The Effectiveness of OpenAI GPT-Generated Definitions Versus Definitions from an English Learners' Dictionary in a Lexically Orientated Reading Task.

This article investigates the effectiveness of AI-generated definitions, specifically those produced by OpenAI's Generative Pretrained Transformer (GPT-3), compared to human-created definitions from the Macmillan English Dictionary (MED) in supporting second-language English learners during a lexically focused reading-comprehension task. The study involved 43 advanced L2 English university students in Spain who completed a multiple-choice reading task with access to either AI-generated definitions, MED definitions, or no definitions. Results showed that students with MED definitions performed significantly better than those without definitions, while no significant difference was found between the AI-generated and no-definition groups or between AI-generated and MED definitions. Additionally, students with access to any definitions took longer to complete the task than those without, but there was no significant time difference between the AI and MED groups. The findings suggest that although AI-generated definitions show promise, human-created dictionary definitions currently provide more effective support for receptive lexical tasks, highlighting the need for further research on AI's role in lexicography and language learning.