Controlling or directing? Text mining to decode supervisor-graduate student relationship

Area of Science:

• Educational Psychology
• Higher Education Studies
• Data Science in Education

Background:

• The Supervisor-Graduate Student Relationship (SGSR) is crucial for academic success and student well-being.
• Existing models of SGSR lack a nuanced, data-driven typology.
• Understanding SGSR variations across disciplines and geo-economic contexts is essential.

Purpose of the Study:

• To develop a novel, data-driven typology of Supervisor-Graduate Student Relationships (SGSR).
• To introduce a 2x2 quadrant model based on control and direction levels.
• To explore disciplinary and geo-economic variations in SGSR types.

Main Methods:

• Text mining of 25,219 evaluation texts from 445 Chinese universities.
• Application of BERT-TextCNN, BERTopic modeling, and BERT embedding-based semantic similarity.
• Analysis of relationship dynamics based on control and direction levels.

Main Results:

• Identification of four SGSR types: Juice Coach, Shadow Hermit, Pilot Captain, and Alchemy Master.
• Social sciences predominantly feature Alchemy Master styles.
• Economically underdeveloped regions exhibit stronger supervisory control.

Conclusions:

• The proposed SGSR model offers theoretical insights and practical applications for higher education institutions.
• Institutions can use this framework as an early warning system to monitor supervisory quality and support students.
• The data-driven approach enables scalable interventions to improve research supervision and protect student well-being.