Leveraging social networks to motivate humans to train agents

摘要: Learning from rewards generated by a human trainer observing the agent in action has been demonstrated to be an effective method for humans teach perform challenging tasks. However, how make learn most efficiently these kinds of reward is still under-addressed. In this paper, we investigate effect providing social-network-based feedback intended engender competitiveness, focusing on its impact trainer's behavior. The results our user study with 85 subjects show that agent's social can induce train longer and give more feedback. Furthermore, performance was much better when social-competitive provided. also making active further increases amount time trainers spend training but does not improve performance.