The information is prepared on the basis of data from the information-analytical system RSF, informative part is represented in the author's edition. All rights belong to the authors, the use or reprinting of materials is permitted only with the prior consent of the authors.

Project titleDevelopment of an algorithm for identifying risk factors for the safety of social networks users based on an analysis of the content and psychological characteristics of its consumers

Keywordssocial networks, online communities, online aggression, online risks, big data, user data, digital footprint, Internet security, psychological characteristics, anxiety, aggressiveness, neuroticism, adolescents, youth

Annotation
The project aims to develop a model for identifying and predicting security risks in social networks based on the analysis of the content (content) of social networks, digital traces, behavioral and psychological data of users. The relevance of the project lies in the increasing gap between the level of involvement of children and adolescents in social networks and understanding of the mechanisms of consumption of online information and the impact of unsafe content on children and adolescents - 90% of social network users. Despite the existing policy of countering and blocking prohibited content on social networks, the number of communities and user accounts that distribute unsafe content is constantly growing. Social networks as the most important institution of socialization make children and adolescents particularly vulnerable to destructive influence due to the as yet undeveloped mechanisms of opposition to negative information and passive, uncritical consumption of information. The scientific novelty of this project is to develop an integrated model of risk factors in social networks based on the identification algorithm for unsafe social network content in conjunction with the psychological characteristics of its consumers. Within this project, for the first time in Russia (in a sample of up to 10,000 users) methods and tools will be created that allow to identify and classify: - forms of unsafe social networks content in the Russian Internet segment; - communities, accounts of users producing and distributing unsafe content; - psychological, individual-typological and pathocharacterological features of consumers of unsafe social networks content among children, adolescents and young people (in conjunction with the patterns of production and consumption of social networks content); - risk groups of users vulnerable to exposure to unsafe content. For the first time will be developed: - an algorithm for identifying users at risk for the consumption and distribution of unsafe content using machine learning methods and analyzing a large array of open user data from social networks; - a tool for monitoring social networks based on digital footprints, behavioral, psychological data of users with the identification of the main safety risk factors and the identification of users belonging to risk groups; - a web application for self-diagnosis of security risks in social networks; - recommendations for working with risk groups among children, adolescents and young people based on current data on information trends in social networks.

Expected results
As a result of the project, results will be obtained that are relevant to social, cognitive and computer sciences, and are comparable to the world level of research, both in terms of the proposed approaches and methods, and in the coverage of respondents. Thanks to the project, new actual data will be obtained on the main risk factors that users have to face in social networks and the connection of these factors with the content of social networks and psychological characteristics of users. As a result of the project, the following results will be obtained: Comprehensive model of identification and prediction of risk factors for security in social networks, based on the analysis of social network content and individual psychological characteristics of users. Classification of forms of unsafe social network content, security threat markers. Psychological, individually-typological and pathocharacterological features of consumers of unsafe content of social networks among children, adolescents and young people. Features of unsafe behavior in social networks for high school students and university students. The algorithm for identifying risk factors for the safety of users of social networks, based on the analysis of content and psychological characteristics of its consumers. The algorithm for identifying users at risk-groups for the distribution and consumption of unsafe content. A web application for monitoring social security risk factors based on digital traces, behavioral, psychological data of users with the identification of major security risks and identification of users at risk. A web application for psychodiagnostics of psychological, individual-typological and pathocharacterological features of users of social networks. Web application for self-diagnosis of security risks in social networks (including for self-diagnosis of psychological characteristics and self-regulation of “unsafe” states). Guidelines for monitoring security risk factors (student, student group, institution, region) - for psychologists and teachers of educational institutions. Database of psychodiagnostic studies of psychological, individual-typological and pathocharacterological features of users of social networks. Database of social network characteristics of users of social networks. 12 publications, including 5 - in journals indexed by the Web of Science or Scopus. Reports and presentations at international and all-Russian conferences. The results of the project are also published publicly on the website of the Foundation for internet development (as agreed with the management of the foundation). There are ample opportunities for practical use of the results, primarily in the social sphere and economy. The project aims to promote the development of the Internet and social networks as a safe space for users, especially children and adolescents. The social significance of the results, first of all, is connected with the need to support projects aimed at: - protection of users (especially vulnerable categories - children, adolescents, youth) from the impact of unsafe information on the Internet and social networks; - promoting the development of computer literacy in the field of digital security of wide sections of the population (children, adolescents and their parents, young people, professionals working with these categories of the population, etc.); - introduction to the practice of development, promising for use in telecommunication networks. As a result of the project, it is also planned to develop guidelines for parents on the safety of children in social networks. The results of the project can be implemented in the educational process of institutions of secondary and higher education of the Tomsk region in the form of the course "Security in social networks / Internet", including in the form of an online course (in this case for students from all over Russia). The results of the project can also be the basis for the development of programs for additional education, advanced training of psychologists, teachers, sociologists, and social workers. Project results can also be applied by managers and experts of education authorities, educational institutions, media education specialists, media representatives, a wide range of professionals involved in the development of the Internet and its impact on the digital generation, as well as anyone interested in the psychology of the Internet and social networks.

