Is There an Algorithm for Happiness?

Published: May 18, 2026

By Jeremy Martin

13 min read

As a boy growing up in Britain, David (Danny) Blanchflower, PhD, and his mates would toss sticks at chestnut trees until hard brown chestnuts, nicknamed “conkers,” rained down. They’d drill holes through them, thread them onto string, and whack at each other in “conker battles.”

“It was utterly irrelevant,” recalls Blanchflower, who now studies happiness as the Bruce V. Rauner Professor of Economics at Dartmouth. “The conkers were just an excuse. It was about being together.”

When he describes those childhood memories now, decades later, there’s warmth in his voice, the wistful signature of dignified meaning. Today, however, few British parents let their children go conkering. The game—along with similar random ways kids of the pre-digital age invented to have fun—has all but died out. And simultaneously something intangible has slipped away: Young people are decidedly less happy than they once were.

Since the 1970s, social scientists could rely on a predictable pattern across life stages: Youth were most mirthful, the middle-aged more miserable, and old timers happier yet again. So universal was the pattern—holding across 146 countries, appearing in over 600 published papers—that Blanchflower says, “The U-shaped curve [of happiness] was one of the most important patterns in social science. Until it wasn’t.”

This graph, from research conducted by Dartmouth economist David (Danny) Blanchflower, PhD, depicts the “U-shaped curve of happiness,” consistently observed by social scientists from the 1970s to the 2010s, in which youth reported the highest levels of satisfaction with life. *Source: Blanchflower, DG. Journal of Population Economics, 2021*

Sometime between 2012 and 2015, young people became among the most morose of the bunch, according to Blanchflower’s research. The trend was global, observed across 167 countries, accelerating in its spread. But what was changing?

An Outbreak of Unhappiness

Hunting for culprits, Blanchflower and his fellow sleuths noticed the timing of this outbreak of unhappiness lined up precisely with when social media truly went viral—drawing young people away from the real world and each other. Adolescents who once spent more than two hours a day together outside of school were, by the mid-2010s, spending closer to 40 minutes together.

By 2025, teenagers were logging about nine hours daily on screens, nearly half of it on social media, according to the Pew Research Center. At the same time, rates of anxiety, depression, and psychological distress surged. Today, roughly 60% of people ages 18 to 22 meet criteria for at least one mental health condition.

“We’re seeing less in-person connectivity, more social isolation, especially since the pandemic,” says Michael Heinz, MD, an inpatient psychiatrist at Dartmouth Health’s Dartmouth Hitchcock Medical Center and assistant professor of psychiatry at the Geisel School of Medicine at Dartmouth. “I’ve seen more struggles that are directly related to smartphone use, to the point where it starts to take the place of social relationships and other meaningful activities.”

But for young people living among those statistics, social media is just a slice of the story. According to Jaime Graft D ’26, a computer science and neuroscience student at Dartmouth, young people today face a maelstrom of pressures from the economic to the existential.

“Everything is extremely expensive. Wages have not kept pace with inflation. The job market is brutal. Most young people doubt they’ll ever be able to buy a house. Then there’s COVID, which kind of ruined college and high school for a lot of people,” Graft says. “There’s been a lot of destigmatization of mental health struggles, which is great. But I think that’s mainly because a lot of people experience these struggles now.”

Young people have also lost access to “third spaces,” Graft says. Slowly and almost imperceptibly, these informal spots where communities once gathered beyond work and home have all but dried up. Bars and restaurants are often cost prohibitive, too, he notes. As a result, “our third spaces are often online.”

Social media aggravates these deeper stresses. Constant visibility raises the cost of failure. Everything is a performance. Selfhood is constantly up for evaluation. “You’re always being monitored. You’re always comparing yourself to other people,” Graft says. “But nobody’s really flourishing.”

The question is whether anything can be done about it.

An Unlikely Counselor

During her sophomore year, Asli Tavasli D ’27, a computer science major at Dartmouth, got a concussion that triggered a cycle she couldn’t escape. Recovery meant falling behind in coursework, isolating herself to catch up, getting sick from lack of sleep, then feeling overwhelmed by stress.

Even with good friends and a supportive partner, she needed help—right then and there.

“With mental health services, you might have to wait awhile for an appointment,” she says.

So late one night, overwhelmed and unable to sleep, Tavasli found herself talking to ChatGPT about how she felt.

