Scholarlink

Collaboration skills are essential 21st-century competencies for success in education, the workforce, and daily life. However, the extent to which these skills generalize across diverse task contexts remains underexplored. This study investigates collaboration processes, team performance, and their robustness across tasks using data from 709 four-person teams engaged in three types of online collaborative tasks: problem-solving, decision-making, and negotiation. Leveraging GPT-4o, we automatically categorized 58,298 turns of communication data with high accuracy and efficiency using a tailored coding framework. Epistemic network analysis was then applied to model collaboration processes and examine how collaboration behaviors manifest within and across tasks. Results revealed that while teams exhibited consistent communication quantity and intensity across tasks, distinct collaboration patterns emerged based on the unique demands of each task type. Unlike collaboration behaviors, team performance did not significantly correlate across different types of tasks, likely due to its reliance on domain-specific factors such as prior knowledge. These findings offer insights into the design of valid and reliable computer-based assessments of collaboration skills across diverse contexts.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Uncovering Transferable Collaboration Patterns Across Tasks Using Large Language Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yang Jiang,
  • Jiangang Hao,
  • Wenju Cui,
  • Emily Kerzabi,
  • Patrick Kyllonen

摘要

Collaboration skills are essential 21st-century competencies for success in education, the workforce, and daily life. However, the extent to which these skills generalize across diverse task contexts remains underexplored. This study investigates collaboration processes, team performance, and their robustness across tasks using data from 709 four-person teams engaged in three types of online collaborative tasks: problem-solving, decision-making, and negotiation. Leveraging GPT-4o, we automatically categorized 58,298 turns of communication data with high accuracy and efficiency using a tailored coding framework. Epistemic network analysis was then applied to model collaboration processes and examine how collaboration behaviors manifest within and across tasks. Results revealed that while teams exhibited consistent communication quantity and intensity across tasks, distinct collaboration patterns emerged based on the unique demands of each task type. Unlike collaboration behaviors, team performance did not significantly correlate across different types of tasks, likely due to its reliance on domain-specific factors such as prior knowledge. These findings offer insights into the design of valid and reliable computer-based assessments of collaboration skills across diverse contexts.