Annotation of the results obtained in 2021
The solution of problems was carried out at three levels of research organization: 1. Horizontal level - collected, structured and analyzed data of deviant content of a social network, individual psychological characteristics of teenagers and young people who are users of the social network VKontakte. 2. Vertical level - studied the relationship between the individual psychological characteristics of users and the fact and specifics of deviant content consumption in the social network (preference for different deviant content types). 3. Cross-level - predicative models were built, tested and validated based on a data set: the severity level of certain individual psychological characteristics; profile data (including likes); the fact and specifics of deviant content consumption on data arrays of 5254 users. Each of levels of the study organization and implementation was based on its own methodology: - Horizontal level: algorithms and methods for searching and collecting data from the social network VKontakte vk.com (API), network analysis, graph method, Phyton, Gephy tools; standardized methods of psychometric and screening studies. - Vertical level: descriptive and frequency statistics, Fisher's test, etc. - Cross-level: computer modeling based on a data set. The main results obtained during the third period of the project: 1. The improved model for predicting risk factors of psychological security based on user data from the social network VKontakte by increasing empirical data from 2420 to 4025 respondents (the Short Dark Triad, the Beck Anxiety Inventory, the Beck Depression Inventory, Self-harm Behavior Scale by N.A. Polskaya, Social networks using questionnaire by V.V. Matsuta). The quality of predicting the risk factors "psychopathy" and "problematic using of social networks" was increased by 2 times (F-measure, %, from 36 to 72 and from 42 to 83). The quality of predicting "anxiety", "depression", self-harm strategies "regain control of emotions", "influencing others", " relief from stress" was increased by 15-25% (F-measure, %, 41 and 63, 41 and 61 , 42 and 62, 30 and 46, 42 and 58, respectively). "Machiavellianism" and "narcissism", which did not receive a prediction at the previous stage, received quality metrics (F-measure, %, 52 and 62). 22-46% accuracy and 45-70% recall were at the previous stage, 47-72% accuracy and 47-100% recall become after improving the predicative model. For the risk factor “changing oneself, searching for new experience”, the model could not be improved: F-measure, %, 35 and 37, accuracy and recall - 26 and 34, 54 and 40. A possible reason for this is the absence of any manifestation of this risk factor in the respondents' digital footprint. The data set is unique for Russian and international researches in this interdisciplinary area. 2. Developed and tested the predicative model for risk factors of psychological security according to user data of the social network VKontakte (taking into account user reactions to content). The model was applied on a sample of 5254 respondents’ relevant data (2426 students and 2828 high school students). In the students’ sample the value of F-measure for most risk factors is 63-71%, completeness is 81-97%, accuracy is 48-58%. The exception was the strategy "change yourself, search for new experience" with the value of the F-measure - 24%, recall - 33% and accuracy - 19%. In the sample of high school students the value of the F-measure for all risk factors is 55-75%, recall - 58-100%, accuracy - 52-75%. In the pooled sample the value of F-measure for most risk factors is 63-71%, recall - 81-97%, accuracy - 48-58%. 3. The updated and enlarged classifier of unsafe content types in the social network VKontakte including 4147 unique deviant communities on 15 categories. In 2019 the classifier included 471 communities, in 2020 - 2251, in 2021 – 2843. 