Laptievа T. Methods of detecting disinformation based on expert assessments

The dissertation is devoted to solving an urgent scientific task related to the development of a scientific and methodological apparatus for detecting disinformation and stopping its spread in the information space on the basis of expert assessments. The purpose of the dissertation is to increase the reliability of detecting disinformation and stopping its spread in the information space on the basis of expert assessments. The object of research is the process of detecting disinformation and stopping its spread in the information space. The subject of the study is models and methods of detecting disinformation in the information space based on expert assessments. The introduction substantiates the relevance of the scientific task related to the development of a scientific and methodological apparatus for detecting disinformation and stopping its spread in the information space based on expert assessments. The connection of the work with scientific programmes, plans and topics is presented, the scientific novelty is given, the practical significance of the results obtained is presented, information on the personal contribution of the applicant is provided, and a list of publications on the topic of the dissertation is presented.The first section examines the current scientific tasks and strategic directions related to the detection of disinformation in the information space. The author analyses the laws of existence of information properties and peculiarities of human perception of information to identify particularly important factors in addressing the issue of disinformation detection. Based on the analysis of the factors influencing the processes of information transmission, modern systems for detecting influences or threats in textual information are identified. In the second section, the cluster analysis method is improved in order to reduce the amount of contextual information for further more thorough analysis of information. The cluster analysis method was improved by integrating two clustering methods: the bee colony method and the genetic algorithm. The third section assesses the factors that significantly affect the spread of disinformation. The mathematical model for predicting the spread of disinformation was improved. The mathematical model for predicting the spread of disinformation has been further developed by expanding the factors influencing the spread of disinformation and introducing normalising coefficients into the probability state matrix for the model of stopping the spread of disinformation in the information space. The processes of stopping the spread of disinformation according to the developed model are modelled on the basis of the calculated transition matrix, taking into account the additional factor of influence on the spread of disinformation. Section 4 assesses the indicators of disinformation detection in the information space. It is proved that an effective method for identifying the veracity of information at the moment is the method of expert evaluation. The strength of the relationship between the indicators of disinformation detection is estimated. 1. For the first time, a method of detecting disinformation and stopping its spread in the information space based on expert opinions is proposed, which differs from the existing ones by using an improved clustering method with the subsequent application of the naive Bayes method for classifying information and a method of detecting disinformation based on expert opinions. 2. A mathematical model for predicting the spread of disinformation has been improved, based on the model of a finite automaton with a given final level of reliability for detecting false information messages with a known initial level of reliability and a set of permissible actions, which makes it possible to implement a multi-step verification of messages with a gradual increase in reliability indicators depending on the nature of the messages and the degree of influence on their content. 3. The clustering method for selecting the most informative groups (clusters) for disinformation recognition was further developed, combining evolutionary modelling and artificial intelligence methods for detecting disinformation; the integration of models leads to the creation of a hybrid agent that alternately performs the functions of adaptive behaviour of a bee colony and a genetic algorithm, which reduces the number of compatible clusters by 5% compared to existing clustering methods. The practical significance of the results obtained is as follows: - the scientific results proposed in the thesis can be used to identify disinformation and stop its spread; - the basic rules and recommendations for identifying disinformation and stopping its spread in the information space are proposed. Based on the results of the study, the practical and theoretical value of the developed methods was confirmed, practical recommendations were given on the application of the dev