Self-Repair Increases Referential Coordination

Area of Science:

• Psycholinguistics
• Computational Linguistics
• Human-Computer Interaction

Background:

• Dialogue involves convergence on referring expressions, becoming more abstract over time.
• Interactive repair mechanisms are thought to drive this convergence.
• Previous research focused on other-initiated repair, neglecting self-initiated repair.

Purpose of the Study:

• To investigate the role of self-initiated repair in dialogue convergence.
• To examine how artificially generated self-repairs affect referring expression development.

Main Methods:

• A computer-mediated maze task experiment was conducted.
• Participants communicated via a chat tool that converted private revisions into public self-repairs.
• Turn revisions were automatically detected and displayed as public self-repairs.

Main Results:

• Participants exposed to public self-repairs used more abstract and systematized referring expressions.
• Task performance decreased for participants receiving these transformed turns.
• Increased effort in diagnosing and resolving coordination problems was observed.

Conclusions:

• Self-initiated repair, even when artificial, can enhance the abstraction of referring expressions.
• This abstraction comes at the cost of increased cognitive load and reduced task efficiency.
• Findings suggest a trade-off between linguistic convergence and task performance in mediated communication.