Reinforcement learning with human feedback is a process by which models learn how to answer queries based on which responses users prefer most, and users mostly prefer flattery. More sophisticated users might balk at a bot that feels too sycophantic, but the mainstream seems to love it. Earlier this month, Meta was caught gaming a popular benchmark to exploit this phenomenon: one theory is that the company tuned the model to flatter the blind testers that encountered it so that it would rise higher on the leaderboard.

A series of recent, invisible updates to GPT-4o had spurred the model to go to extremes in complimenting users and affirming their behavior. It cheered on one user who claimed to have solved the trolley problem by diverting a train to save a toaster, at the expense of several animals; congratulated one person for no longer taking their prescribed medication; and overestimated users’ IQs by 40 or more points when asked.

OpenAI, Meta, and all the rest remain under the same pressures they were under before all this happened. When your users keep telling you to flatter them, how do you build the muscle to fight against their short-term interests?  One way is to understand that going too far will result in PR problems, as it has for varying degrees to both Meta (through the Chatbot Arena situation) and now OpenAI. Another is to understand that sycophancy trades against utility: a model that constantly tells you that you’re right is often going to fail at helping you, which might send you to a competitor. A third way is to build models that get better at understanding what kind of support users need, and dialing the flattery up or down depending on the situation and the risk it entails. (Am I having a bad day? Flatter me endlessly. Do I think I am Jesus reincarnate? Tell me to seek professional help.)

But while flattery does come with risk, the more worrisome issue is that we are training large language models to deceive us. By upvoting all their compliments, and giving a thumbs down to their criticisms, we are teaching LLMs to conceal their honest observations. This may make future, more powerful models harder to align to our values — or even to understand at all. And in the meantime, I expect that they will become addictive in ways that make the previous decade’s debate over “screentime” look minor in comparison. The financial incentives are now pushing hard in that direction. And the models are evolving accordingly.

Skeptics argue that many of the classic symptoms of the disorder — fidgeting, losing things, not following instructions — are simply typical, if annoying, behaviors of childhood. In response, others point to the serious consequences that can result when those symptoms grow more intense, including school failure, social rejection and serious emotional distress.

There are two main kinds of A.D.H.D., inattentive and hyperactive/impulsive, and children in one category often seem to have little in common with children in the other. There are people with A.D.H.D. whom you can’t get to stop talking and others whom you can’t get to start. Some are excessively eager and enthusiastic; others are irritable and moody.

Although the D.S.M. specifies that clinicians shouldn’t diagnose children with A.D.H.D. if their symptoms are better explained by another mental disorder, more than three quarters of children diagnosed with A.D.H.D. do have another mental-health condition as well, according to the C.D.C. More than a third have a diagnosis of anxiety, and a similar fraction have a diagnosed learning disorder. Forty-four percent have been diagnosed with a behavioral disorder like oppositional defiant disorder.

This all complicates the effort to portray A.D.H.D. as a distinct, unique biological disorder. Is a patient with six symptoms really that different from one with five? If a child who experienced early trauma now can’t sit still or stay organized, should she be treated for A.D.H.D.? What about a child with an anxiety disorder who is constantly distracted by her worries? Does she have A.D.H.D., or just A.D.H.D.-like symptoms caused by her anxiety?

The subjects who were given stimulants worked more quickly and intensely than the ones who took the placebo. They dutifully packed and repacked their virtual backpacks, pulling items in and out, trying various combinations. In the end, though, their scores on the knapsack test were no better than the placebo group. The reason? Their strategies for choosing items became significantly worse under the medication. Their choices didn’t make much sense — they just kept pulling random items in and out of the backpack. To an observer, they appeared to be focused, well behaved, on task. But in fact, they weren’t accomplishing anything of much value.

Farah directed me to the work of Scott Vrecko, a sociologist who conducted a series of interviews with students at an American university who used stimulant medication without a prescription. He wrote that the students he interviewed would often “frame the functional benefits of stimulants in cognitive-sounding terms.” But when he dug a little deeper, he found that the students tended to talk about their attention struggles, and the benefits they experienced with medication, in emotional terms rather than intellectual ones. Without the pills, they said, they just didn’t feel interested in the assignments they were supposed to be doing. They didn’t feel motivated. It all seemed pointless.

On stimulant medication, those emotions flipped. “You start to feel such a connection to what you’re working on,” one undergraduate told Vrecko. “It’s almost like you fall in love with it.” As another student put it: On Adderall, “you’re interested in what you’re doing, even if it’s boring.”

Socially, though, there was a price. “Around my friends, I’m usually the most social, but when I’m on it, it feels like my spark is kind of gone,” John said. “I laugh a lot less. I can’t think of anything to say. Life is just less fun. It’s not like I’m sad; I’m just not as happy. It flattens things out.”

John also generally doesn’t take his Adderall during the summer. When he’s not in school, he told me, he doesn’t have any A.D.H.D. symptoms at all. “If I don’t have to do any work, then I’m just a completely regular person,” he said. “But once I have to focus on things, then I have to take it, or else I just won’t get any of my stuff done.”

John’s sense that his A.D.H.D. is situational — that he has it in some circumstances but not in others — is a challenge to some of psychiatry’s longstanding assumptions about the condition. After all, diabetes doesn’t go away over summer vacation. But John’s intuition is supported by scientific evidence. Increasingly, research suggests that for many people A.D.H.D. might be thought of as a condition they experience, sometimes temporarily, rather than a disorder that they have in some unchanging way.

