The paper is a pioneer work in mapping the methodological problems, which have been introduced into sociology by ABM as an important innovation. It sheds light on the sources and the background of ABM, comparing the latter with EBM. It shows also the main characteristics of multi-agent models and their wide-scaled applicability. The author focuses especially on the systematization of basic methodological problems. He underlines the relationship between simulation and classical experiment, as well as the idealized character of multi-agent models and their effective applying in the development and testing of the sociological theory.