Several areas of multi-agent research, such as large-scale agent organization and experience-based decision making, demand novel perspectives and efficient approaches for mul-tiscale information analysis. A recent breakthrough in harmonic analysis is diffusion geometry and diffusion wavelets, which offers a general framework for multiscale analysis of massive data sets. In this paper, we introduce the "diffusion" concept into the MAS field, and investigate the impacts of using diffusion distance on the performance of solution synthesis in experience-based multi-agent decision making. In particular, we take a two-dimensional perspective to explore the use of diffusion distance and Euclidean distance in identifying 'similar' experiences-a key activity in the process of recognition-primed decision making. An experiment has been conducted on a data set including a large collection of battlefield decision experiences. It is shown that the performance of using diffusion distance can be significantly better than using Euclidean distance in the original experience space. This study allows us to generalize an anytime algorithm for multi-agent decision making, and it also opens the door to the application of diffusion geometry to multi-agent research involving massive data analysis.