Supporting Math Understanding in Title I Schools: Creating a Text Answer-AI-Tutor (TAAIT)

The project team will conduct 5 rounds of interviews and prototype adjustments for the student, teacher, and instructional-facing experiences. They will test different methods for improving the ability to score and produce feedback for students and a concept of training in the one in-house LLM. These tests will answer the question that all S2S Phase 1 projects need to answer: Schools "believe the solution could meet their needs." 

Title I middle schools that assign problems from Illustrative Mathematics, a freely available OER math curriculum used across all states in the U.S will participate in the project. Once designed and validated, TAAIT will be a feature in ASSISTments that provides immediate feedback and support to students when they submit text answers to open response problems. It will also provide actionable data to teachers and school-level instructional leaders. 

The R&D plan will include research design and methods. It will include both (1) micro studies at the student-level using our data instrumentation infrastructure that has a randomization engine (to implement quick AB experiments) and (2) a summative typical IES-efficacy study that uses school-level randomization that allows us to compare learning with and without TAAIT. For the micro studies, we will compare two different prompts to create the feedback. For the summative study the control condition will be with ASSISTments but not TAAIT (compared to ASSISTments+TAAIT). For the micro studies, we will compare how well the student does on the next problem. For the school-level studies, our ultimate goal will be to compare differences in state test scores.