4. Results of content analysis of identified communities with insecure content, which demonstrate: 1) High heterogeneity of the communities content "hate and aggression", "violent content", "depressive content", "suicidal content", "self-harm", "psychoactive substances". Publications posted in the community of one thematic category refer to communities of other thematic categories. 2) Different (not always high) proportion value of the content in communities the publications with insecure content. Posts with neutral content are often found in communities with insecure content. The ratio of neutral and unsafe posts differs depending on the specific category. 3) The need to improve the assessing accuracy of the severity of user interest in a particular category of insecure content by: - ranking communities with unsafe content by the proportion of unsafe content in them, and assigning a degree of "toxicity" to each community depending on the size of this proportion; - taking into account when classifying content not only the verbal, but also the non-verbal form of information presentation (images, video, audio), because for some categories this form carries the main semantic load (violent content, psychoactive substances, self-harm); - including in the analysis not only user subscriptions to communities with unsafe content, but also user reactions (likes, comments, reposts) to individual publications with unsafe content. 6. Description of the established relations (р<0.05) of the individual psychological characteristics with the unsafe content consumption of deviant communities in the social network VKontakte based on a data set of 5254 people: 2426 students and 2828 high school students. Psychopathy – depressive content, hate and aggression, psychoactive substances, self-harm. Machiavellianism – hate and aggression, psychoactive substances. Narcissism – depressive, suicidal content, hate and aggression, psychoactive substances, self-harm. Anxiety, depression, problematic using of social networks – depressive, suicidal, violent content, psychoactive substances, self-harm. Influencing others, regain control of emotions, relief from stress – depressive, suicidal, violent content, hate and aggression, psychoactive substances, self-harm. Change yourself; search for new experience has similar relations, excluding suicidal content. 7. Description of relations between individual psychological characteristics and the digital footprint data. Sex characteristics, profile status, number of subscribers, subscriptions to communities with unsafe content, and user reactions to certain clusters (coronavirus pandemic, love, dating and relationships, success, etc.) are the most significant having a statistically greater weight) digital footprint data for predicting anxiety, depression, types and strategies of self-harm, problematic using of social networks, machiavellianism, narcissism, psychopathy. 8. Complex data-driven for identifying and predicting risk factors for the users’ psychological safety in social networks. It was developed using data mining methods and algorithms of a large-scale study on data sets of more than 20 thousand users containing more than 1 million measurements. The model has 4 components: a) individual psychological characteristics; b) the pattern of information consumption in the social network; c) the specifics of the content consumed in the social network; d) characteristics of the digital footprint (including features of online representation, online behavior and temporal context). Simplified model form is the Intraindividual risks + Interindividual risks + Information risks / Digital footprint = Risk factors for the users’ psychological safety in social networks. Intraindividual risks (according to the factor weight and the F-measure) are a high level of anxiety, depression, neuroticism, psychopathy, the self-harm presence (all p<0.05). The vulnerability risk to online threats increases in direct proportion to the increase in the severity of the above signs of psychological distress and/or sharpening of personal characteristics (all p<0.05). Interindividual, including communication and behavioral risks are the negative connotation of