How to Stay Curious While Avoiding Distractions

My wife doesn’t even roll her eyes anymore when I don’t finish the simplest of household tasks. “Are you distracted?” she asks, even though she knows the answer. Luckily, I have my back now, because if there’s one person in this household who’s more likely to be halfway through putting on their shoes or brushing their teeth because they suddenly remember something they wanted to read, watch, or listen to, it’s my 13-year-old son. If they make Distracted an Olympic sport, I’m putting money on him being a medal contender.

Of course, my wife is less strict with him than I am.

“He gets distracted because he’s so curious,” she said. And the comment stuck in my mind, partly because I had read almost exactly the same thing from design guru Don Norman, who wrote, “My curiosity often leads me to insights that have helped me in my career. So why does this beautiful, creative quality of curiosity get the pejorative term ‘distraction’?” These are ideas to ponder. Yet surely there is a distinction to be made between the essential quality of curiosity and its evil twin, distractibility.

Janelle Shane’s exploration of AI, You Look Like a Thing and I Love You (2019), sheds light on the question under controlled conditions by looking at the behavior of curious and distractible AI systems. As Shane explains, AI systems are often trained through a form of trial and error, with a “reward function” that decides which experiments should be considered a success and which a failure. For example, you could teach a computer to ride a virtual bicycle in a simulated 3D environment by rewarding the distance traveled and penalizing the number of times the bike falls over.

The challenge comes when the reward function misses what the human programmers really wanted. Perhaps the AI ​​avoids the risk of falling by leaving the bike on the ground, or maximizes distance traveled by rocking in a large circle, or even turning the bike upside down and pedaling. These aren’t just theoretical possibilities. One algorithm was designed to sort a list of numbers and simply removed the list, instantly ensuring that no number was out of order.

These are fairly simple problems. The more complex the desired behavior, the easier it is to accidentally reward the wrong thing. But there is a clever and effective approach to training computers to solve a fairly wide range of problems: reward curiosity. More precisely, reward the computer when it encounters situations where it finds the outcome unpredictable. It will then look for something it has not seen before.

Shane writes: “A curiosity-driven AI will learn to move through a video game level so that it can see new things, avoiding fireballs, monsters, and death pits, because when they hit it, it sees the same boring death scene.” Death should be avoided, not for its own sake, but because it is terribly predictable.

This is all fascinating in itself, and illustrates why humans have evolved a sense of curiosity. But AI systems, like 13-year-old boys, can also be curious to the point of distractibility. Ask a curiosity-driven AI, for example, to teach itself to play a Pac-Man-like game in which ghosts move randomly through a maze, and you’ll struggle: the AI ​​doesn’t have to do anything to satisfy its curiosity, because unpredictable ghosts are endlessly fascinating. Or, as Shane explains, a curiousbot quickly learns how to navigate a maze unless there’s a TV on one of the maze walls showing a series of random images. “Once the AI ​​found the TV, it was mesmerized.” Just like my son. Or, for that matter, me.

This problem is well-known enough to AI researchers that it has a name: the “noisy TV problem.” And for a clever programmer, it can be solved. Unfortunately, our modern world is full of distractions that are designed to grab our attention as perfectly as a noisy TV is designed to grab the attention of a curiousbot, and we can’t simply reprogram ourselves to avoid these intellectual empty calories.

One solution is defensive: avoid loud TVs. Delete your social media account (or at least remove the app from your phone and install two-step verification to make it hard to log in). Don’t sleep with your phone in the bedroom. Turn off all but the essential notifications. We all know this, and if you can bring yourself to do it, it works. But a second approach focuses more on the positive. In addition to trying to avoid sheer novelty, we should seek out things that make us curious. This is easier than you might think, because thoughtful curiosity builds knowledge, and knowledge builds thoughtful curiosity.

As Ian Leslie explains in his 2014 book Curious: The Desire To Know and Why Your Future Depends on It, human curiosity typically requires a reasonable basis of facts to back it up. “The curiosity zone is next to what you already know,” he writes.

I think that’s true. I’m much more curious about new ideas in areas I already know a little about, like economics, board games, or gymnastics, than about subjects I don’t have an intellectual foothold in, like anthropology, knitting, or hockey.

So the plan for both distractible members of the Harford household should be the same: keep learning. The more you know, the more you’ll gravitate toward something deep, rather than the next thumbnail YouTube recommends.

Written for and first published in the Financial Times on August 23, 2024.

Loyal readers may appreciate the book that started it all: The Undercover Economist.

I have set up a storefront on Bookshop in the US and UK. Links to Bookshop and Amazon may generate referral fees.