Learning to Be Inclusive Through Large Language Models

• In Event: AERA Graduate Student Research-in-Progress Session 5
  In Roundtable Session: Identity, Engagement, and Inclusion in Learning and Instruction (Table 15)

Abstract

Achieving inclusion, the degree to which individuals experience belongingness and uniqueness within a group (Shore, et. al., 2011), has been a recognized goal within educational contexts due to inclusion's connections to improved collaboration, sustained learning, and equitable instruction. Since scaling measurement and providing automated feedback with existing qualitative and self report methods is not feasible, this dissertation focuses on operationalizing inclusion as a set of observable, multimodal (linguistic and gestural) behaviors that can be detected by large language models, which provide opportunities for automatic detection and generation of near-immediate feedback on inclusion quality. The final model is tested in an experimental design where groups of learners complete tasks and receive feedback on inclusion quality with the model.

Author

• Armanda Lewis, New York University