profile status; posting, reposting, “likes” and other online reactions (“emoji”) to content (all p<0.05). Informational risks are a) the subscriptions to groups/communities and/or network activity in groups/communities classified in the categories of violent, suicidal, depressive content, hate and aggression, psychoactive substances use, self-harm (using the developed and updated Unsafe Content Classifier); b) the online presence of "friends" or subscriptions to personalities demonstrating posting, reposting, "likes" and other online reactions to content in the specified categories. Subscribing to communities from 2 or more unsafe content categories or subscribing to 2 or more communities significantly increases the risk of the social media users’ psychological security (all p<0.05). The digital footprint risks established as part of the study of the relationship between individual psychological, behavioral and network characteristics and the predictive models construction are the number of "friends" online; the number of profile followers; nighttime (corresponding to the time zone) of online activity (according to the weight factor and F-measure). The overall level of psychological security of a particular user is determined based on a set of specified data, taking into account the fact that the presence of risk signs from more than 2 categories increases the user's vulnerability to online threats. 9. Patents: Matsuta V.V., Goiko V.L., Feshchenko A.V. Database with the results of psychological diagnostics and user’s social network VKontakte data of university youth. Matsuta V.V., Goiko V.L., Feshchenko A.V. Database with the results of psychological diagnostics and user’s social network VKontakte data of school students. 10. Databases of individual psychological (anxiety, depression, types and strategies of self-harm, problematic using of social networks, machiavellianism, narcissism, psychopathy), and social network characteristics (sex, age, education, marital status, place of residence; profile data, subscriptions, posts, reactions to content) of 2426 students and 2828 school students. 11. Consolidated, processed and prepared for analysis dataset (results of psychodiagnostics, profile data, subscriptions, posts, user's network of connections, reactions to content) based on which it is possible to build new models and test hypotheses. It is a data set of 4025 students (updated by increasing the number of respondents) and 2828 school students by the Short Dark Triad, the Beck Anxiety Inventory, the Beck Depression Inventory, the Self-harm Behavior Scale by N.A. Polskaya, the Social networks using questionnaire. The dataset is unique because VKontakte is the only social network that allows the use of open user data. Other social networks (Facebook, Instagram, Messenger, Twitter) closed this opportunity for researchers a few years ago. They are blocked now in accordance with the requirements of Russian law. However, the dataset and how it can be used to build predicative models can be extended to other social networks. 12. Web application for self-diagnosis of psychological characteristics and self-regulation of unsafe emotional states and behavioral manifestations. The application include psychodiagnostics the block with the formation of a report and recommendations for the respondent; bot (digital assistant); the block of self-regulation and correction methods and techniques (with a tracker) by sections: anxiety, depression, stress, panic attacks, self-harm; block of results evaluation. It was developed based on the psychological research results, the practices of cognitive behavioral therapy, mindfulness, relaxation and art therapy https://t.me/patronus_tsu_bot?start=start_FPJMei65 13. Publications in scientific journals: 4 publications, including 4 publications in editions indexed by Web of Science or Scopus (including 3 Q1), 1 publication in edition included in the Russian Science Citation Index. 14. Publications in electronic media (10) and an updated website about the project: http://frpss.tilda.ws

