Electrophysiological markers of adaptive co-representation in joint language production: Evidence from human–robot interaction

Authors:
Publisher website
Google Scholar
Abstract
This study aimed to assess the extent to which human participants co-represent the lexico-semantic processing of a humanoid robot partner. Specifically, we investigated whether participants would engage their speech production system to predict the robot’s upcoming words, and how they would progressively adapt to the robot’s verbal behaviour. In the experiment, a human participant and a robot alternated in naming pictures of objects from 15 semantic categories, while the participant’s electrophysiological activity was recorded. We manipulated word frequency as a measure of lexical access, with half of the pictures associated with high-frequency names and the other half with low-frequency names. In addition, the robot was programmed to provide semantic category labels (e.g., “tool” for the picture of a hammer) instead of the more typical basic-level names (e.g., “hammer”) for items in five categories. Analysis of the stimulus-locked activity revealed a comparable event-related potential (ERP) associated with word frequency both when it was the participant’s and the robot’s turn to speak. Analysis of the response-locked activity showed a different pattern for the category and basic-level responses in the first but not in the second part of the experiment, suggesting that participants adapted to the robot’s lexico-semantic patterns over time. These findings provide empirical evidence for two key points: (1) participants engage their speech production system to predict the robot’s upcoming words and (2) partner-adaptive behaviour facilitates comprehension of the robot’s speech.
Funding Information
  • Ramón y Cajal Research Program (RYC2018-026174-I)
  • Agence Nationale de la Recherche (ANR-18- CE28-0013)
  • Agence Nationale de la Recherche (ANR-18-FRAL-0013-01)
  • Aix-Marseille Université

This publication has 89 references indexed in Scilit: