In tech companies, the mandate across marketing and R&D has long been clear: move fast, capture attention, and optimize for engagement. But when the end-user is a teenager or a child, the traditional metrics of success—like daily active users and time spent—collide with the fundamentals of healthy development.

My observation from working on teen and child focused technologies—including VR, social media, AI, and ed-tech platforms—is that conversations tend to orient around making these technologies “safe for kids”, but this to me seems like a low bar. We should be holding ourselves to a higher standard; aspiring to build technology that is actually good for young people.

And I’m not alone. I had conversations with leading experts across education, media, social science, and technology which revealed powerful insights about how we build good products for developing minds.

Special thanks to Michael Preston at Sesame Workshop, Mark Sparvell at Microsoft, Dr. Cecilie Steenbuch at Copenhagen Business School, and Gary Cooper, formerly at Amazon.

1. Design for social bonds—with boundaries

The public conversation surrounding youth and technology is loud, anxious, and often flattens a complex reality into a single narrative: screen time is bad. Generalized tech anxiety can miss reality of what kids are actually seeking.

As Michael Preston, Executive Director of the Joan Ganz Cooney Center at Sesame Workshop, points out, this narrative washes away vital nuance. “Kids are not small adults,” Preston says. “there’s just a gap, in terms of understanding kids’ developmental needs.”

Social connections are of profound importance to developing minds. Mark Sparvell, an education expert at Microsoft, challenges the narrative that young people are mindlessly addicted to their devices. “Kids aren’t addicted to tech. It’s their friends that they’re addicted to.” And he points to studies showing how some technologies can significantly facilitate peer connection, emotional support, and positive new relationships.

And yet, kids don’t have the biological ability to self-regulate the ways adults do. Emphasizing the need for built in safeguards that inhibit mis-use and over-reliance, Cecilie Steenbuch, a social scientist studying digital influence at Copenhagen Business School, notes that youth possess frontal lobes that are not fully developed. This limits their ability to anticipate long-term consequences and heightens their neural sensitivity to reward signals, particularly when they feel “watched” by their peers online.

2. Build critical thinking into any AI experience

If social media was built on the architecture of peer validation, the era of generative AI introduces an entirely new paradigm. Chatbots and AI-friends are not just interactive or personalized; they are intimate. Unlike static screens, AI talks back. It mimics human empathy and forms synthetic relationships with young users.

In order for AI to be good for kids, it must help them build strong muscles around critical thinking. Gary Cooper, who formerly worked on responsible tech and AI at Amazon, warns of the “sycophantic aspect” of AI bots that eagerly agree with whatever a user says. When a child asks a question, the AI responds with an unwavering confidence that the child inherently accepts as a statement of fact. This creates a dangerous “confidence trap“, Sparvell notes that when a user’s baseline knowledge on a topic is low, they are far more likely to accept an AI’s output without question.

Poorly trained AI can also create an unrealistic and unhelpful model for real-world human relationships. Educational media of the past, like Sesame Street’s Oscar the Grouch, was explicitly designed to teach kids how to navigate difficult, grumpy peers and understand the environmental factors behind someone’s mood. But AI that never challenges a child fails to teach them the friction and emotional intelligence necessary to sustain healthy, resilient human relationships.

3. Change KPIs from "time spent" to "effective use"

Tech platforms are inherently incentivized to keep users engaged. But optimizing for continuous scrolling puts companies in direct conflict with children’ s wellbeing. To fix this, tech companies must fundamentally rethink their metrics of success.

“The traditional KPIs of time spent and day-over-day users… I don’t know that that gives us the right information about if this an effective thing for kids.” Gary Cooper argues that R&D and product teams must define and measure “effective use”. Did the interaction with an AI agent challenge the child’s critical thinking? Did it encourage them to step away from the screen and research a topic at their local library?

Too often, companies build mass-market products optimized for adults, and simply slap restrictive guardrails on them to create a “kids’ version”. As Cooper puts it, tech teams must learn to “go slow to go fast,” building child wellbeing into the foundation rather than trying to patch harmful impacts downstream.

4. Defy outdated Silicon Valley “truisms”

How do we design digital products that actually align with a child’s developmental needs? This requires de-programming from many UX guidelines that have gone without challenge or questioning, until now:

Design for the “off-screen” moment

Preston recalls the original philosophy behind Sesame Street, envisioned by Joan Ganz Cooney in the 1960s: “What’s most interesting about the show is what happens when the TV’s turned off”. Modern tech should not aim to trap a child inside a digital ecosystem, but rather inspire them to go play, collaborate, and explore in the physical world.

Introduce “positive friction”

Today’s media is optimized for rapid, continuous call-and-response. But deep learning requires a slower pace. Preston advocates for the awkward, patient pauses of old-school shows like Blue’s Clues—giving kids the time and space to summon their own curiosity. Steenbuch similarly advocates for “positive friction” in user interfaces to deliberately slow down consumption and push users into a more deliberative, reflective state of mind.

Utilize “pleasant frustration”

Sparvell notes that tech shouldn’t make things effortlessly easy. Learning requires a “Goldilocks zone” of pressure and support. He calls this “pleasant frustration”—an environment where the challenge is demanding enough to build mastery, but supportive enough to keep the child from giving up.

Guided release

Instead of handing a child an open-ended AI tool, products should be designed for “guided release.” Tech should also provide scaffolding, teach the skills, and gradually release control to the user as they build competence and agency.

5. Get on the rug

Finally, product teams cannot rely on guesswork, adult assumptions, or parental beta-testers to understand what kids need. To build better tech, designers must bring children into the process as equal partners. Preston advocates for getting “on the rug” with diverse groups of kids.

By co-designing with kids—sitting at eye-level, utilizing tangible materials, doing hands-on activities and treating their ideas with respect—designers can bypass the limitations of a child’s verbal articulation and tap into their unvarnished, authentic needs and ideas.