The Evolution of Reframing Recommendation

Reinforcement learning is really helping us reframing recommendation to take into account the long term user satisfaction. So something that we understand is like, let's not try to predict the next click, butlet's try to predict a habit. And so what does what habit mean is like you try to look at correlation between the next click and habit because habit is a way to understand where user will go next in the journey. I can see this kind of idea being rolled out to kind of wider businesses, other use cases.