[1A] Poster Presentation

Mathematics / Statistics

[1A-067-PF] Do Large Language Models Try to Look Good?Mitigating social desirability bias in LLMs using forced-choice measurement and a Thurstonian item response theory model

*Kensuke Okada¹, Yui Furukawa¹, Kyosuke Bunji² (1. Graduate School of Education, The University of Tokyo, 2. Graduate School of Business Administration, Kobe University)

Keywords:

forced-choice measurement,Thurstonian item response theory model,large language models