ML-SPEAK: A theory-guided machine learning method for studying and predicting conversational turn-taking patterns

Area of Science:

• Psychological Sciences
• Computational Social Science
• Team Dynamics

Background:

• Predicting team dynamics from personality traits is a persistent challenge.
• Existing models like input-process-output lack dynamic capabilities for complex team interactions.
• Understanding team composition's impact on processes is crucial for research and practical applications.

Purpose of the Study:

• To develop a computational model for analyzing conversational turn-taking in self-organized teams.
• To investigate the relationship between individual personality traits and team communication dynamics.
• To predict group communication patterns based solely on team trait composition.

Main Methods:

• Developed the ML-SPEAK computational model focusing on conversational turn-taking patterns.
• Trained the model on conversational data from teams with known trait compositions.
• Evaluated model performance using simulated data and real-world data from student teams.

Main Results:

• The ML-SPEAK model accurately predicts speaking turn sequences, outperforming baseline models.
• The model reveals novel relationships between team members' personality traits and their communication patterns.
• Demonstrated the model's ability to predict group communication based on team trait composition.

Conclusions:

• The ML-SPEAK model provides a data-driven, dynamic approach to understanding team processes.
• It bridges the gap between individual characteristics and emergent team communication patterns.
• Offers potential for informing team process theories and optimizing team staffing and training.