Overview / Event description
AI has brought about new trends in producing written texts. While some positively highlight its role as a writing assistant, others are concerned about its influence on ethical practices. Some researchers underline the limitations of AI text generation in simulating natural human communication (Sardinha, 2024), a "writing style" gap between AI-generated and human-written text (Ma et al., 2023) and a lack of specificity, depth, and accurate source referencing (Amirjalili et al., 2024).
The presentation will share the study findings on the linguistic analysis of AI and human-written texts by foundation-year students in HEI. The learners produced a short report on a collaboratively conducted study within a semester. We surveyed students (n=302) to explore their experiences in academic writing and language skills that developed in collaboration. We then compared students' perceptions with their actual marks, i.e. 42 randomly selected AI (n=21) and student-generated (n=21) papers. For data comparison, the researchers used the Tool for the Automatic Analysis of Lexical Sophistication (TAALES) to measure the lexical frequency and range, word N-grams, academic vocabulary and semantic properties of student papers (Allen & McNamara, 2015).
Most students claimed collaborative writing helped generate ideas and improve academic writing skills and language, owing to the peer feedback on grammar and lexical errors. Findings indicate a generally positive alignment between students' self-reported responses and the lecturers' comments on their papers - consistent lexical and syntactic choices reflecting more personalisation, hedging, and appropriate genre-specific terminology. TAALES analysis showed that human-authored texts contained more commonly used words with higher lexical frequency values (8.65) than AI-generated texts (8.31). Its lexical range has also been found to vary marginally in diversity (67.96) compared to AI texts (65.27), which means the latter tends to be more formulaic and repetitive. However, regarding semantic similarity, AI-written content indicated greater internal coherence following a more structured lexical pattern. Finally, human writers have demonstrated more variety in ideas and word choice, while the content produced by AI has lower conceptual flexibility. Based on these findings, EAP teachers could consider using AI-generated content to incorporate corpus-based classroom activities for varying vocabulary, improving authenticity in genre-specific language and style, and as an initial organizational guide in academic writing.
Objectives
- Examine students’ perceptions of collaborative academic writing.
- Compare linguistic features of AI and human-written texts using TAALES.
- Assess alignment between student self-reports and lecturer feedback.
- Explore teaching uses for AI-generated content.
Facilitators
Iroda Saydazimova is a lecturer and Academic English module leader at Westminster International University in Tashkent, Uzbekistan. She mainly teaches EAP and Academic and Research Writing modules at different levels. Her current professional and research interests focus on testing and assessment, academic writing, and motivation in language acquisition.
Liliya Makovskaya is a senior lecturer in the Global Education Department of Westminster International University in Tashkent, Uzbekistan. She has experience in teacher training and material design in international projects. Her research interests lie in assessment, second language writing, feedback, academic vocabulary, discourse analysis, and higher education.
V1 respect individual learners and diverse groups of learners
V3 use scholarship, or research, or professional learning, or other evidence-informed approaches as a basis for effective practice
V5 collaborate with others to enhance practice
K3 critical evaluation as a basis for effective practice