Ink and Algorithm: Exploring Temporal Dynamics in Human-AI Collaborative Writing
Project Overview
The document explores the integration of Generative Artificial Intelligence (GAI) in collaborative writing within educational settings, emphasizing tools like CoAuthor that enhance human-AI interaction during writing tasks. Through a study utilizing time-series clustering, it examines how GAI is employed over time and its effects on human writing behaviors, uncovering specific usage patterns that influence cognitive processes involved in writing. The findings indicate that recognizing these distinct patterns can assist educators in creating more effective writing tools that cater to the varied needs of students. Ultimately, the study underscores the potential of GAI not only to support collaborative writing but also to inform teaching strategies and enhance learning outcomes in educational contexts.
Key Applications
CoAuthor
Context: Collaborative writing tasks in educational settings, targeting students and writers.
Implementation: Used a dataset from 1,445 writing sessions to analyze interaction patterns with GAI.
Outcomes: Identified four distinct temporal patterns of GAI usage, revealing correlations with cognitive writing behaviors.
Challenges: Limited understanding of the dynamics of human-AI interactions and reliance on coarse categorization of AI usage.
Implementation Barriers
Technical
Lack of understanding of how human learning can be effectively supported with GAI technologies.
Proposed Solutions: Further research into human-AI collaborative processes and the development of tailored writing tools.
Adoption
Low trust in AI suggestions among some users, leading to reluctance in using GAI.
Proposed Solutions: Enhancing the quality of AI-generated suggestions to better align with user expectations.
Project Team
Kaixun Yang
Researcher
Yixin Cheng
Researcher
Linxuan Zhao
Researcher
Mladen Raković
Researcher
Zachari Swiecki
Researcher
Dragan Gašević
Researcher
Guanliang Chen
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Kaixun Yang, Yixin Cheng, Linxuan Zhao, Mladen Raković, Zachari Swiecki, Dragan Gašević, Guanliang Chen
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai