Transforming Item Generation in Modern Assessment Systems with Generative AI

• In Event: Automatic Item Generation

Abstract

This study examines how generative AI can aid the process of item generation. Items generated by prompting the ChatGPT 4.0 model are compared with items generated by humans in terms of costs and psychometric properties. Meticulous psychometric linking design is applied to ensure the comparability of the item parameters.

Authors

• Yunting Liu

• shreya bhandari, University of California, Berkeley

• Zachary A. Pardos, UC Berkeley