Supporting Ambitious and Equitable Science Learning: Leveraging Artificial Intelligence in Developing Multimodal Formative Feedback

This project plans to investigate fundamental questions in STEM learning and learning environments, specifically how to design and integrate artificial intelligence (AI) to support equitable K-12 science education. The project will explore a promising solution involving multimodal AI to design automated feedback for formative science assessments. Insights from this research will illustrate how AI feedback can be applicable to a variety of assessment modalities and science content, promote students’ science identities, and inform instructional goals and practices.

Multimodal formative assessments—incorporating textual, audio, and visual representations—create expansive spaces for students to demonstrate their scientific understanding. Following the assessments, in-time feedback that highlights and further elicits students’ ideas can deepen their science practices and participation. This project will generate design principles to develop in-time AI feedback for multimodal formative assessment, and understanding of how such feedback supports science learning and teaching.

There will be three project phases mapping onto the three project years. Year 1 will include three iterative design cycles to develop, evaluate, and refine the AI feedback using large multimodal models. These cycles will be guided by ambitious and equitable science learning frameworks.

They will be conducted in collaboration with a Design Team of middle school teachers and the Advisory Board, along with iterative testing with teachers and students in different assessment contexts. This approach ensures that the designed feedback is ecologically valid and aligned with a range of instructional settings. Year 2 will involve a year-long classroom pilot of different AI feedback conditions (static versus interactive feedback) with 60 sixth graders, to examine how the feedback supports science practices, science identities, and instructional practices. Year 3 will focus on analysis and dissemination.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.