科学技術予測調査手法に関する数理的分析

デルファイ調査及びリアルタイム・デルファイ法に関するエージェントシミュレーション

抄録

The "Delphi" related methods fostered the development of multiple survey practices. However, their mathematical theoretical validity and prediction accuracy have not been investigated in depth, especially from the perspective of mathematical or qualitative method.

In this paper, we run virtual subject experiment with agent-based simulations to verify the validity of the information obtained from two methods: a Delphi survey for a repeated questionnaire and a real-time Delphi method aimed to present the response status when a systematic information feedback is implemented according to predefined conditions /protocols.

In the results, the Delphi method emerges as a way to consolidate opinions and make them converge. On the contrary, the degree of consolidation/convergence ensured by the real-time Delphi method varies greatly across multiple simulations, even within the same respondents' group. Moreover, this method cannot be validated theoretically with respect to the statistical group response, as the basic Delphi principle is not satisfied.