What we miss when we’re missing relationship data

Connection x usage = impact

Lately, my eyes keep wandering to danah boyd‘s seminal book, It’s Complicated, sitting on my bookshelf. I first read it over a decade ago, but its lessons are reverberating today.

boyd’s careful, empathetic analysis of how young people were relating to and through social media stood the test of time because she wasn’t just writing about tech, she was writing about us.

Core to her argument is how we use technology reveals us to ourselves. It mirrors complex social patterns far beyond our screens. That’s a critical insight, especially when it comes to young people who, as boyd described, tend to use technology in far more sophisticated and nuanced ways than meets the eye (aka than adults typically give them credit for).

The Rithm Project‘s latest report on youth & AI feels like a glimpse into that same phenomenon in the present-day wild west of Gen AI. The phenomenal research, led by Alison Lee with Michelle Culver, analyzes data from a survey of ​​2,383 young people ages 13-24 on how they’re using—or not using—AI.

Besides asking young people how they’re using AI, the Rithm team did something else important: they asked about the social circles those users occupied, including how much they felt they belonged and how much they felt like a burden to others because of their problems.

That revealed important (and dramatic) distinctions hidden behind usage patterns. Users engaging in the exact same online behaviors actually clustered into distinct profiles—what the researchers called personas—with very different social risk profiles.

For example, the researchers found that a total of 18% of young people reported relying on AI for emotional support. But those users fit into two different personas: “social processors” and “private processors.” Social processors leaned on AI in emotionally tricky situations, but the majority reported always having someone to turn to IRL. Private processors, on the other hand, were far less connected offline–only 7% felt they always had someone to turn to. Private processors were also more than twice as likely to report feeling the urge to use AI more and more, suggesting a clear risk of emotional overreliance.

Source: The Rithm Project, 2026

In other words, the consequences of turning to AI aren’t just defined by the use case, but by the relationships that are or aren’t within reach for its users.

Another pair of personas showed similar distinctions within the same use case. The researchers found that a total 40% of young people surveyed reported using AI just for information and tasks—what adults might think of as productivity, or non-relational use cases. But Rithm’s data reveals that about half of those young people reported feeling far more connected (“The Intentional Connector”) than their peers (“The Brave Face”). Those less-connected Brave Faces were also more than twice as likely to report feeling like a burden to others with their problems.

That suggests very different risk profiles: even if they aren’t befriending their chatbots and talking to them about their emotions, Brave Faces still face a real relational risk if they are inclined to depend on AI, rather than humans, in moments of need.

Source: The Rithm Project, 2026

This study is pushing us to think carefully about how we talk about the risks AI poses to social connection, because, when it comes to the relational impacts of AI, people’s specific social circumstances may be just as important as their use of the technology itself.

That doesn’t mean we should stop interrogating AI companies’ predatory design decisions that scale emotional manipulation or fake empathy. But it does mean that if you’re worried about youth digital health and safety today, attending to kids’ offline social lives may be more potent than focusing on digital tools alone.

The unfortunate reality? Most of the time, we understand very little about young people’s social lives and needs.

Measuring the missing metrics

In my own work on social capital, I’ve spearheaded efforts to build better relationship data–especially in schools–for the last decade. It hasn’t gone great.

The efforts are often met with deep skepticism, overwhelm, or occasionally apathy. While there are some bright spots, like the Search Institute’s Developmental Relationships Survey, by and large, schools fly blind on the social health of their students. In the last year, I’d bet most have spent far more time implementing cell phone bans than finding out who their students know and can turn to, and how comfortable they feel reaching out to others for help.

Civil society isn’t much better: while community-based organizations and mentoring programs invest deeply in youth relationships, they’re rarely funded to build or measure relationships to outlast interventions.

That stark lack of relationship data means that, while we know AI poses risks, we actually know very little about which young people are most vulnerable to overrelying on AI to the detriment of social connection—or what interventions would actually safeguard them as chatbots offer an alluring alternative to real-world connections.

Building beyond individual agency

I think there’s another reason relationship data matters: it can keep us honest about the structural interventions this moment deserves, not just the individual ones.

It’s common today to hear calls to build young people’s capacity and agency to use AI on their terms; given the risks AI poses, we should equip young people to shape their relationship with this technology rather than have algorithms drive them.

But I also think that Rithm’s data illuminates why that agency-driven approach is necessary but insufficient. Young people should be equipped to hold their own…but they should also be held by the communities around them, and by each other. They should be capable of self-determination but also immersed in contexts that offer scaffolds and support for the parts of them that are still developing.

If we don’t understand the social context they operate in, that’s nearly impossible to do with any real attunement. The Brave Face might ace a digital literacy curriculum and still quietly deplete his weak-tie network and deepen help-seeking avoidance. The Private Processor could develop savvy strategies for working through challenges with AI, but fail to see that leaning on those around her can be profoundly meaningful, and that when we burden one another, we build reciprocity.

Absent relationship data, we risk defaulting to defining individuals by their individual traits and behaviors, and trying to arm them with everything they might need to navigate the world on their own. While that’s a rational strategy, it falls into the very trap that we need to avoid.

In the final chapter of It’s Complicated, boyd wrote, “Technology makes the struggles youth face visible, but it neither creates nor prevents harmful things from happening, even if it can be a tool for both. It simply mirrors and magnifies many aspects of everyday life, good and bad.”

Rithm’s study is a first-of-its-kind look into that mirror right now. What we’re seeing isn’t just a nuanced story of AI’s journey into young people’s lives, but that social connectedness will play a big part in shaping how the technology shapes them. We can keep debating the technology, and trying to keep up with the impact these new, anthropomorphic tools will have on rising generations. But the more pressing question may be the one we’ve been avoiding asking for far longer: who do young people actually have to turn to?