[Purpose/significance] To explore how algorithmic folk theories evolve from dispersed individual perceptions to collective consensus through multi-agent interactions, thereby providing insights into the formation process and evolutionary mechanisms of algorithmic folk theories. [Method/process] From the perspective of knowledge diffusion, this study employs an Agent-Based Modeling (ABM) simulation approach, conceptualizing information users as heterogeneous cognitive agents. By incorporating metrics such as Shannon entropy, structural entropy, and LMC complexity, it investigates the evolutionary process of algorithmic folk theories as well as the diffusion and distribution patterns of algorithmic knowledge. [Result/conclusion] The evolution of algorithmic folk theories undergoes four stages:emergence, development, stabilization, and decline. The diffusion network evolves from fragmentation to integration,while knowledge distribution exhibits three patterns: consensus, equilibrium, and polarization. Individual cognitive ability and foundational knowledge level are identified as key factors influencing the evolution of folk theories, and excessive disparities may lead to knowledge gaps and cognitive polarization.