Designing for Development First

When Wayne and I started building Zondle way back in 2012, we didn't begin with "how do we make children spend more time on the platform?" We began with "what does the research say about how children actually learn?"

That question has driven our work for 14 years. Wayne's background in AI ethics and education research, his connections with researchers at Oxford, UCL, The Open University, and Bristol, these relationships have shaped how we think about building platforms for children. We've been refining pedagogical approaches and working with universities since 2012. What we have now with oodlü is a really strong starting point, built on over a decade of learning what works and what doesn't.

The research that shaped our early thinking came from Professor Paul Howard-Jones at the University of Bristol. His work on uncertain rewards and dopamine response influenced how we approach engagement and learning. His research showed something counterintuitive: uncertain rewards generate a stronger dopaminergic response than certain ones.

When a child answers a question correctly and knows they'll definitely get a reward, there's a dopamine spike when they see the question. When they answer correctly and might get a reward, the dopamine builds whilst they wait to find out. The uncertainty creates anticipation. The anticipation creates a teachable moment.

We built oodlü Quickfire around this principle. When children answer questions, a wheel of chance determines the reward. The rising anticipation as the wheel spins creates the optimal moment for teachers to provide feedback and scaffold learning. The research suggests this approach enhances both engagement and retention compared to predictable reward structures.

This is one example of designing for development first. Starting with evidence about how learning works, then building engagement mechanisms around that evidence rather than hoping educational content can be retrofitted onto engagement mechanics designed for retention.

Our learning framework has four layers. We've mostly implemented the first layer, which includes retrieval practice, interleaving, elaborated feedback, elaborative interrogation, and stop-and-think pauses. Each technique is backed by research showing measurable improvements in long-term retention and understanding.

Retrieval practice means testing helps learning more than repeated studying alone. Interleaving means mixing different types of problems improves retention compared to blocked practice. Elaborated feedback means prompting learners to reconsider their reasoning before providing explanation. Elaborative interrogation means asking why answers are true to encourage deeper processing.

Stop-and-think is particularly interesting. Requiring students to pause before answering promotes deeper cognitive processing, reduces guessing, and leads to stronger understanding. We designed specific mechanics to create those pauses naturally within gameplay rather than feeling like imposed delays. I built the software alongside Professor Kaśka Porayska-Pomsta (Director of the UCL Knowledge Lab) and was actively involved in the research.

So these aren't our theories. They're findings from researchers at Oxford, UCL, The Open University, Bristol. We're implementing established evidence about how learning works rather than inventing our own pedagogical approach and hoping it functions.

The difference between this and most platform development is the starting point. Most children's platforms start with engagement mechanics proven to maximise time on platform, then look for ways to add educational value. We're starting with evidence about how learning works, then designing engagement mechanics that support rather than undermine those processes.

Does this make building harder? Yes. Completely. It's easier to optimise purely for engagement when you're using well-tested retention mechanics from gaming and social media. Adding the constraint of "this also needs to support genuine developmental outcomes based on research" limits design choices significantly.

But it creates something more valuable. These research-backed mechanisms create a fundamental platform we can build on and expand into the consumer market. The same principles that support educational outcomes, retrieval practice, uncertain rewards, and collaborative learning also support general nurturing and development. This framework works whether we're talking about curriculum learning or the broader developmental benefits of open-world experiences.

Layers two through four of our framework aren't implemented yet. They involve more sophisticated applications of collaborative learning, metacognition, and adaptive challenge. Each layer adds developmental techniques with research backing their effectiveness.

The work Wayne Holmes has done on AI in education informs how we think about these layers. His research emphasises that AI should guide and mediate rather than replace human interaction in learning. The architecture we're building keeps adults in the loop whilst using evidence-based techniques to enhance developmental outcomes.

This approach creates something sustainable. Build something children genuinely want to engage with, using research-backed techniques that support actual learning and development. The mechanisms work for educational contexts and for the broader consumer platform we're building.

The research exists. Evidence about retrieval practice, uncertain rewards, collaborative learning, and metacognitive development. Most platforms building for children ignore it because engagement-first design is faster and more proven from a business perspective.

We're choosing differently. Development first. Engagement built around evidence about how learning works. The order matters.

We'd love to hear your thoughts on this. Find us on the social channels linked at the top of the page.