By Nadeeshani Jayasena (Portland State University, OR)
Information
This poster presents the preliminary results of a Spring 2026 pilot study at Portland State University (PSU) involving the integration of MathGPT.ai into the "Application of Statistics for Business" curriculum. Unlike general-purpose LLMs, MathGPT.ai is designed to provide Socratic tutoring for quantitative problems. This research explores how students in two distinct environments (one in-person and one fully online) utilize the AI platform to navigate complex statistical concepts like probability distributions, hypothesis tests and linear regression. We examine whether the AI serves as a "first-responder tutor" and how this impact student confidence and problem-solving persistence. This was implemented at PSU, a large, urban, public research university with primarily undergraduate Business majors (Freshman/Sophomores) and two sections of approximately 35 students each. One section meets in person (F2F) twice weekly; the other is fully asynchronous online. Both sections use the same curriculum and assessments, providing a controlled environment to test the AI's impact across different learning modalities. We are evaluating efficacy through a mixed-methods approach including: Pre/Post-Pilot Surveys: Measuring "Statistics Anxiety" and "Attitudes Toward AI" at the beginning and middle of the Spring term. Usage Analytics: Analyzing MathGPT.ai logs to track the "depth" of student conversations (number of follow-up questions vs. one-off queries). Comparative Performance: Tracking quiz and midterm scores against historical averages for these PSU sections. Student Reflections: Short qualitative "exit tickets" where students describe one specific concept that the AI helped them "discover" or understand more deeply.