Persona based co-operative multi-agent reinforcement learning to enhance user experience

Abstract

A system can determine a first persona for a user identity. The system can generate a web page based on the first persona, wherein respective reinforcement learning agents determine respective content of respective content modules of the website, wherein the reinforcement agents operate on an environment that comprises a group of states, a group of actions, and a group of rewards that are based on the group of states, and respective actions of the group of actions, and wherein a reward of the respective reinforcement learning agents that is increased via reinforcement learning is based on a sum of respective rewards of the respective reinforcement learning agents in transitioning the user identity to a second persona with respect to the website. The system can, in response to receiving first action data indicative of user identity interaction with the web page, transition the user identity to the second persona.