Final Statement: Chinese higher education students perceive Generative AI as a multifaceted support for academic writing that enhances the writing process, improves performance, and positively affects the affective domain, despite challenges related to AI limitations, student capabilities, and task constraints.

Objective: The main goal of this study was to explore Chinese higher education students' perceptions and experiences of Generative AI-assisted academic writing, specifically examining the expected roles of AI in writing tasks, perceived advantages, and potential challenges.

Methods: The researchers conducted a qualitative study with 20 Chinese students from a Sino-British international university, spanning bachelor's, master's, and doctorate levels with varying AI literacy and English proficiency levels. Participants completed academic writing tasks using a ChatGPT4-embedded writing system developed by the research team. The researchers then conducted semi-structured interviews (60-90 minutes) to gather data on students' experiences, perceptions, and perspectives. Data was analyzed using both inductive and deductive thematic analysis.

Key Findings:

• Students expected GenAI to serve in three primary roles:
  1. Multi-tasking writing assistant (search engine, thought provoker, entry-level writer, proofreader)
  2. Private virtual tutor on-demand
  3. Digital peer providing social support during the writing process
• Students identified several advantages of GenAI-assisted writing:
  1. Productive writing process - AI enhanced ideation, planning, drafting, and revision stages
  2. Improved writing performance - Better quality, faster writing speed, and expanded topic knowledge
  3. Enhanced affective domain - Increased joy in writing, fostered question generation, provided perceived support, and promoted self-efficacy
• Students highlighted multiple challenges with GenAI-assisted writing:
  1. AI-related limitations - Hallucinations, lack of contextual understanding, absence of higher-order thinking, limited human awareness, poor cultural awareness in language, inadequate relationship and pedagogical skills, lack of interoperability and explainability
  2. Student-related barriers - Insufficient AI literacy, negative attitudes toward AI, limited higher-order thinking, inadequate task topic knowledge, and deficient writing skills
  3. Task-related constraints - Time limitations affecting effective interaction with AI

Implications: This study contributes to understanding GenAI's impact on academic writing by capturing student perspectives, offering implications for both educational AI design and instructional approaches. The findings suggest:

1. Need for developing human-centered AI in education that prioritizes students' needs, characteristics, and experiences
2. Importance of cultivating student capabilities in prompt engineering to effectively communicate with AI systems
3. Value of fostering higher-order thinking skills to optimize AI-assisted learning
4. Potential for applying post-humanist theories like Actor-Network Theory to conceptualize student-AI interactions as a symbiotic relationship

The research indicates that while GenAI can significantly enhance the academic writing experience, its implementation requires careful consideration of both technical capabilities and pedagogical integration.

Limitations: The study has several limitations:

1. Small sample size of 20 Chinese students from a single international university, which may not fully represent diverse student perspectives
2. Participants' one-time interaction with the research team's system may not reveal perceptions that evolve with extended use
3. Context-specific findings from a Sino-British university environment may limit generalizability
4. Focus on Chinese students may not capture perspectives of students from different cultural backgrounds

Future Directions: The researchers suggest future research should:

1. Employ quantitative methods with larger sample sizes and diverse student populations
2. Examine different learning tasks beyond academic writing (argumentative discussions, creative writing)
3. Consider various student characteristics (attitude toward AI, interaction fluency)
4. Implement longitudinal research designs in actual classroom settings to explore perception changes over time
5. Utilize qualitative data analysis software or text-mining techniques for alternative data analysis approaches
6. Develop comprehensive theoretical frameworks for understanding student-AI interactions in educational contexts
7. Integrate insights from cognitive and behavioral learning theories to inform AI design in educational settings

Title and Authors: "Exploring students' perspectives on Generative AI-assisted academic writing" by Jinhee Kim, Seongryeong Yu, Rita Detrick, and Na Li.

Published On: July 31, 2024

Published By: Education and Information Technologies (2025) 30:1265–1300, a Springer journal.