Instructor and AI Roles in the Chemistry Classroom: Future Science Teachers' Perceptions in a ChatGPT-Enhanced Formative Assessment

Synthesis: Ratniyom, Boonphadung, Intaraprasit, and Chumkaeo (2026) analyze how pre-service science teachers across achievement levels perceive the distinct roles of human instructors versus ChatGPT in a ChatGPT-enhanced formative assessment of stoichiometry. The human instructor was viewed as an adaptive expert, toggling between Simplifier and Elaborator roles depending on learner achievement; ChatGPT was perceived as a personalized self-regulated-learning tool whose role shifted by achievement — a Patient tutor for low-achievers, a Personal Coach for medium-achievers, and an Intellectual Sparring Partner for high-achievers. The authors propose an Instructor–AI Synergistic Learning Ecosystem model reframing human–AI collaboration as complementary partnership rather than competition.

Design

• Data: secondary qualitative analysis of interviews with pre-service science teachers, stratified into high-, medium-, and low-achieving groups.
• Context: ChatGPT-enhanced formative assessment in stoichiometry learning.
• Setting: 13 pre-service science teachers from a single Thai university (case study).

Findings: distinct instructor vs. GenAI roles

The human instructor was predominantly viewed as an adaptive expert, adept at toggling between two roles based on learners' achievement:

• Simplifier for lower-achieving students (breaking down complex content)
• Elaborator for higher-achieving students (extending and deepening understanding)

ChatGPT was perceived as a personalized tool for self-regulated learning, its role shifting with achievement:

• Patient Tutor for low-achievers — non-judgmental, always-available support
• Personal Coach for medium-achievers — guided practice and Feedback
• Intellectual Sparring Partner for high-achievers — challenging and probing ideas

The Instructor–AI Synergistic Learning Ecosystem model

The study's principal contribution is an Achievement-Based Instructor–AI Synergistic Learning Ecosystem model that reframes human–AI collaboration not as competition but as complementary partnership. It positions the instructor as the adaptive expert delivering context-aware feedback while GenAI functions as a personalized SRL tool scaling individualized formative feedback — turning the long-standing challenge of scaling formative assessment in large classes into a pedagogical opportunity. This connects directly to the knowledge base's Human AI Collaboration concept.

What this means for practice

• Instructors. Match the AI workflow to achievement level instead of running one procedure for the whole class: low-achievers in this study wanted step-by-step detail from ChatGPT ("show the detailed calculation, because I'm the type of person who wants to know where the numbers come from"), medium-achievers wanted guided practice and feedback, and high-achievers wanted their ideas challenged.
• Instructors. Keep the final explanation of difficult content yourself: high-achievers preferred the instructor on hard topics because AI explanations did not connect to what they had already learned, and students became confused when ChatGPT used a problem-solving strategy that diverged from the one taught in class.
• Instructors. Standardize the prompt every student must submit with their answers and require the generated feedback to be handed in, so the formative assessment record is comparable across the class.
• Instructors. Use the AI layer as the always-available first pass for routine checking and self-regulated practice, and reserve your own time for the adaptive, context-aware feedback that matters most.
• Faculty developers. Use the Achievement-Based Instructor–AI Synergistic Learning Ecosystem model as a blueprint in teacher preparation: have pre-service teachers design an AI-enhanced assessment activity and state explicitly which role the AI plays for which group of learners.

Limitations

The findings are context-specific (13 pre-service teachers, single university, stoichiometry with ChatGPT); perceptions relied on self-reported descriptions rather than full interaction logs; and the model lacks a developmental/longitudinal dimension.

Citation

Ratniyom, J., Boonphadung, S., Intaraprasit, M., & Chumkaeo, P. (2026). Instructor and AI roles in the chemistry classroom: Future science teachers' perceptions in a ChatGPT-enhanced formative assessment. Chemistry Teacher International, advance online publication.