The neural architecture of theory-based reinforcement learningHow does the brain build mental models of rich, dynamic domains, such as video games?