Table of Links
2. Background
2.1 Effective Tutoring Practice
2.2 Feedback for Tutor Training
2.3 Sequence Labeling for Feedback Generation
2.4 Large Language Models in Education
3. Method
3.1 Dataset and 3.2 Sequence Labeling
3.3 GPT Facilitated Sequence Labeling
4. Results
6. Limitation and Future Works
APPENDIX
B. Input for Fine-Tunning GPT-3.5
C. Scatter Matric of the Correlation on the Outcome-based Praise
D. Detailed Results of Fine-Tuned GPT-3.5 Model's Performance
2.1 Effective Tutoring Practice
Effective tutoring plays an important role in enhancing student learning by integrating academic knowledge with the capability to address students’ socio-motivational needs [13, 19, 48, 39]. However, equipping tutors with these skills proves challenging, given the limited active learning opportunities that bring situational, scenario-based experiences to the professional development of tutors [7]. Thus, current tutor training for tutors in addressing the social-emotional and motivational aspects of student learning remain underdeveloped[7, 52].
Our study focuses on a particular aspect of tutoring practice: the delivery of effective praise. Praise is a fundamental tutoring practice during the human tutoring process, consistently shown to have a positive impact on student motivation, engagement, and learning outcomes[25, 29, 55]. Research highlights that for praise to be effective, it should be: (1) sincere, earned, and truthful; (2) specific by giving details of a student’s strengths; (3) immediate, with praise given right after the student’s action; (4) authentic, avoiding repetitive phrases like “great job” which diminishes meaning and becomes predictable, and (5) focused on the learning process rather than innate ability [55]. Existing literature categorizes praise into three types: effort-based Effort, outcome-based Outcome, and person-based Person [29, 55, 8, 7]. Effort-based praise emphasizes the student’s learning process (e.g., “I love your effort that you put into the writing...”). Outcome-based praise highlights a student’s achievements, like scoring high on an assignment or solving a problem correctly, and it’s sometimes linked to less effective praise strategies such as “Good job!”. Person-based praise attributes success to innate qualities beyond the student’s control such as “You are smart!” and is often, similar to outcome-focused praise, considered less effective [29].
Training novice tutors to provide more effective praise (i.e., effort-based praise) requires a comprehensive understanding of both the desirable and less favorable elements of their praise responses. For tutors to refine their skills effectively, they should engage in a feedback process to know how well their responses align with the effective praise in tutoring[7, 55]. However, manual feedback generation by expert tutors poses significant challenges due to its time-consuming and labor-intensive nature. This underscores the necessity for exploring automated feedback systems within tutor training programs. Such systems could offer scalable and timely feedback, thereby enhancing tutors’ ability to effectively address student motivation issues.
This paper is available on arxiv under CC BY 4.0 DEED license.
Authors:
(1) Jionghao Lin, Carnegie Mellon University (jionghal@cs.cmu.edu);
(2) Eason Chen, Carnegie Mellon University (easonc13@cmu.edu);
(3) Zeifei Han, University of Toronto (feifei.han@mail.utoronto.ca);
(4) Ashish Gurung, Carnegie Mellon University (agurung@andrew.cmu.edu);
(5) Danielle R. Thomas, Carnegie Mellon University (drthomas@cmu.edu);
(6) Wei Tan, Monash University (wei.tan2@monash.edu);
(7) Ngoc Dang Nguyen, Monash University (dan.nguyen2@monash.edu);
(8) Kenneth R. Koedinger, Carnegie Mellon University (koedinger@cmu.edu).