Table 2 Coding scheme for moderating variables.

Grade level

1. Primary (ages 7–13)

2. Secondary (ages 14–17)

3. College (ages over 18)

Type of course

1. STEM and related courses (programming, mathematics, science, physics, etc.)

2. Language learning and academic writing (English, academic writing courses, research methodology courses, etc.)

3. Skills and competencies development (teacher training courses, developing creative problem-solving skills courses, etc)

Learning model

1. Personalized learning (An instructional approach that tailors learning experiences to individual student's needs, preferences, and pace, as exemplified by Song and Song (2023))

2. Problem-based learning (A student-centered method where learners develop knowledge and skills by solving authentic, real-world problems, as demonstrated by Urban et al. (2024))

3. Project-based learning (An active learning approach in which students engage in extended projects to explore complex topics, as illustrated in Küchemann et al. (2023))

4. Contextual learning (A learning approach that situates knowledge within specific, real-life contexts, as shown in Lu et al. (2024))

5. Reflective learning (A process of reviewing, analyzing, and evaluating one’s learning process or outcomes, followed by necessary adjustments, as exemplified by Escalante et al. (2023))

6. Mixed

Duration

1. ≤1 week

2. 1–4 weeks

3. 4–8 weeks

4. >8 weeks

Role of ChatGPT

1. Intelligent tutor (primarily functions as a teacher, providing personalized guidance, feedback, and assessment to students, as exemplified by Escalante et al. (2023))

2. Intelligent partner (acts as a peer, engaging in dialogue and interaction with students to simulate the behavior of classmates, as shown in Li et al. (2024))

3. Intelligent learning tool (focuses more on offering auxiliary functions, such as resource recommendations, knowledge retrieval, and learning progress tracking, without simulating human interaction, as exemplified by Song and Song (2023))

4. Mixed

Area of ChatGPT application

1. Assessment and evaluation (using ChatGPT to provide assessments and feedback on student learning, as exemplified by Escalante et al. (2023))

2. Tutoring (mainly refers to students using ChatGPT on their own during or after class to solve difficulties encountered in learning, as shown in Essel et al. (2024))

3. Personalized recommendation (mainly refers to ChatGPT providing corresponding learning resources and suggestions based on student's individual needs, as exemplified by Song and Song (2023))

4. Mixed

Learning Effects

1. Higher-order thinking (primarily measured through scales and questionnaires, as shown in Lu et al. (2024))

2. Learning perception (primarily assessed through scales and questionnaires, as shown in Chen and Chang (2024))

3. Learning performance (primarily evaluated through tests and exams, as demonstrated in Urban et al. (2024))