Akademik Veri Yönetim Sistemi
5th International BRITISH Congress on Interdisciplinary Scientific Research & Practices, London, İngiltere, 5 - 07 Şubat 2026, cilt.1, sa.1, ss.14, (Özet Bildiri)
Artificial intelligence (AI) is broadly defined as systems that emulate humans’ complex thinking and problem-solving abilities. Large language models (LLMs) have driven major advances in AI. This study comparatively analyzes the capacity of two LLM-based natural language processing systems, ChatGPT and DeepSeek, to interpret test data for flexible pavement hot mix asphalt (HMA) materials in highway engineering and to assess their technical consistency. The evaluation draws on results from Marshall stability and Semi-Circular Bending (SCB) tests conducted on HMA specimens. Each model received the same data set, test methods, and explanatory information. Model responses were evaluated across five dimensions: (i) technical accuracy, (ii) compliance with engineering terminology, (iii) data interpretation competence, (iv) ability to generate applicable engineering recommendations, and (v) internal consistency. Reliability was assessed using the modified DISCERN (mDISCERN) scale, adapted from applications in the health sciences to engineering, and the Global Quality Scale (GQS). In this study, mDISCERN comprises five items scored from 1 to 5, with higher scores indicating greater reliability; GQS scores ranged from 1 (worst) to 5 (best). To mitigate bias, two independent transportation engineering academics performed the ratings; in cases of disagreement, a third independent expert adjudicated. Inter-rater agreement was reported using Cohen’s kappa (κ). Thus, the engineering performance of the models was compared objectively using both qualitative and quantitative approaches. The results aim to clarify the feasibility of employing AI-based language models as supportive digital decision-making tools for flexible pavement design and analysis. The study further contrasts the models’ command of technical terminology, integrity in data interpretation, and capacity for original inference. Overall, the work proposes a methodological framework for the use of AI in transportation engineering and intends to inform education, engineering practice, and digital decision-support systems while delineating the limits and potential application areas of AI in engineering.