Publications

1. Chudinov S., Serbina G., Matsuta V., Peshkovskaya A. Social media that kills: a network study on virtual groups related to school shooting and suicide Digital Communications and Networks, - (year - 2022)

2. Evseev V.I., Peshkovskaya A.G. No Link Exists between Physical Aggression and Self-Injuries in Young Men Who Self-Harm European Neuropsychopharmacology, Volume 53, Supplement 1, Pages S146-S147 (year - 2021) https://doi.org/10.1016/J.EURONEURO.2021.10.193

3. - База данных результатов психологической диагностики и пользовательских данных социальной сети «Вконтакте» вузовской молодежи -, 2021621266, 15.06.2021 (year - )

4. - Ученому ТГУ дали европейскую премию за работу по профилактике суицида Сайт Томского государственного университета, Ученому ТГУ дали европейскую премию за работу по профилактике суицида // Томский государственный университет, 26 октября 2021 (year - )

5. - Цифровой след расскажет о психологическом портрете студента // Томский государственный университет Сайт Томского государственного университета, Цифровой след расскажет о психологическом портрете студента // Томский государственный университет, 22 июля 2021 (year - )

6. - Ученые ТГУ выявили у молодежи рост интереса к опасному контенту в VK Сайт Томского государственного университета, Ученые ТГУ выявили у молодежи рост интереса к опасному контенту в VK // Томский государственный университет, 17 мая 2021 (year - )

7. - В Сибири интерес молодежи к опасному контенту вырос вдвое Аиф-Томск, В Сибири интерес молодежи к опасному контенту вырос вдвое // Аиф-Томск, 18 мая 2021 (year - )

8. - Томские ученые выявили у молодежи рост интереса к опасному контенту в VK РИА "Сибирь", Томские ученые выявили у молодежи рост интереса к опасному контенту в VK // РИА "Сибирь", 18 мая 2021 (year - )

9. - В Томске создан метод получения психологического портрета по «цифровому следу» Мультимедийный портал «ПОИСК», В Томске создан метод получения психологического портрета по «цифровому следу» // Мультимедийный портал «ПОИСК», 22 июля 2021 (year - )

10. - Томские ученые научились создавать психологический портрет по «цифровому следу» Томский Обзор, Томские ученые научились создавать психологический портрет по «цифровому следу» // Томский Обзор, 22 июля 2021 (year - )

11. - В ТГУ разработали программу, составляющую психологический портрет студента Агентство новостей ТВ2, В ТГУ разработали программу, составляющую психологический портрет студента // Агентство новостей ТВ2, 22 июля 2021 (year - )

12. - Программу, составляющую психологический портрет студента, разработали в Томском госуниверситете Интерфакс России, Программу, составляющую психологический портрет студента, разработали в Томском госуниверситете // Интерфакс России, 22 июля 2021 (year - )

13. - Томские ученые заявили, что интерес молодежи к небезопасному контенту в VK вырос в 2 раза Портал Новости vtomske.ru, Томские ученые заявили, что интерес молодежи к небезопасному контенту в VK вырос в 2 раза // Портал Новости vtomske.ru, 18 мая 2021 (year - )

