Speaker
Description
This presentation explores the use of generative language models by students in tertiary education contexts and how the use of text classification might be applied to detect LLM-generated writing. Using essays from L1 Japanese learners of English and Llama 2, three classifiers (naive Bayes, fastText, and a Llama 2-based model) are compared for accuracy. Findings provide insights into mitigating AI use in academic writing while maintaining assessment validity.
Summary
This presentation explores the use of generative language models by students in tertiary education contexts and how the use of text classification might be applied to detect LLM-generated writing. Using essays from L1 Japanese learners of English and Llama 2, three classifiers—naive Bayes, fastText, and a Llama 2-based model—are compared for accuracy. Findings provide insights into mitigating AI use in academic writing while maintaining assessment validity.
| Teaching Context | College and university education |
|---|