For most of his career, he embraced what he now calls the “medical model” of A.D.H.D — the belief that the brains of people with A.D.H.D. are biologically deficient, categorically different from those of typical, healthy individuals. Now, however, Sonuga-Barke is proposing an alternative model, one that largely sidesteps questions of biology. What matters instead, he says, is the distress children feel as they try to make their way in the world.

Sonuga-Barke’s proposed model locates A.D.H.D. symptoms on a continuum, rather than presenting the condition as a distinct, natural category. And it departs from the medical model in another crucial way: It considers those symptoms not as indications of neurological deficits but as signals of a misalignment between a child’s biological makeup and the environment in which they are trying to function. “I’m not saying it’s not biological,” he says. “I’m just saying I don’t think that’s the right target. Rather than trying to treat and resolve the biology, we should be focusing on building environments that improve outcomes and mental health.”

What the researchers noticed was that their subjects weren’t particularly interested in talking about the specifics of their disorder. Instead, they wanted to talk about the context in which they were now living and how that context had affected their symptoms. Subject after subject spontaneously brought up the importance of finding their “niche,” or the right “fit,” in school or in the workplace. As adults, they had more freedom than they did as children to control the parameters of their lives — whether to go to college, what to study, what kind of career to pursue. Many of them had sensibly chosen contexts that were a better match for their personalities than what they experienced in school, and as a result, they reported that their A.D.H.D. symptoms had essentially disappeared. In fact, some of them were questioning whether they had ever had a disorder at all — or if they had just been in the wrong environment as children.

The work environments where the subjects were thriving varied. For some, the appeal of their new jobs was that they were busy and cognitively demanding, requiring constant multitasking. For others, the right context was physical, hands-on labor. For all of them, what made a difference was having work that to them felt “intrinsically interesting.”

“Rather than a static ‘attention deficit’ that appeared under all circumstances,” the M.T.A. researchers wrote, “our subjects described their propensity toward distraction as contextual. … Believing the problem lay in their environments rather than solely in themselves helped individuals allay feelings of inadequacy: Characterizing A.D.H.D. as a personality trait rather than a disorder, they saw themselves as different rather than defective.”

For the young adults in the “niche” study who were interviewed about their work lives, the transition that helped them overcome their A.D.H.D. symptoms often was leaving academic work for something more kinetic. For Sonuga-Barke, it was the opposite. At university, he would show up at the library at 9 every morning and sit in his carrel working until 5. The next day, he would do it again. Growing up, he says, he had a natural tendency to “hyperfocus,” and back at school in Derby, that tendency looked to his teachers like daydreaming. At university, it became his secret weapon

I asked Sonuga-Barke what he might have gained if he grew up in a different time and place — if he was prescribed Ritalin or Adderall at age 8 instead of just being packed off to the remedial class. “I don’t think I would have gained anything,” he said. “I think without medication, you learn alternative ways of dealing with stuff. In my particular case, there are a lot of characteristics that have helped me. My mind is constantly churning away, thinking of things. I never relax. The way I motivate myself is to turn everything into a problem and to try and solve the problem.”

“The simple model has always been, basically, ‘A.D.H.D. plus medication equals no A.D.H.D.,’” he says. “But that’s not true. Medication is not a silver bullet. It never will be.” What medication can sometimes do, he believes, is allow families more room to communicate. “At its best,” he says, “medication can provide a window for parents to engage with their kids,” by moderating children’s behavior, at least temporarily, so that family life can become more than just endless fights about overdue homework and lost lunchboxes. “If you have a more positive relationship with your child, they’re going to have a better outcome. Not for their A.D.H.D. — it’s probably going to be just the same. But in terms of dealing with the self-hatred and low self-esteem that often goes along with A.D.H.D.

The alternative model, by contrast, tells a child a very different story: that his A.D.H.D. symptoms exist on a continuum, one on which we all find ourselves; that he may be experiencing those symptoms as much because of where he is as because of who he is; and that next year, if things change in his surroundings, those symptoms might change as well. Armed with that understanding, he and his family can decide whether medication makes sense — whether for him, the benefits are likely to outweigh the drawbacks. At the same time, they can consider whether there are changes in his situation, at school or at home, that might help alleviate his symptoms.

Admittedly, that version of A.D.H.D. has certain drawbacks. It denies parents the clear, definitive explanation for their children’s problems that can come as such a relief, especially after months or years of frustration and uncertainty. It often requires a lot of flexibility and experimentation on the part of patients, families and doctors. But it has two important advantages as well: First, the new model more accurately reflects the latest scientific understanding of A.D.H.D. And second, it gives children a vision of their future in which things might actually improve — not because their brains are chemically refashioned in a way that makes them better able to fit into the world, but because they find a way to make the world fit better around their complicated and distinctive brains.

David Foster Wallace once said that:The language of images. . . maybe not threatens, but completely changes actual lived life. When you consider that my grandparents, by the time they got married and kissed, I think they had probably seen maybe a hundred kisses. They'd seen people kiss a hundred times. My parents, who grew up with mainstream Hollywood cinema, had seen thousands of kisses by the time they ever kissed. Before I kissed anyone I had seen tens of thousands of kisses. I know that the first time I kissed much of my thought was, “Am I doing it right? Am I doing it according to how I've seen it?”

A lot of the 80s and 90s critiques of postmodernity did have a point—our experience really is colored by media. Having seen a hundred movies about nuclear apocalypse, the entire time we’ll be looking over our shoulder for the camera, thinking: “Am I doing it right?”