This study presents an approach to decision making for autonomous driving based on reinforcement learning and shared autonomy. Game theory is used to model the interaction between the human driver and the machine, while model predictive control (MPC) is used in the actual control task. The decision model used by the machine is based on Reinforcement Learning (DRL) based on Deep Q Network (DQN) derived from naturalistic driving. To verify the performance of RL based decision model, its performance is compared with performances of other decision models, namely the conventional human driver behavior model, the rule based decision model and the data-driven decision model. The mean velocities and moving distances from highway driving simulations are compared and analyzed to compare their performance. It is shown that the RL based decision model demonstrates the best performance among the decision models considered.