Gravar-mail: Dynamic Interaction between Reinforcement Learning and Attention in Multidimensional Environments