The effects of having a robot as a task partner on lexical retrieval

Previous research suggests that a task partner’s speaking affects own language production. In single-subject settings, naming latencies for semantically related objects increase with each additional semantic category member (cumulative semantic interference effect, Howard et al., (2006). In dual-subject settings, in which two partners take turns naming pictures, naming latencies for semantically related objects also increase by semantic category members named by the partner, suggesting a simulation of lexicalization processes on behalf of the task partner (Kuhlen & Abdel Rahman, 2017). In the current experiment we investigate whether this partner elicited semantic interference extends to a joint task setting with a humanoid robot (Pepper, SoftBank Robotics) as partner, assessing the role of partner identity on lexical simulation. Thirty-six participants (preregistered at AsPredicted.Org, 2019), name semantically related objects either by themselves, or by taking turns with a humanoid robot – Pepper. We assess the degree of cumulative semantic interference experienced when previous semantically related pictures are named by the robot (Joint Naming) as compared to when they are presented visually only (Single Naming). Additionally, we assess whether differences in robot perception, attitudes to technology, and individual differences in personality and social competence affect the degree of partner-elicited semantic interference when interacting with a robot. Naming latencies for the trials in which participants name the picture are modelled using linear mixed models as a function of the predictors: Naming Condition (Joint Naming vs. Single Naming), Ordinal Position (ordinal position 1 to 5), and Experimental Block (1 to 2). If the robot is conceived similar to a human task partner we expect participants’ naming latencies to increase not only in response to the number of semantically related pictures participants previously named themselves, but also additionally, in response to semantically related pictures named by the robot. If the robot is not conceived similar to a human task partner, we do not expect to find an additional slowing of naming latencies as a result of the robot naming the semantically related pictures (i.e. no significant Naming Condition x Ordinal Position interaction). The data collection for the project has commenced with 19 participants but paused due to the Coronavirus/COVID-19 lockdown. The results of this study will contribute novel insights to our understanding whether having a robot as a task partner affects one’s own language production. Further, the results will shed light on the role of the task partner’s identity on simulating their lexicalization processes.