论文信息 - Leveraging social networks to motivate humans to train agents

Leveraging social networks to motivate humans to train agents

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

Shimon Whiteson | Hayley Hung | W. Bradley Knox | Guangliang Li

[1] Brett Browning,et al. Learning by demonstration with critique from a human teacher , 2007, 2007 2nd ACM/IEEE International Conference on Human-Robot Interaction (HRI).

[2] Andrea Lockerd Thomaz,et al. Reinforcement Learning with Human Teachers: Evidence of Feedback and Guidance with Implications for Learning Performance , 2006, AAAI.

[3] Shimon Whiteson,et al. Using informative behavior to increase engagement in the tamer framework , 2013, AAMAS.

[4] Farbod Fahimi,et al. Online human training of a myoelectric prosthesis controller via actor-critic reinforcement learning , 2011, 2011 IEEE International Conference on Rehabilitation Robotics.

[5] Jude W. Shavlik,et al. Giving Advice about Preferred Actions to Reinforcement Learners Via Knowledge-Based Kernel Regression , 2005, AAAI.

[6] Peter Stone,et al. Interactively shaping agents via human reinforcement: the TAMER framework , 2009, K-CAP '09.