Abstract <p>A new approach to analyzing the development of scientific knowledge is proposed, based on the concept of information dynamics of evaluative theories. Scientific knowledge is considered as a dynamic system striving to minimize informational tension. A methodology has been developed that uses the average mutual information to aggregate evaluations considering the semantic load and internal consistency of knowledge. Using climate models and evaluations of the age of the universe as examples, the application of the proposed approach regarding the evolution of scientific theories is demonstrated.</p>