This paper explores the efficacy of generative artificial intelligence (GenAI) for value-focused thinking, specifically its ability to generate high-quality sets of objectives for organizational and policy decisions. Overall, we find that while most GenAI-generated objectives are individually viable, the objective sets as a whole exhibit substantial shortcomings. They often include nonessential considerations, omit important objectives, and lack structure due to redundancy and poor decomposability. A key issue is the tendency of GenAI to include means objectives, even when explicitly instructed not to do so.

At the same time, we show that the quality of objective sets can be significantly improved by applying best practices in prompting and incorporating decision analysis (DA) expertise. The findings highlight the importance of a human-in-the-loop approach: GenAI is useful for generating initial objective ideas, but expert input from decision analysts is essential before using the results to support real-world decision making.

To operationalize this, we present and demonstrate a four-step approach that combines the complementary strengths of GenAI and decision analysts.

Simon, Jay; Siebert, Johannes U. ChatGPT vs. Experts: Can GenAI Develop High Quality Organizational and Policy Objectives? Decision Analysis (in press). https://doi.org/10.1287/deca.2025.0387