Do we talk differently to humans and robots when assembling a shelf together?

What is it about?

The nature of interactions likely differs when humans talk to robots compared to other humans. However, it is unclear how they differ in terms of the linguistic dimension of speech acts - i.e. utterances performing different actions, e.g. directives, expressives, or commissives. This study examines the speech acts of adult German speakers in a Wizard-of-Oz experiment involving the collaborative assembly of an IKEA shelf with either another human or a robot arm. In our experiment, the wizard tele-manipulates the robot and plays predefined responses. We compared audio and video recordings of human–robot (H-R) and human–human (H-H) interactions, categorizing them based on speech acts and lexical complexity. This study is the first to provide a comparative linguistic-pragmatic perspective on H-R and H-H interaction in an assembly task, with potential implications for robot design for fluent language-based cooperation. Our results indicate that H-R communication is more directive, simpler, and wordier than H-H in a comparable situation.

Featured Image