As much as some technologists believe AI can write the next best-selling novel or create art that inspire us, most of us would agree that we are far from being there (and even before we get there, ethical and legal issues have already begun to surface, and rightfully so).
We trained our model to understand cognitive development research and curriculum theories, and then set it out to wade through and curate the vast amount of existing multimedia content made for kids out there on the world wide web. We built the delivery system to prescribe the content and craft a unique learning experience for every child, but the content itself has to be human, original works. They include the incredibly original phonics songs by Gracie’s Corner, or Scratch Garden’s hilariously engaging videos about adding and subtracting, or Jack Hartmann rapping number bonds, to name a few.
Not all kid content is wholesome and educational. These amazing creations had to be unearthed from oceans of toy-unboxing videos and addictive 3D animations touted as “educational” that have flooded YouTube. The indie creators (many of whom are teachers) that release some of the most pedagogically appropriate and innovative content are never breaking out; 1 in 5 videos targeting young children circle back to CocoMelon simply by sheer volume. We could not rely on existing media platforms to inform our model of what is “good” content. That would only widen gaps and deepen biases. We had to pivot.
We went through months and months of research with a diverse team of media, child psychology, and social-emotional learning experts (representing 15 ethnicities), and decided that we need to create our own representative, inclusive, developmentally appropriate content that presents real-life situations where children had to problem-solve.
Instead of contracting studios, we went across the Bay Area, to College Track and the Kapor Foundation to find young people that have traveled an incredible amount of “distance” – born and raised in underserved, low-income households and Title I Schools, fighting their way through ceiling after ceiling to attend the top colleges across the nation – and asked them to write stories that would inspire and encourage 3 to 9 year old children. Their stories highlight the emotional vocabulary, their resilient inner voice, and social awareness that was critical to their own pursuit of growth and character development. The content centered around 5 pillars – “Emotion and Feelings”, “Family & Better Together”, “Self-Care”, Growth Mindset” and “Preserving Resources”.
The pivot to inviting 15-20 year old underrepresented young people (24 in total, more than 60% of overall staff) craft stories to teach Social Emotional Skills proved to be one of our best moves. Their stories were unconventional, original, and featured characters that hit very close to home. My favorites include the biracial child that brought a “different-looking” lunch to school and getting nervous about what others may say; the hard-working pony that wore a unicorn horn every day because only unicorns are worth billions (in Silicon Valley); the Moon who had to unlearn societal standards of beauty to truly appreciate it’s bumpy surface. We employed the same young people to produce short videos from each story script, and released these videos on the Yeti Confetti™ Kids YouTube channel. Within the first 3 months, we gathered 6 million impressions, 150,000 views, and created jobs in tech and media for underrepresented young people.
We are now training a specific language model that allows children to decide what each character should do and to turn every narrative into a Critical Thinking, multi-ending story playground in-app, in partnership with Stanford.
