Extended Abstract: The Impact of Interactive AI Feedback on STEM Students' Writing of Lay Research Summaries

Tzipora Rakedzon, Ayelet Baram-Tsabari, Elad Segev, Roy Yosef

Abstract: The integration of generative AI tools in academic writing, particularly for STEM students and researchers, is a growing trend. These tools may save time and enhance grammatical accuracy; however, they can also be used pedagogically to help students improve their own academic and science communication skills. Despite this, studies evaluating the impact of AI tools on students' writing outcomes, specifically in STEM, are limited. This research focuses on whether individualized AI-based feedback helps STEM graduate students improve their writing of short research summaries for a lay audience. To this end, students participated in the task in one of four feedback conditions: a group that receives feedback on appropriate jargon use before each iteration; a group that receives interactive feedback from ChatGPT with rewrite suggestions; a group that receives interactive feedback both about jargon and ChatGPT rewrite suggestions; and a comparison group that does not receive any feedback before each rewrite. Results indicate that AI-based feedback helped students improve their use of jargon and readability, though no significant improvement was observed in message distillation. In this presentation, we share the empirical research and available online task for use in STEM academic and science writing programs.

External IDs:dblp:conf/ipcc/RakedzonBSY24