← Back to Volume 48, Issue 1 Contents
MODELING SEMANTIC DRIFT IN CHATGPT RESPONSES USING MARKOV CHAINS: AN EXCEL-BASED SIMULATION APPROACH
Ashwini Modi
Mathematics Department, SIWS College, Wadala, Mumbai.
Volume: 48Issue: 1Pages: 10–15Year: 2026ISSN: 0970-5120
⬇ Download Full PDF
Abstract
ChatGPT and other large language models generate text by guessing one word at a time based on what came before. While their responses often sound smooth and logical, they sometimes slowly drift off-topic or repeat themselves. In this study, we use a simple Markov chain model—built and tested in Excel—to track how the meaning of ChatGPT’s replies shifts over time. By simulating these “semantic states,” we uncover patterns in how the model stays on track, loops back, or wanders. Our results show that even basic probability tools can help us peek into the structure behind ChatGPT’s behavior, making it easier to fine-tune prompts and improve how we interact with these models.
Keywords: ChatGPT, Semantic Drift, Markov Chains, Excel Simulation, LanguageModels
MSC 2020: 60J20 (Primary) 60J35 (Secondary).
How to Cite
Ashwini Modi, MODELING SEMANTIC DRIFT IN CHATGPT RESPONSES USING MARKOV CHAINS: AN EXCEL-BASED SIMULATION APPROACH, Journal of the Indian Academy of Mathematics, 48(1) (2026), 10–15.
References
- Weatherstone, C., et al. (2025). Quantifying Latent Semantic Drift in Large LanguageModels Through Self-Referential Inference Chains. arXiv preprint
- Kobanov, N., et al. (2025). Context-Preserving Gradient Modulation for Large Language Models: A Novel Approach to Semantic Consistency in Long-Form Text Generation. arXiv:2502.03643.
- Boyer, K. E., et al. (2010). Modeling Dialogue Structure with Adjacency Pair Analysis and Hidden Markov Models. Proc. Intelligent Tutoring Systems.
- Shaarawy, H. Y. (2023). Cohesion and Coherence in Essays Generated by ChatGPT: A Comparative Analysis to University Students’ Writing. Badr University in Cairo.
- Bunt, H., and Petukhova, V. (2023). Semantic and Pragmatic Precision in Conversational AI Systems. Frontiers in Artificial Intelligence, 6, Article 896729
© Journal of the Indian Academy of Mathematics