Supporting Undergraduate Students' Learning in Practical Chemistry Courses through AI-Supported Experimental Design

Synthesis: Yim and Lui (2026) report an authentic pedagogical approach integrating AI chatbots into an upper-division undergraduate analytical chemistry laboratory course: students used AI to design lab manuals for analyzing real-world samples, then implemented them through hands-on experimentation, with the manuals reviewed by independent testing and certification professionals. Surveys and focus-group Feedback indicated significantly enhanced student confidence in conducting experiments and soft skills — critical thinking, Problem Solving, analytical abilities, and experimental design. A key finding was the shift in staff roles from the traditional "cookbook" demonstration model toward guiding and advising students as they address unexpected issues from student-designed lab manuals.

The pedagogical approach

• Authentic task: students designed a lab manual for analyzing real-world samples using AI, then implemented it hands-on.
• Validation: lab manuals were reviewed by independent testing and certification professionals to ensure the accuracy and reliability of AI-generated content.
• Evaluation: surveys and focus-group interviews with students and staff.

Findings

• Confidence and soft skills: the approach significantly enhanced students' confidence in conducting experiments and their critical thinking, problem-solving, analytical abilities, and experimental design skills.
• AI's role: AI was particularly beneficial for ideation, scoping, and managing language-related tasks; with effective prompting it provided useful background information and generic experimental procedures.
• Student discernment: many students preferred to consult their own literature reviews for critical details, expressing concerns about potential inaccuracies in chemical calculations and the reliability of advanced chemistry content — evidence of evaluative engagement with AI output.
• Staff role shift: instead of demonstrating established experiments ("cookbook"), staff now encounter a broader range of unexpected issues arising from student-designed lab manuals, focusing on guiding and advising rather than supplying definitive answers.

This complements the knowledge base's framing of critical AI use and human–AI collaboration: students learned to interrogate and verify AI-generated content rather than accept it uncritically.

What this means for practice

• Instructors. Have students design their own lab manual with AI support and then implement it themselves: in this upper-division analytical chemistry course the design-then-implement sequence raised confidence in experimental design from a median of 3 with 52% neutral (n = 29) before the project to a median of 4 with 60% agreeing or strongly agreeing (n = 32) afterward.
• Instructors. Route AI-generated manuals through independent testing and certification professionals — students cited that external validation, alongside successful completion of their own experiment, as a main reason for their confidence gains.
• Instructors. Plan for the staff role shift from demonstrating "cookbook" experiments to guiding and advising students through unexpected problems, and budget for the significant workload increase staff reported, which varied with experiment complexity.
• Learners. Use AI for ideation, scoping, and language-related tasks while verifying critical details against your own literature review — the students in this course were wary of inaccuracies in chemical calculations and advanced chemistry content and checked them independently.

Limitations

• The evaluation rests on 31 students and 3 staff who volunteered for pre/post surveys and focus groups in a single upper-division laboratory course run once, so students who declined may hold different views.
• Interview coverage was partial: data came from 7 of the 8 laboratory groups, one group did not attend at all, and two groups were represented by only a single member.
• Groups of 4–5 students, forced by laboratory space and equipment limits, meant individuals took part only in assigned steps — as one student put it, during standard solution preparation "only two students could participate, while the others simply waited behind them" — so self-reported confidence may rest on incomplete understanding.
• All outcomes are self-assessments and focus group perceptions, with no comparison group and no objective measure of experimental design skill.

Citation

Yim, K.-H., & Lui, M. Y. (2026). Supporting undergraduate students' learning in practical chemistry courses through AI-supported experimental design. Journal of Chemical Education, 103(6), 3022–3030.