Anupam Jena, a professor and physician at Harvard, and an economist, and one of the co-authors of this study. Jena replied: “The now ubiquitous use of GPS could, in theory, mitigate the spatial memory based benefits that prior generations of taxi or ambulance drivers may have developed.” (He was quick to add, though, that this doesn’t necessarily mean that the rest of us—who aren’t relying on spatial navigation skills in our daily work—should turn off our GPS for brain health.)

I’m not suggesting we all become ambulance drivers to ward off Alzheimer’s. But I do think there’s a broader principle here about the sometimes surprising benefits of cognitive effort. Just as those students in the ChatGPT study did better when they wrestled with essay prompts before turning to AI, it’s good for our brains to have regular challenges that aren’t entirely offloaded to technology.

The MIT study’s most hopeful finding was that the “brain-first” students could use AI productively and without a learning decrement after they’d done some of their own thinking. They used ChatGPT to refine and enhance work they’d already wrestled with, not as a replacement for thinking. Measures of their brain activity showed strong connectivity between regions, but, more importantly, they remembered what they wrote. Maybe that’s a good model: Think first, tool second. I certainly don’t plan to forsake GPS, but for several years I have tried to navigate before I rely on it—not because I’ve been thinking about brain health, but because I want to learn my way around.

The question for the rest of us: What are we outsourcing that we should be doing ourselves? And what price might we pay for taking the easy route?

To be clear, I think there are myriad benefits of AI for individual users. And I’ll take on research that supports that in future posts. Hopefully, if you stick with me over time, we can home in on some answers regarding not whether to use AI tools, but how to use them to support rather than stifle our own thinking and learning.

Thank you for the time you took to read this post.

If you’re interested in the benefits of wayfinding, you might enjoy Alex Hutchinson’s book The Explorer’s Gene—which was featured in the “Why You Should Get Lost More Often” post.

And if a friend sent you this post, please consider supporting Range Widely by subscribing.

Finally, if you appreciated something here, please share it.

Until next time…

David