Annotation of the results obtained in 2019
The goal of the project is to develop the methodology and algorithms to analyze harmful and dangerous content in the Russian-speaking segment of social networks, which are used a lot by children, teenagers and young people, as well as to identify individual and psychological characteristics of the social media users predetermining their vulnerability in terms of such dangerous content. Our goal for year 1 was to verify the assumption on the correlation between the information security risk factors and consumption of information in social media. Our tasks for year 1 were: - to develop the methodology and algorithms to analyze harmful and dangerous content in social networks; - to develop the classifier of dangerous content types in social networks; - to develop the algorithm to identify individual and psychological characteristics of the social media users predetermining their vulnerability in terms of such dangerous content; - to develop software to identify users from risk groups in terms of distribution and consumption of dangerous content by analyzing their subscriptions to the corresponding communities. To achieve our goal and tasks during year 1 of our project we arranged and performed studies in 3 areas: - a psychodiagnostic study of 10735 respondents by means of a specialized online platform to identify their psychological, personal-typological characteristics; - a study of a digital track of 70555 respondents with tools and algorithms of big data analysis of Vkontakte (Russian social network); - a study of characteristics of dangerous content posted by deviant communities in Vkontakte. To study psychological, personal-typological characteristics characteristics of respondents we applied psychodiagnostic methods: ‘Big Five Questionnaire’, ‘Scale of depression, anxiety and stress’ by A. Buss and M. Perry, ‘Scale of anxiety and depression’ by A.T. Beck, ‘Internet-addiction test’ by K. Young, ‘Scale of self-harming behaviour’ by N.A. Polskaya, a questionnaire ‘Body modifications and self-harm’ by N.A. Polskaya and A.S. Kabanova. For data processing we applied methods of descriptive statistics and Fisher’s exact test. To study the digital footprint of respondents we applied: a Python3-based algorithm to automatically collect API-based data from VKontakte. To upload the users’ profiles we applied users.get, to receive identification numbers of communities a user is subscribed to we applied getSubscriptions, while to upload the content from a user’s wall we applied wall.get. We also applied the algorithm to collect information on the time users spend in Vkontakte and the algorithm to identify sentiments of posts (Sentiment Analysis at the linguistic analysis platform PolyAnalyst). To study the dangerous content on Vkontakte we applied the method of the linguistic vocabulary construction (Natural Language Toolkit). For our search we used linguistic markets and Kribrum, the social media monitoring software. To identify users from risk groups in terms of generation, spreading and consumption of dangerous content we applied the ‘snowball’ method. Our study confirmed the assumption on the correlation between the information security risk factors and consumption of information in VKontakte. We identified correlations between users’ psychological characteristic of various levels (anxiety, aggression, depression, extraversion, friendliness, consciousness, openness to experience, and neurotism) with subscriptions to communities with dangerous content and time spent in this social network. To clarify nature and characteristics of the identified correlations we are planning further investigations. Our concrete findings after year 1 of working on the project are: 1. We created the classifier of types of dangerous content in Vkontakte including 3050 deviant communities from 13 categories: those promoting alcohol and drugs abuse, smoking, etc.; depression, images of cruel murders, blood, dead bodies; aesthetics of evil, nationalism, hatred, dangerous manipulations with one’s body, weapons, vulgarity, school shooting, aesthetics of death, suicide; true crime. 2. Wу crated an application to analyze a user’s subscriptions in Vkontakte to identify his\her interest to dangerous contents, which is superior in terms of scope and accuracy as compared to other available systems for monitoring the behavior of children in social media: https://lk.opendata.university/methods/raw/37. 3. We created an application for diagnostics of psychological characteristics of VK users and for collection of open data from its users’ accounts: http://ivik.org. Wе recommend these platforms for use at educational and other institutions working with teenagers and young people. 4. Wе compiled databases of the VK users: - data on psychological, personal-typological characteristics and digital footprint of school and college\university students from Tomsk, Tyumen, Voronezh, Nizhny Novgorod, Moscow, Sevastopol (11396 students); - social-demographic data and data on the digital footprint of school and college\university students from Tomsk and Barnaul (59820 students). 5. We described personal-typological characteristics of VK users of dangerous content. 6. Wе wrote and published papers in following journals: 4 in journals indexed by Web of Science or Scopus, 1 in a journal from the Russian Science Citation Index. On top of that, after year 1 we have publications in online media focused on the project (12 references) and in the project website: frpss.tilda.ws

Publications

1. Matsuta V., Mundrievskaya Y., Serbina G., Peshkovskaya A., Goiko V., Feshchenko A. School shooting communities on social networks: identification mechanisms and network structure Annual Review of Cybertherapy and Telemedicine, - (year - 2020)

2. Matsuta V.V., Mundrievskaya Y.O., Serbina G.N., Mishchenko E.S. Анализ текстового контента девиантных онлайн-сообществ (на примере сообществ скулшутинга) Гуманитарный научный вестник, №3, с. 90-101 (year - 2020) https://doi.org/10.5281/zenodo.3763848

3. Matsuta V.V., Mundriyevskaya J.O., Serbina G.N., Peshkovskaya A.G. Identification Strategy of Deviant Communities in Social Media (as Exemplified by School Shooting) Social and Behavioral Sciences, - (year - 2020)

4. Peshkovskaya A., Mundrievskaya Y., Serbina G., Matsuta V., Goiko V., Feshchenko A. Followers of school shooting online communities in Russia: age, gender, anonymity and regulations Advances in Intelligent Systems and Computing, - (year - 2020)

