Supporting decision-making in engineering design by analizing team discussions with language models

Abstract

This article explores the potential of Large Language Models (LLMs) like ChatGPT to automate the application of the Analytical Hierarchy Process (AHP) within engineering design. We hypothesize that the wealth of information embedded in decision-making discussions, despite their informal nature, can be leveraged to extract AHP parameters and generate data-driven decision support. We demonstrate this through two case studies: (1) selecting an electronic architecture for an underwater sensor array and (2) choosing a extraction technology for mining. The results show that LLMs can successfully identify and weight decision criteria from unstructured discussions, with primary criteria weights closely matching manual AHP implementations. While evaluation consistency varied, the automated approach maintained reasonable alignment with expert judgments in final alternative rankings. The LLM-based approach also demonstrated advantages in processing efficiency and ability to capture implicit criteria not formally considered in manual processes. These findings suggest that LLMs can effectively support AHP implementation in engineering decision-making, though their output should complement rather than replace expert judgment. This work contributes to the growing body of research on AI-supported decision-making in engineering design, offering insights into both the potential and limitations of automated multi-criteria decision analysis.