Massively multiplayer online role-playing games (MMORPGs) are played in complex, persistent virtual worlds. Over time, the landscape of these worlds evolves and changes as players create and personalise their own virtual property. However, many existing approaches to the design of the non-player characters that populate virtual game worlds result in characters with a fixed set of preprogrammed behaviours. Such characters lack the ability to adapt and evolve in time with their surroundings. In addition, different characters must be individually programmed with different behaviours. Motivated reinforcement learning offers an alternative approach that achieves more adaptive characters. A game world can be populated with characters that use the same agent model, but which develop different behaviours over time based on their individual experiences in the world.