RESEARCH ON EVALUATION METHODS FOR THE USE OF GENERATIVE AI FINDINGS IN TECHNICAL JUDGEMENTS FOR ROAD BRIDGE INSPECTIONS
Japanese Journal of JSCE · 2026 · 82 (8) · n/a
DOI: 10.2208/jscejj.25-00073 ↗Kurz gefasst
Fachartikel in der Zeitschrift Japanese Journal of JSCE, 2026. Verfasst von Takumi KOBAYASHI und 5 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben.
Abstract (Originalsprache)
The rapid evolution of artificial intelligence in recent years has gone beyond the scope of merely replacing humans, and is becoming a factor that is prompting the redefinition of the roles traditionally played by humans in various areas of society. In this study, we evaluated the reliability of LLM-generated findings from various perspectives, with a view to building a future system in which humans and generative AI collaborate in making technical decisions in road bridge inspections, where logical reasoning is important. The results of the evaluation showed that the commercially available model, for which no additional training was conducted, showed a lower degree of similarity in the words used than that of a skilled engineer, but a higher degree of similarity in the sentence. In addition, the similarity between the responses of skilled engineers and those of the generated AI was evaluated multilaterally using five indices (Bag of Words, Term Frequency-Inverse Document Frequency, Jaccard Coefficient, Network Graph, and Sentence-BERT) based on unstructured questions. The following example shows a series of validation methods to evaluate the influence of other biases and other factors by using semi-structured questions.
Zitieren
APA
KOBAYASHI, T., MORI, S., HATORI, T., NAKAHATA, K., OHSUMI, M., & DAIKOKUYA, N. (2026). RESEARCH ON EVALUATION METHODS FOR THE USE OF GENERATIVE AI FINDINGS IN TECHNICAL JUDGEMENTS FOR ROAD BRIDGE INSPECTIONS. Japanese Journal of JSCE, 82(8), n/a. https://doi.org/10.2208/jscejj.25-00073
Vancouver
KOBAYASHI T, MORI S, HATORI T, NAKAHATA K, OHSUMI M, DAIKOKUYA N. RESEARCH ON EVALUATION METHODS FOR THE USE OF GENERATIVE AI FINDINGS IN TECHNICAL JUDGEMENTS FOR ROAD BRIDGE INSPECTIONS. Japanese Journal of JSCE. 2026;82(8):n/a. doi:10.2208/jscejj.25-00073
BibTeX
@article{KOBAYASHI2026,
author = {KOBAYASHI, Takumi and MORI, Shinichiro and HATORI, Tsuyoshi and NAKAHATA, Kazuyuki and OHSUMI, Michio and DAIKOKUYA, Nobuhide},
title = {RESEARCH ON EVALUATION METHODS FOR THE USE OF GENERATIVE AI FINDINGS IN TECHNICAL JUDGEMENTS FOR ROAD BRIDGE INSPECTIONS},
journal = {Japanese Journal of JSCE},
year = {2026},
volume = {82},
number = {8},
pages = {n/a},
doi = {10.2208/jscejj.25-00073}
}
EUVIMED