Rex to the rescue: Evaluating the effectiveness of real-time AI-driven stealth assessment and support for mathematical modeling in virtual science investigations

Stealth assessment, rooted in evidence-centered design (ECD), provides an unobtrusive means of assessing students’ competencies in online environments. This paper presents a formative, performance-based assessment in the Inquiry Intelligent Tutoring System (Inq-ITS), designed using ECD and integrating AI-driven stealth assessment to provide targeted, real-time support for students’ mathematical modeling in science investigations. We examined the effectiveness of the targeted support from an AI-based pedagogical agent, Rex, in ninth-grade physical science classrooms. Results showed that students with lower math proficiency had significant learning gains when Rex was available, compared to similar students without Rex. These findings highlight the power of stealth assessment to provide real-time data and support, while offering empirical evidence of how students are learning essential science and math competencies aligned to the NGSS.