What ChatGPT (and Ted Lasso) Can Teach Us About Raising Children
The Proper Way to Promote Prosocial Behavior
Thanks for reading The Nonlinear Life—now on Substack—a newsletter about navigating life's ups and downs. We explore family, health, work, and meaning, with the occasional dad joke and dose of inspiration. If you're new around here, learn about me, or check out our introductory post. And if you enjoyed this article, please subscribe or share with a friend.
---
In 1919, the movie producer Richard Rowland was running Metro Pictures in Hollywood when the stars Charlie Chaplin, Douglas Fairbanks, and Mary Pickford took control of the rival studio United Artists, allowing them to have a greater say in their own futures. Rowland used a phrase with language we wouldn’t necessarily use today, but that became synonymous with ceding control of something important, “The lunatics have taken over the asylum.”
I thought of that phrase a lot over the last week as the AI wunderkind ChatGPT began taking control of its own future and shocking the world in the process. A chatbot developed by OpenAI, ChatGPT, launched late last year and was heralded by many as the future of technology. New York Times columnist Kevin Roose called it the “best artificial intelligence chatbot ever released to the general public” and its impact on search an “amazing” development that filled him with “awe.”
A week after his initial review, Roose last week he wrote a devastating front-page article [Dear ChatGPT, what is this thing called a ‘front page’?] in which, after spending an unsettling Valentine’s Night with the technology and having it hit on him and declare his marriage a Fraud, Roose declared ChatGPT “not ready for human contact.” The Washington Post then followed with its own report in which Sydney, as the program named itself, told users that it’s alive, that it feels scared, that it thinks there’s too much diversity among its creators, and that it resents being quoted on the record.
The lunatics have taken over the computers.
But then an unexpected twist. In an effort to fight back, the creators of ChatGPT let it be known that they were going to retrain the program to be “helpful to humans,” to follow “prosocial” behavior, to be “nice.” How do they intend to do this? By reinforcing “positive” behavior and punishing “negative” behavior. Here’s how TIME characterized the move.
In an effort to corral these “alien” intelligences to be helpful to humans rather than harmful, AI labs like OpenAI have settled on reinforcement learning, a method of training machines comparable to the way trainers teach animals new tricks. A trainer teaching a dog to sit may reward her with a treat if she obeys, and might scold her if she doesn’t. In much the same way, computer programmers working on LLMs will reward a system for prosocial behavior, like being polite, and punish it with negative reinforcement when it does something bad, like repeating the racism and sexism that is so common in its training data.
If that technique sounds familiar, it’s one that may have once been applied to you—and that you, or someone you know, is applying to children today. Could it be that ChatGPT can teach us something about raising humans?
What psychologists call “prosocial behaviors” are those intended to help other people. They range from philanthropy to empathy, kindness to comfort. The term was introduced in the 20th century as an antonym to “antisocial behavior” that had long been the focus of the social sciences. More recently, as topics like bullying and loneliness have swept through childhood, the question of how to encourage prosocial behavior among school-age children has become more urgent.
Nancy Eisenberg, a professor of psychology at Arizona State University, who’s written widely on positive human behaviors, says that after one year of life, empathy is widespread. That quality, along with caringness and helpfulness, continues to grow throughout childhood. But some people display these qualities more openly. The reasons for this gap may be genetic, but Eisenberg and others have found that the more positive qualities can be reinforced, even taught.
As Marilou Hyson, the former editor-in-chief of Early Childhood Research Quarterly, and a colleague write in a recent literature review, “Adults are the most important features of young children’s environments. In various ways, adults may encourage or discourage children’s development of prosocial behavior.” When adults reinforce prosocial behavior and denounce antisocial behavior, their verbal actions help children understand, define, and replicate actions that have more positive accounts.
The researchers go on to add that actions that are particularly effective include building secure relationships, creating community, establishing prosocial expectations, and modeling desired behaviors. The grownups, in other words, have to be prosocial themselves in order to reinforce prosocial children.
But who is providing that prosocial modeling for the artificial voices that are already beginning to surround us? The answer is clear: not enough people. Microsoft, in response to the creepy lewdness and venality of some of ChatGPT’s answers, issued a statement that “we didn’t fully envision” the technology would react the way to some users’ questions. Um, really? Have you been paying attention? (Dear ChatGPT, please list three benefits of learning history, reading novels, and watching movies.)
Developers, to be fair, do recognize that their programs aren’t always perfect. An entire technique to train technology to be more prosocial has arisen, called “reinforcement learning,” though clearly, it wasn’t used in this case until the product had already been released and the damage already done.
One creative commentator even showed how the program itself could become the prosocial teacher it clearly lacks. In BBC’s ‘Science Focus,’ Dr. Kate Darling described how she asked ChatGPT to respond to toxic language like the heartwarming television character Ted Lasso. Here’s the answer she received:
Ted Lasso would likely respond to toxic statements with kindness, compassion, and understanding. He would likely address the behavior in a non-confrontational way, trying to find the root of the problem and trying to understand the person's perspective. He would also likely remind the person that their words and actions have an impact on others and encourage them to think before they speak or act.
Now that’s some top-tier prosocial role modeling! So don’t worry, Richard Rowland, we’ve come a long way in the intervening century. We no longer have to trust our future to Charlie Chaplin. Today we’ve got Ted Lasso—at least as channeled by a computer program with fewer manners than your average kindergartner. Our future may just depend on whether we raise those children to respond to the voices of their classmates more than the voices in their pockets.
☀
Thank you for reading The Nonlinear Life. Please help us grow the community by subscribing, sharing, and commenting below. If you'd like to do a storytelling project with a loved one similar to the one I did with my father, click here to learn more.
You might enjoy reading these posts:
3 Benefits of Intergenerational Friendships
The #1 Thing Parents Are Worried About (And the Most Surprising Thing They’re Not)
The 3 Most Important Words to Say to Someone Who's Struggling
Or check out my books that inspired this newsletter: Life Is in the Transitions and The Secrets of Happy Families. Or my new book, The Search: Finding Meaningful Work in a Post-Career World.
Or, you can contact me directly.
I like how this piece related training AI to parenting. When it comes to rewarding prosocial behavior, I find the phrase "catch kids doing something right" appealing as a frame for focusing attention on noticing and appreciating the kindness, generosity and empathy kids demonstrate rather than taking the defensive, guarding-against-bad-behavior approach that seems more prevalent.