LLM-Facilitated Persona-Based Delphi Method for Multi-Criteria Idea Evaluation

This study proposes a novel multi-perspective idea evaluation framework by combining the traditional Delphi method with AI-generated personas and Large Language Models (LLMs). Unlike conventional approaches that rely on predefined evaluation criteria, our method enables each persona to independently select and apply its own criteria based on unique attributes. Through iterative rounds facilitated by an LLM, evaluations converge while retaining diversity and minimizing bias. We demonstrate this approach through the assessment of multiple product ideas, highlighting how persona attributes such as age and gender influence evaluation perspectives. The results suggest this framework enhances both the efficiency and inclusivity of early-stage idea screening and consensus building.