The paper examines the use of Jev, a decision model, in reinforcement learning (RL) environments. It investigates how Jev can fulfill the requirements of RL systems and proposes algorithms that employ Jev as a reference policy, exploration judge, and replay rater to improve RL performance.