Join Me: Empirical Means for Gathering Collaborative Affordances

With robots becoming more competent at navigating the world outside of tightly controlled spaces, it is becoming more relevant to consider how they will share space with and collaborate with human users. The notion of collaborative affordance has been recently brought to robotics literature as a way to describe action possibilities emerging from agents cooperating, and knowledge representation patterns have been described. However, this now raises new questions as to where collaborative affordance knowledge comes from. One natural answer would be humans, since they are the most competent at handling the world and they are the ones who may share their space with robots. In this paper, we look at design considerations for human user studies aimed at acquiring and validating collaborative affordance knowledge. The underlying assumption will be that some combination of human users and artificial agents must complete a task together, with various data on performance being collected and analyzed. We identify several issues related to the kinds of data to collect, the tasks to use for studies, how the ``artificial'' partners are controlled. We treat this as a design space with trade-offs which we describe.