The paper is devoted to the problems of identifying the types of intelligence (artificial or natural) in automated systems that are used in the education sphere. The proposed approaches are based on the use of models and methods of game theory. It is assumed that agents (players) in the studied game situations can be both people (owners of natural intellectual abilities) and software systems (owners of artificial intelligence). The term “separation” refers to procedures that result in the classification of players (agents) into classes of carriers of natural (anthropogenic) or artificial intelligence. Separation procedures are based on the hypothesis that different types of intelligences in game situations will correspond to stable characteristic models of strategic behavior. In the paper we propose separation algorithms based on different classes of games, both strategic and cooperative. In particular, on the simplest bimatrix games “Rock-Paper-Scissors “and “family dispute”. The advantages of algorithms based on games of the class “family dispute” is due to the multiplicity of equilibrium situations. Thus, there appear additional possibilities for revealing the patterns of behavior characteristic of different types of intelligence. We also consider separately algorithms based on cooperative game models, assuming separation based on the choice of stable choices of certain divisions. The methods and approaches investigated in the article are particularly relevant in the light of new problems and challenges faced by the sphere of education at the stage of global digital transformation.