University of Twente Student Theses
As of Friday, 8 August 2025, the current Student Theses repository is no longer available for thesis uploads. A new Student Theses repository will be available starting Friday, 15 August 2025.
Analysis of the impact of readily available AI on Academic Text : Lexical Diversity & Syntactic Complexity
Delden, Felix van (2025) Analysis of the impact of readily available AI on Academic Text : Lexical Diversity & Syntactic Complexity.
PDF
676kB |
Abstract: | The widespread adoption of AI writing tools, such as ChatGPT, has raised questions about their influence on academic writing, particularly among non-native English (L2) speakers. This study examines how these tools have shaped lexical diversity and syntactic complexity in 3223 bachelor’s and master’s theses from a technical department at a Dutch university, written in english, spanning from 2018–2025. All theses from the department were analyzed, with three programs highlighted— including two technical programs (theses are completed in half a semester) and a creative technology program (theses are completed in one semester)—as these had the most published theses between 2018 and 2025. Theses from pre-LLM (2018–2022) and post-LLM (2023–2025) periods were contrasted using two lexical diversity metrics (vocd-D, MTLD) and three syntactic complexity metrics (mean sentence length, clauses per sentence, sentence length variation). Results show a significant rise in lexical diversity after 2022: vocd-D in- creases by 8.3 % and MTLD by 13.6 %, suggesting broader vocabulary use with AI support. In contrast, syntactic complexity remains stable within narrow bounds across all programs (clauses per sentence: 0.75–0.78; mean sentence length: 15.2–15.6 tokens; sentence length variation: 8–10 words). Program-level patterns persist, creative technology theses exhibit higher subordination and sentence length, possibly influenced by their longer writ- ing period and narrative style, while technical theses stay concise under tighter semester and format constraints. These findings highlight a nuanced AI impact: richer vocabulary with- out greater syntactic complexity. Limitations include a single-department, predominantly L2 sample and unknown student backgrounds. Future work should expand to L1 writers, other disciplines, controlled AI- vs. human- written corpora, and model-specific analyses to strengthen AI-content de- tection methods. |
Item Type: | Essay (Bachelor) |
Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
Subject: | 02 science and culture in general, 50 technical science in general, 54 computer science, 81 education, teaching |
Programme: | Business & IT BSc (56066) |
Link to this item: | https://purl.utwente.nl/essays/107398 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page