Edward Deci and Richard Ryan have spent forty years showing that intrinsic motivation depends on three psychological needs: autonomy, competence, and relatedness. AI tools for kids violate two of them by default. They replace the kid’s autonomous creative move with an AI-generated finished thing. The studio is built around preserving all three. The Keep / Review / Undo loop is the autonomy. The wizards are the competence. The cohort and the gallery are the relatedness.
The framework, briefly.
Self-determination theory (SDT) is the most influential motivation framework in modern psychology, in continuous development by Edward Deci and Richard Ryan since the 1980s.1,2 The core claim is that humans have three psychological needs whose satisfaction predicts intrinsic motivation, well-being, and learning outcomes:
- Autonomy. The sense that what you’re doing is your choice, not externally imposed.
- Competence. The sense that you’re getting better at something that matters to you.
- Relatedness. The sense that you’re connected to others who see what you’re doing.
The three are not interchangeable. SDT predicts that satisfying two and undermining the third produces motivated-but-disengaged behavior, or motivated-but-anxious behavior, or any other unstable shape. All three need to be present, in some measure, for the kid to be in the "intrinsic motivation" mode.
Decades of empirical work, in classrooms, in sports, in workplaces, in addiction recovery, have replicated the basic structure. SDT is one of the most robust findings in motivation research. It’s the framework most learning designers reach for when they want to understand why a tool succeeds or fails to keep kids engaged.
The "AI as audience" failure mode.
Most AI tools for kids work like this. The kid types a prompt; the AI generates a finished thing; the kid looks at the thing. Sometimes the kid is delighted. Sometimes they re-prompt. Either way, the kid’s role is consumer of AI output.
SDT predicts what happens to the kid’s motivation in this loop. Autonomy is partially satisfied (the kid chose to prompt) but partially undermined (the kid didn’t actually make the thing; the AI did). Competence is undermined (the kid isn’t getting better at anything; the AI got better at predicting what they wanted). Relatedness is unaffected (the AI isn’t a peer).
The net effect: kids enjoy AI tools that work this way for a while. They lose interest faster than parents expect. SDT predicts this is because none of the three needs are getting consistent reinforcement. The autonomy is hollow, the competence isn’t building, and there’s no peer relationship.
Worse, kids who use AI tools that operate in "audience" mode for a long time develop what motivation researchers call amotivation: a low-energy, low-engagement state where the kid doesn’t see the point of starting projects on their own. The AI made the thing. Why would the kid?
What the studio does differently.
Tell and Show is built so the kid does the imagining and the AI does the typing. The kid’s autonomy is preserved by structure: every wizard requires the kid to name the change in their own words; every Inkie proposal is held back until the kid sees it and decides; every decision (Keep / Review / Undo) is the kid’s, not the AI’s.
Autonomy → the kid is the decider.
The studio runs on the principle that the kid is the author of every change. The AI never applies a change without the kid’s explicit decision. The Keep button is small and the Review button is right next to it and the Undo button is one click away. The kid presses one of the three after every Inkie move. They don’t feel like AI is doing the project; they feel like AI is suggesting moves on a project they’re running.
Competence → the kid’s taste compounds.
Each decision the kid makes builds their taste. The first time they undo a change, they’ve learned something about what they don’t want. The fifth time they review, they’ve developed a sense of when Inkie’s second proposal will be better than the first. By the end of a project, they have a decision log they can read, concrete evidence that their judgment got sharper. Competence isn’t abstract; it’s in the log.
Relatedness → the gallery and the cohort.
The published gallery exists so kids can see other kids’ shipped work. The cohort program adds synchronous peers. The "you ship at a real URL friends can play" pitch is also relatedness scaffolding. The kid’s work isn’t private; it has an audience that recognizes them as a maker. Relatedness becomes structural, not optional.
The boundary we won’t cross.
This is the design constraint we keep returning to. AI for kid creative work has to preserve autonomy. The instant a tool starts generating finished creative output for a kid (here’s the story you asked for, here’s the song you described, here’s the game you imagined) it has violated the SDT model. The kid feels like an audience.
We won’t ship features that operate in audience mode, even when they would be easier or more impressive in marketing. We won’t add a "make me a game" button. We won’t add a "write me a story" button. The wizards always operate on a project the kid is already making. The kid is always the one running the loop.
The kid does the imagining. The AI does the typing. The design constraint we won’t violate
Said differently: the question we ask before shipping any feature is "does this preserve or violate the kid’s autonomy?" If it preserves it, we ship. If it violates it, we change the design or skip the feature. There are no exceptions. The motivation model says this is what makes the difference between a tool kids use for years and a tool they use for two weeks.
Why parents notice this within a session.
The behavioral marker parents see is small but consistent. A kid using a Tell and Show studio for the first hour will say things like "I want to…" or "let me try…". First-person ownership language. A kid using a "make-me-a-game" tool for the first hour says things like "let’s see what it makes". Third-person observer language. The kid’s language tracks their sense of authorship.
That language is what we want. Not because it’s the marketing line. Because it predicts, per forty years of motivation research, that the kid will still be working on the project on Sunday afternoon. Which is the marketing line, but it’s also the design constraint, and they happen to point at the same shape of product.
References
- Edward L. Deci & Richard M. Ryan, Intrinsic Motivation and Self-Determination in Human Behavior, Plenum, 1985. The foundational text.
- Richard M. Ryan & Edward L. Deci, "Self-Determination Theory and the Facilitation of Intrinsic Motivation, Social Development, and Well-Being," American Psychologist, 2000. The most-cited synthesis.
- Edward L. Deci, Richard Koestner, & Richard M. Ryan, "A meta-analytic review of experiments examining the effects of extrinsic rewards on intrinsic motivation," Psychological Bulletin, 1999. A useful counterpoint on what undermines intrinsic motivation.
- For SDT applied to digital media specifically: Andrew Przybylski, C. Scott Rigby, & Richard Ryan, "A motivational model of video game engagement," Review of General Psychology, 2010.