Speaker
Description
Generative artificial intelligence (GenAI) has expanded visual support for second-language vocabulary learning, but research focuses mainly on recall and retention. This study examined whether learner-generated AI imagery supports deeper lexical knowledge in meaning precision, collocational knowledge, contextual application, and productive use. Sixty-five first-year Vietnamese university students at approximately CEFR B1 learned 24 academic words in a counterbalanced within-subjects design. Half were studied through student-generated AI images (AI-IMG) and half through matched verbal semantic elaboration (VERB). Both conditions required learners to interpret each word and construct a meaningful situation; only AI-IMG added a ChatGPT-generated visual representation. Vocabulary knowledge was assessed before instruction, immediately afterward, and three weeks later. Separate analyses showed small-to-moderate AI-IMG advantages in meaning precision, contextual application, and productive use, but little difference in collocational knowledge. Learner-generated AI imagery therefore appears to support semantic and contextual dimensions of vocabulary depth selectively rather than improve all aspects of lexical knowledge equally.
Keywords: AI-generated images; vocabulary depth; generative AI; ChatGPT; EFL vocabulary learning; academic vocabulary
