NIE x RI collab on generative AI for STEM assessment

Teachers, what if your go-to AI chatbot could craft top-notch STEM assessment contents in seconds? A recent collaborative study from Raffles Institution (RI) and National Institute of Education (NIE) has shown that cutting-edge generative AI technologies like GPT-4 can do just that but with an important twist. Researchers combined the well-known "chain-of-thought" prompting where chatbots are instructed to consider intermediate reasoning steps, with an additional novel prompt to use coding language. This strategy achieved much higher accuracy and quality, compared to other approaches, in generating challenging questions and multistep solutions for numerous topics in chemistry, physics, and mathematics. The work opens doors to accurate, personalized, and scalable generation of contents for learning and assessment.

 

The work is a product of a year-long collaboration between a group of STEM teachers from Raffles Institution led by Chemistry teacher Kuang Wen Chan, and researchers from two departments in NIE, Farhan Ali from Learning Sciences and Assessment and Joonhyeong Park from Natural Sciences and Science Education.

The research is free to read at:

Chan et al. (2025). Automatic item generation in various STEM subjects using large language model prompting. Computers and Education: Artificial Intelligence, 8, 100344.

Teachers can refer to the supplementary information here that provides a rich source of prompting strategies for various STEM subjects and topics.