A young woman looks down at her phone. A pixel effect blurs half her face. A wall of plants is behind her. — Asli Tavasli D ’27, a computer science major at Dartmouth, has helped build Evergreen, a first-of-its-kind generative AI platform to provide students with personalized wellness support, along with faculty and more than 200 other student workers. Tavasli and her peers may be facing a mental health crisis, but they are also designing solutions together. Credit: Robert Gill

“I knew from working in tech that we shouldn’t 100% trust AI,” she says. “But I wanted to talk to someone. I wanted to hear from someone. And it actually helped me.”

That gave her pause. If a chatbot that was not designed for mental health support could provide some relief in a moment of crisis, what might be possible with one purpose-built for well-being?

The Blueprint for Well-Being

Such technology would need to be built upon a proven model for flourishing.

In 1998, psychologist Martin Seligman, PhD, then at the University of Pennsylvania, founded positive psychology, a field devoted to understanding not just mental illness, but what allows people to thrive. Wellness, Seligman’s research suggests, is more than the absence of disease. It is the presence of five core parameters: Positive emotion, Engagement, Relationships, Meaning, and Accomplishment—collectively known as “PERMA.” When these elements are strong, studies link them to everything from greater job satisfaction to stronger social bonds.

But “a lot of the things that bring joy and fulfillment take work,” Heinz notes. “They take investment, time, and authentic engagement.”

During Blanchflower’s childhood, time, attention, and social connection made that work easier, and lasting happiness more attainable. For Graft and Tavasli, those same resources face strong headwinds—and are challenged by digital systems designed for fast relief.

“The ease of things that bring you pleasure or solve our problems has eroded our sense of fulfillment,” Tavasli says. Instead of working for happiness by taking a long hike, making plans with friends, or struggling to get the right answer, “we turn to easy shortcuts. Every indulgence is at our fingertips.” But it is the struggle, Tavasli points out, that produces lasting contentment. Skip it, and you lose the real reward.

Seligman calls this state of well-being—the kind built through effort, engagement, and meaningful challenge—“flourishing.” And for young people like Tavasli and Graft, the conditions that support it have become harder to sustain.

Fighting Fire With Flourishing

If immediate gratification is the problem, then a mental health intervention would need to provide the opposite. It would need to strengthen resilience and foster the slower, more deliberate art of flourishing.

That challenge sits at the heart of Evergreen, a first-of-its-kind generative AI platform being built at Dartmouth by students and faculty to provide students with personalized wellness support.

The app aims to guide users through building the best conditions and daily routines for flourishing. It leverages the principles of positive psychology—PERMA—to guide behavior away from addiction and isolation and toward connection and fulfillment.

When a student first opens Evergreen, they select wellness areas to focus on, such as social connection or personal growth. They can ask the AI chatbot for help with specific challenges like social anxiety, time management, or relationships. The chatbot responds in the tone of a trusted peer rather than a clinical therapist.

The app adapts its suggestions to individual needs based on data students choose to share. That might include sleep patterns, academic schedules, physical activity, or location on campus. The algorithm learns what conditions support each student’s wellbeing and delivers brief “nudges” at moments when they’re most likely to help.

“Evergreen might sense you’re feeling lonely,” says Andrew Campbell, PhD, Albert Bradley 1915 Third Century Professor of computer science at Dartmouth and co-lead on the project. “It might nudge you to grab coffee with a particular friend—because it’s learned that works for you.”

Using an app to get students off their apps may seem paradoxical, especially when the average college student already spends nearly seven hours a day on their phone, according to a 2023 study published in the journal Clinical Psychological Science.

“For most young people, rejecting smartphones wholesale isn’t realistic,” says Lisa Marsch, PhD, founding director of the Center for Technology and Behavioral Health at Dartmouth (CTBH), the Andrew G. Wallace Professor within the Departments of Psychiatry and Biomedical Data Science at Geisel, and co-lead on Evergreen. “They’ve become essential tools for participating in society,” she says. But, “technology is neutral. The question is how people are taught to use these tools.”

Tavasli, who in addition to her studies has served as a software developer on the Evergreen team, adds that “we have given ourselves access to everything apart from mental health…. Now we need to focus on what that uncontrolled and unmediated access has turned us into.”

When the Drug Is Digital

From 2009 to 2019, depression and suicidality among U.S. high school students spiked by 40%, even as use of most substances shrank by 20–40%, according to research by Jacob Borodovsky, PhD, a senior research scientist and epidemiologist at CTBH who studies addiction.

