Linguistic alignment with an artificial agent: A commentary and re-analysis.

Area of Science:

• Cognitive Psychology
• Human-Robot Interaction
• Linguistics

Background:

• Cirillo et al. (2022) suggested automatic conceptual alignment in joint picture naming with a social robot based on response proportions.
• Their study examined adaptation to the robot's lexical choices, specifically category vs. basic names.
• This work provides a commentary and complementary analysis of the existing dataset.

Purpose of the Study:

• To conduct a complementary analysis of conceptual alignment using response times (RTs) as a measure of cognitive processing.
• To investigate the automaticity versus strategic nature of alignment in a joint picture naming task with a social robot.
• To assess the generalizability of observed alignment effects.

Main Methods:

• Re-analysis of the openly available dataset from Cirillo et al. (2022).
• Employed a multiverse approach focusing on response times (RTs).
• Conducted comprehensive visual exploration of response proportions and leave-one-out robustness checks.

Main Results:

• Alignment in the Category condition (superordinate label) correlated with longer RTs and greater variability, suggesting strategic processing.
• Alignment in the Basic condition showed shorter RTs and reduced variability, consistent with the basic-level advantage.
• Alignment effects were primarily observed in specific semantic categories and for low-frequency or novel stimuli, indicating limited generalizability.

Conclusions:

• Findings support the view that aligning with category labels in this task is a strategic, effortful process, not an automatic one.
• The limited generalizability suggests that observed effects may reflect individual task strategies rather than true linguistic alignment.
• Recommends future research directions for linguistic alignment, emphasizing Open Science principles.