Conceptual alignment in a joint picture-naming task performed with a social robot

Area of Science:

• Cognitive Science
• Human-Robot Interaction
• Computational Linguistics

Background:

• Understanding human conceptual alignment is crucial for effective human-robot collaboration.
• Previous research has primarily focused on human-human interaction, leaving human-robot conceptual alignment under-explored.

Purpose of the Study:

• To investigate whether humans exhibit conceptual alignment when performing a language task with a robot.
• To determine if adaptation to a robot's conceptual choices influences human word selection.

Main Methods:

• A joint picture-naming task was conducted with 24 French native speakers and a robot.
• Participants and the robot took turns naming objects from 15 semantic categories.
• The robot was programmed to use superordinate category names for a subset of categories.

Main Results:

• Participants' responses showed alignment with the robot's conceptual choices.
• Participants increasingly used superordinate names when objects belonged to categories where the robot also used superordinate names.
• This indicates adaptation to the robot's linguistic behavior.

Conclusions:

• Humans demonstrate conceptual alignment with robots during language tasks.
• Adaptation to a robot partner can influence human word choices, even in non-human interaction.
• Findings suggest a basis for more natural and intuitive human-robot communication.