This inverse relationship hinted that something other than traditional drug use might have colonized the same neural pathways involved in addiction. To Borodovsky, that made sense, and the culprit was evident: algorithmically driven platforms like social media. The parallels between substance use and these platforms run deeper than most people realize. While “not all screen time is created equal,” Borodovsky says, social media platforms operate on neurological principles similar to addictive substances. They both supply intermittent rewards, bursts of validation at unpredictable intervals, that trigger quick dopamine hits.

Over time, this creates behavioral patterns that mirror addiction: compulsive use, withdrawal symptoms, and diminishing returns from everyday pleasures. Young people are particularly susceptible, as their prefrontal cortex—the very region of the brain needed to resist compulsive behavior—does not fully mature until the mid-twenties.

But research by Kyle S. Smith, PhD, associate professor of psychological brain sciences at Dartmouth, suggests dopamine is not really a “pleasure chemical.” More accurately, dopamine helps the brain learn which cues are worth paying attention to and which behaviors are worth repeating.

“Dopamine tells your brain to engage and ‘do that again,’” Smith says, which is why, when it comes to addiction, it is “a big part of the problem.”

This learning system did not evolve to keep humans scrolling. It evolved to help humans pursue connection, purpose, and meaning—all core ingredients of flourishing.

Humans rarely act on long-term intentions alone, so when well-timed cues intervene at a decisional moment, they can more effectively influence behavior, prompting us to move our body, reach out to a friend, or pause before scrolling on social media.

According to behavioral science and cognitive psychology, these “nudges” can reduce cognitive load and activate habit-forming pathways, making meaningful behaviors easier to repeat until they become second nature.

The digital realm has provided powerful vehicles for delivering these behavioral cues. While devices often foster less authentic connection than in-person interaction, says Michael Heinz, MD, an assistant professor of psychiatry affiliated with the Geisel School of Medicine at Dartmouth and Dartmouth’s Center for Technology and Behavioral Health (CTBH), smartphones’ ubiquity may make them the easiest way to “meet people where they’re at.”

But when nudges become encoded in algorithms, ethical questions can arise: Who decides what behaviors are “good” enough to be nudge-worthy? How strong should nudges be? A core principle of nudging is that it preserves people’s freedom to choose. Yet if a nudge is designed to influence behavior, can it truly preserve that freedom?

This Algorithm Does Not Want Your Attention

The premise behind Evergreen is to redirect that learning system back to its original purpose. And to do that, Evergreen turns the addictive algorithms on their head: While social media algorithms are built to keep people online, Evergreen optimizes to drive them off their phones.

An algorithm, after all, is just a tree of logic that intends to solve a problem or direct decisions.

The app’s suggestions draw from behavioral science: Evergreen’s conversational AI was trained on more than 250 dialogues to incorporate habit formation, motivational interviewing, and behavioral activation. Together, these approaches honor how people actually change. Gentle prompts steer attention toward movement, social connection, and purpose rather than passive engagement—offering small, doable actions that can build momentum when motivation is low. If a user has been sedentary too long, the app might suggest physical activity, which generates positive emotion.

Unlike social media platforms, Evergreen does not sell data, run advertisements, or share information with third parties. Sensing capabilities are designed not for surveillance, but for pattern recognition. Students decide what data, if any, to share. If they opt in to sensing, the system can adapt suggestions based on behavioral patterns—sleep, physical activity, dining hall swipes, location changes. Whereas social media uses behavioral data to keep you engaged with content, Evergreen uses it to disengage you from the screen.

The Offline Community in Your Pocket

What distinguishes Evergreen from most wellness apps isn’t just its algorithmic design, but where it comes from. The platform has been developed by Dartmouth students and faculty, with support from CTBH at Geisel. More than 200 paid Dartmouth students work or have worked flexibly, term by term, on the project.

Intentionally built locally, Evergreen knows the names of Dartmouth buildings, the locations of study spaces, and routes between dorms and dining halls. When it suggests meeting someone for coffee, it can suggest the café. By grounding its “nudges” in local context, Evergreen aims to narrow the gap between intention and action.

“Since the people making Evergreen are students,” Tavasli says, “you basically get the collective Dartmouth ‘hive-mind’ experience on your phone. An upperclassman telling you, ‘Hey, when the weather made me depressed, I borrowed a sun lamp from Baker- Berry [Library].’ It’s like another Dartmouth student talking to you.”

Graft also previously worked on Evergreen’s mobile development, helping to develop its underlying architecture. One feature he helped build allows students to signal on the app when they’re open to meeting others for a meal.

Screenshot of the Evergreen app — Evergreen’s AI assistant responds to student stress in real time, nudging users toward focus, action, and eventually, disengagement from the screen. *Image courtesy of the Evergreen team*

“It’s not very normal to just walk up to a stranger in the dining hall and sit down with them,” Graft says. “People don’t want to face rejection. Setting that status lets you know who else wants to make new connections.”

It’s a small intervention, but a revealing one—a digital tool designed explicitly to lower the friction to analog connection. A modern counterpart to Blanchflower’s conker games that still forces students to do the work of flourishing.

The Evergreen team is careful to articulate that the app is not a therapist, doesn’t diagnose or treat illness, and is not meant to replace any existing Dartmouth resources. But it does include built-in safeguards, directing users to Dartmouth Student Health Service at Dick’s House or the National Suicide Prevention Lifeline if it detects language suggesting crisis or imminent harm.

Importantly, says Iris Cao D ’26, a senior studying economics and philosophy who has worked on Evergreen’s dialogue design, the goal is “to build technology that understands people not just as data points, but as stories in progress.”

Will it Work?

Not all students are convinced. When Dartmouth announced Evergreen in 2025, The Dartmouth published several opinion pieces questioning the approach. Some students worried the app might become a substitute for human connection. Others raised concerns about data privacy, even with opt-in sensing, and whether the investment might be better directed toward expanding counseling staff or existing wellness programs.

Evergreen is currently undergoing rigorous testing under institutional review board oversight. The rollout has been intentionally gradual, beginning with a pilot group of roughly 200 students, then expanding to 600, with plans to eventually make the platform available to the full undergraduate population.

“We’re going to have over 100,000 human hours that are part of developing the system. It’s not a light-touch approach,” says Nicholas Jacobson, PhD, technical lead for the Evergreen project, associate professor of biomedical data science and psychiatry at Geisel, and director of the Treatment Development and Evaluation Core at CTBH. “It’s going to be very deeply ingrained in ways that are really trying to make these safe and effective.”

The safety protocols for Evergreen build upon Jacobson’s prior work. In 2024, along with Heinz and others, Jacobson developed Therabot, the first fully generative AI therapy chatbot to undergo a clinical trial, the successful results of which were published in March 2025. That trial created a roadmap for designing and testing these tools in safe, controlled clinical trials, and provided a foundation for the Evergreen project. While Evergreen is not designed to provide therapy, its chatbot is modeled after Therabot’s software.

Evergreen is part of a broader push to improve student mental wellness at Dartmouth. In 2023, President Sian Leah Beilock, PhD a five- year “Commitment to Care” strategic plan that expanded clinical staffing, increased access to free telehealth counseling, and reframed mental health as a campus-wide responsibility rather than an individual failing—and it’s already working. From 2021 to 2024, Dartmouth saw a decline in students at risk for moderate to severe depression from 33% to 24%, and a decrease in students at risk for moderate to severe anxiety from 27% in 2021 to 23% in 2024. Furthermore, 93% of undergraduates surveyed in 2024 agreed that they knew where to get professional help for their mental health, up 10 percentage points since 2021.

Evergreen extends this work upstream. Rather than waiting for students to reach crisis, it aims to support the everyday habits that make crises less likely.

For now, the Evergreen app is a hypothesis in motion, a carefully designed experiment rather than a proven solution. Ultimately, research will reveal whether the platform fosters flourishing and changes behavior. If it proves effective, Marsch says, the long-term goal is to adapt Evergreen for use at other institutions and eventually beyond higher education.

Marsch is under no illusions that Evergreen can fix all that ails young people’s ability to flourish today. “A lot of young people are growing up in a world that feels unstable,” she says. “Climate change, geopolitical conflict, economic precarity—social media amplifies those anxieties, but it doesn’t create them out of nothing.”

Evergreen “is not a panacea,” Marsch adds. “Technology alone can’t solve structural problems like economic inequality or climate anxiety. But it can help give people skills to navigate the world they’re in.”

Graft, too, resists the idea that an app can repair a society in which housing feels out of reach, work feels insecure, and the future feels uncertain. The value in technology, for him, resides in helping the hearts and minds of young people to believe that, even in the face of challenges, flourishing is still possible.

Blanchflower’s childhood games were never really about chestnuts. The conker battles were simply a pathway to connection and joy. A return to that past may be impossible. But the modern blueprint for well-being can still be found—and built—together.

Additional reporting by Eva Botkin-Kowacki

Download This Issue