Human-Based and Gen-AI Academic Writing: AComparative Linguistic Analysis of Cohesion and Coherence

Abstract

This research does a comparative linguistic analysis of cohesion and coherence in human generated and AI generated academic essays, using Halliday and Hassan's Cohesion theory and Charolles’ Coherence theory. With the rise of artificial intelligence, particularly big language model such as ChatGPT has intensified the interest in assessing the effectiveness of AI in academic writing. This research examines 100 essays including 50 written by humans and 50 by artificial intelligence (ChatGPT) with an emphasis on the use of cohesive devices, additive adversative, causal and temporal conjunctions with anaphoric and cataphoric references, substitution, ellipsis and lexical choices , collocation and reiteration. This investigation reveals that while AI generated essays outperform in temporal conjunctions (65.79%) and cataphoric references (74.03%). But human writing exhibit more proficiency in additive conjunctions (64.12%) and anaphoric references (60.63%). Humans related articles exhibit more superior coherence, showcasing the complexity and depth in argumentation. In contrast, AI generated essays are more systematic and logically organized framework however lack the nuance development of concepts of human writing. The findings highlight the strength and limitations of AI in replicating human-like academic writing, suggesting areas for further enhancement in AI-assisted educational tools instruments and writing technology.