Developing a generative AI chatbot to support students' higher-order thinking in STEM education

Higher-order thinking is a key component of science, technology, engineering, mathematics (STEM) reasoning and supports students' ability to reflect on their ideas, make connections, and generalize across contexts. We report a new approach to the use of generative AI in STEM education, in which we are developing a chatbot that aims to engage students in higher-order thinking, rather than giving instructional content. We describe a proof of concept test of whether interaction with our generative artificial intell…

By the Truvace deskOctober 15, 20262 min read1 sourcePeer-reviewedTRV-2026-1384· cite & verify

Developing a generative AI chatbot to support students' higher-order thinking in STEM education — "Chatbots linked" by Ashhstarr is licensed under CC BY-SA 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-sa/4.0/.

In brief

Researchers built a generative AI chatbot, the UCI Math ThinkBot, meant to prompt students to reflect rather than deliver instructional content. Thirty-three fifth- and sixth-graders were taught ratio comparison strategies in a 30-minute lesson and completed an assessment. They then discussed their solutions with the chatbot, and the transcripts were analyzed.

Eliciting higher-order thinking is hard in ordinary classroom discourse, so a tool that prompts it could support STEM reasoning. The work is a proof of concept, and the authors describe the findings as suggestive. It remains unclear whether the effect holds at larger scale, or whether it improves learning.

Main points

The chatbot was designed to engage students in higher-order thinking instead of giving instructional content.
The sample was 33 fifth- and sixth-grade students working on proportional reasoning problems.
Students first received a 30-minute lesson on symbolic and number line strategies for comparing ratios.
Transcripts were analyzed both qualitatively and quantitatively for higher-order thinking talk.
Children's higher-order thinking talk was positively associated with the chatbot's.

The gain

In a pilot with 33 fifth- and sixth-grade students, a prompting chatbot drew out higher-order thinking talk about ratio strategies, and students' talk rose with the chatbot's own.

The rundown

The study tested whether dialog with the Math ThinkBot could elicit higher-order thinking talk, abbreviated HOTT. Students explained their solutions and compared strategies in conversation with the chatbot after completing the lesson and assessment.

Analysis of the transcripts found HOTT in both the chatbot's prompts and the children's responses. The authors note this matters because eliciting such talk can be a challenge in classroom discourse.

Children's HOTT was positively associated with the HOTT the chatbot produced. The authors conclude that human-AI dialog can lead to instructional HOTT discourse and may be a promising approach for supporting higher-order thinking and STEM reasoning.