A New Experimental Perspective
W. Bradley Knox, Brian D. Glass, Bradley C. Love, W. Todd Maddox, Peter Stone · 2012
Human beings are a largely untapped source of in-the-loop knowledge and guidance for computational learning agents, including robots. To effectively design agents that leverage available human expertise, we need to understand how people naturally teach. In this paper, we describe two experiments that ask how differing conditions affect a human teacher's feedback frequency and the com- putational agent's learned performance. The first experiment considers the impact of a self-perceived teaching role in con- trast to believing one is merely critiquing a recording. The second considers whether a human trainer will give more frequent feedback if the agent acts less greedily (i.e., choos- ing actions believed to be worse) when the trainer's recent feedback frequency decreases. From the results of these ex- periments, we draw three main conclusions that inform the design of agents. More broadly, these two studies stand as early examples of a nascent technique of using agents as