A COMPARATIVE ANALYSIS OF SEMANTIC AND PRAGMATIC FEATURES IN AI-GENERATEDAND HUMAN-WRITTEN TEXTS: COHERENCEAND MEANING CONSTRUCTION

Abstract

The proposed research will compare semantic and pragmatic aspects of AI-generated and human-written texts in terms of their coherence and meaning-making. It will be aimed at determining the lexical cohesion, sentence complexity, thematic consistency, tone modulation, politeness strategies, and figurative language of both types of text. The study background underscores the progressing nature of AI language models, such as GPT-3 and GPT-4, which are capable of producing semantically accurate text but may have difficulty with pragmatic features such as tone and adaptation to a particular context. The theoretical framework is informed by linguistic theory of semantics (Halliday and Hasan, 1976), pragmatics (Grice, 1975), and coherence (Schiffrin, 1987), which focuses on how human communication is better than AI in creating context-sensitive meaning. The study will utilize a comparative analysis methodology, which will involve the use of thematic analysis and descriptive statistics to analyse the data. The datasets used in data collection consist of AI-generated texts of GPT models and human-generated texts of publicly available materials of different genres. Purposive sampling is used as the sampling method as the relevant texts are selected on which the analysis will be conducted, and final 100 texts (50 AI-generated and 50 human-written) are selected. Results indicate that AI-generated texts are semantically sound but they lack semantic tone variation, figurative language and theme continuity. Written works by people show better discourse markers and pragmatic comprehension, have a better coherence and a communication contextually suitable.