AIED

The Limits of Text-Only LLMs for Think-Aloud Transcript Correction: A Multimodal Comparison in the Context of a Puzzle Video Game

Mon Jun 29, 9:30 AM–9:45 AM · North 210

Part of Multimodal LLMs for Educational Understanding and Assessment

Generative AI & Large Language Models Multimodal Learning Analytics NLP & Discourse Analysis

Compares text-only and multimodal LLMs for correcting think-aloud transcripts, revealing the limits of text-only approaches in a puzzle video game context.

Authors

Nasim Eshgarf, Zack Carpenter, Craig Anderson