The solution of the problem of bluff detection in the game «I-doubt-it» based on reinforcement learning