A Digital Stylistic Exploration of AI-Generated Text: Prompt Engineering for Journalistic and Legal Genres in ChatGPT-4
DOI:
https://doi.org/10.25130/Lang.10.2.P1.13Keywords:
Keywords: ChatGPT-4, digital stylistics, journalistic genre, legal genre, prompt engineering, stylistic features.Abstract
This study seeks to investigate practical and efficient methods for the stylistic analysis of Generative Pre-trained Transformer-4 (ChatGPT-4) ’s responses across journalistic and legal genres. This research aims to determine how prompt design influences stylistic variation and evaluate the Artificial Intelligence’s (AI) capacity for self-analysis. Using an experimental model, three prompt types: minimal, tone-specified, and constraint-based, were designed to generate text for qualitative analysis of lexical, grammatical, and textual features. The findings suggest that minor adjustments to the prompts can produce significant differences in rhetorical tone, word choice, and syntactic complexity. Legal responses exhibited greater structural precision and formality than journalistic ones, which favored brevity and neutrality. However, ChatGPT-4’s self-generated stylistic analysis remained superficial, prioritizing tone over grammatical detail, and was less accurate than professional stylisticians' analysis, who offer more precise classification of stylistic features and more reliable categorization. It is concluded that prompt engineering is central to achieving genre-appropriate AI text while emphasizing the necessity for continued human expertise in rigorous stylistic interpretation. Accordingly, it is recommended that structured prompt-design frameworks be developed and that expert human oversight be formally integrated into automated stylistic interpretation and evaluation processes.
References
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., … Amodei, D. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877–1901.
Das, A., Sanyal, D., & Chakraborty, T. (2024). StyloAI: Distinguishing AI-generated content with stylometric analysis. arXiv. https://arxiv.org/abs/2405.10129
Hoover, D. L., Culpeper, J., & O’Halloran, K. (2023). Digital stylistics: Computational approaches to style in literature. Routledge.
Jurafsky, D., & Martin, J. H. (2023). Speech and language processing (3rd ed.). Pearson.
Leech, G., & Short, M. (2007). Style in fiction: A linguistic introduction to English fictional prose (2nd ed.). Pearson Education.
Miller, S. (2022). Stylistics in the age of artificial intelligence. Palgrave Macmillan.
Prasanna, S., & Karthik, T. R. (2024). A comparative analysis of AI-generated text and human academic writing. Frontiers in Education, 9, Article 1347421. https://doi.org/10.3389/feduc.2024.1347421
Reynolds, L., & McDonell, K. (2021). Prompt programming for large language models: Beyond the few-shot paradigm. arXiv:2102.07350.
Short, M. (1996). Exploring the language of poems, plays and prose. Routledge.
