Document Type : Exploratory
Authors
ATU
Abstract
This study aimed to design and validate an AI-based Delphi method using the imitation game (Turing Test). The main research question was how the Turing Test could serve as a novel approach to assess the validity of AI-based Delphi results. This applied–developmental study involved 30 participants, including HR specialists, AI experts, and independent evaluators. Secondary data were collected, and the outputs of three language models—Microsoft Copilot, ChatGPT, and Gemini—were used as inputs in the Turing Test to analyze the similarity between human and AI-generated Delphi results. These data were extracted from the authors’ previous studies, which had identified the personal development needs of HR managers using the Delphi method.
As the first study to apply the Turing Test for evaluating a research method, the findings demonstrated that AI-based Delphi can produce data that are consistent and reliable in comparison to human responses. The results indicate that large language models effectively simulate human behavior and response styles, making it difficult for participants to distinguish between human and machine-generated answers. These findings strengthen the validity of the AI-based Delphi method and suggest that it can serve as an effective research tool without compromising the quality of human-level analysis. without compromising the quality of human-level analysis
Keywords
Main Subjects
)