The complexity of personalizing education lies in understanding both how people learn and the fundamental structure of human knowledge. In achieving our goals, we knew a pioneering approach would be needed.
Sana's technology makes innovative use of deep neural networks. These models take content, answers, response times and an array of contextual information as input, before processing through a number of different network layers containing millions of neuron-like connections. One neural network, the "prediction network", predicts how the student will interact with content. Another, the "planning network", selects the content that will result in the highest increase in engagement over time.
The sheer number of online education users around the globe means that we can gather data in bigger numbers than ever. Information can also be collected with far more regularity — daily or even hourly. With every new student, Sana gets new insights about how people learn. More students mean more data and more meaningful results. Sana is already producing insights and discoveries from billions of data points gathered from tens of millions of students.