5. Feshchenko A., Mundrievskaya Y., Peshkovskaya A., Goiko V. Psychological safety of students in social networks: the search for dangerous content and identifying its consumers Proceedings of, - (year - 2020)

6. - Приложение ТГУ позволит вычислять в соцсетях агрессивных подростков РИА Томск, Приложение ТГУ позволит вычислять в соцсетях агрессивных подростков // РИА Томск, 18 февраля 2020 г. (year - )

7. - Специальный алгоритм отыщет во «ВКонтакте» школьников с повышенной тревожностью Агентство новостей ТВ2, Специальный алгоритм отыщет во «ВКонтакте» школьников с повышенной тревожностью // Агентство новостей ТВ2, 13 февраля 2020 г. (year - )

8. - В Томске разрабатывают алгоритм выявления агрессии и тревожности в соцсетях Информационное агентство "Красная весна", В Томске разрабатывают алгоритм выявления агрессии и тревожности в соцсетях // Информационное агентство "Красная весна", 13 февраля 2020 г. (year - )

9. - В России разрабатывают метод выявления в соцсетях школьников, склонных к агрессии Медийно-сервисный интернет-портал "Рамблер", В России разрабатывают метод выявления в соцсетях школьников, склонных к агрессии // Медийно-сервисный интернет-портал "Рамблер", 13 февраля 2020 г. (year - )

10. - В России с помощью специального алгоритма найдут подростков, склонных к агрессии и тревожности Независимое педагогическое издание "Учительская газета", В России с помощью специального алгоритма найдут подростков, склонных к агрессии и тревожности // Независимое педагогическое издание "Учительская газета", 15 февраля 2020 г. (year - )

11. - В России разрабатывают метод выявления в соцсетях школьников, склонных к агрессии Интернет-издание Letidor, В России разрабатывают метод выявления в соцсетях школьников, склонных к агрессии // Интернет-издание Letidor, 13 февраля 2020 г. (year - )

12. - Ученые создают алгоритм поиска школьников с повышенной тревожностью Научно-популярный портал Naked Science, Ученые создают алгоритм поиска школьников с повышенной тревожностью // Научно-популярный портал Naked Science, 12 февраля 2020 г. (year - )

13. - В России создают алгоритм поиска детей с повышенной тревожностью Российское агентство международной информации "РИА Новости", В России создают алгоритм поиска детей с повышенной тревожностью // Российское агентство международной информации "РИА Новости", 13 февраля 2020 г. (year - )

14. - "ВКОНТАКТЕ" НАЙДУТ ШКОЛЬНИКОВ С ПОВЫШЕННОЙ ТРЕВОЖНОСТЬЮ ВТРК Радио "Маяк", "ВКОНТАКТЕ" НАЙДУТ ШКОЛЬНИКОВ С ПОВЫШЕННОЙ ТРЕВОЖНОСТЬЮ // ВТРК Радио "Маяк", 13 февраля 2020 г. (year - )

15. - Во «ВКонтакте» будут искать агрессивных детей с помощью алгоритма Источник актуальных новостей Recipe.Ru, Во «ВКонтакте» будут искать агрессивных детей с помощью алгоритма // Источник актуальных новостей Recipe.Ru, 13 февраля 2020 г. (year - )

16. - Учёные создают алгоритм поиска школьников с повышенной тревожностью Сайт Томского государственного университета, Учёные создают алгоритм поиска школьников с повышенной тревожностью // Сайт ТГУ, 12 февраля 2020 г. (year - )

17. - Во «ВКонтакте» будут искать агрессивных детей с помощью алгоритма Сайт "Софтодром", Во «ВКонтакте» будут искать агрессивных детей с помощью алгоритма // Сайт "Софтодром", 13 февраля 2020 г. (year - )

Annotation of the results obtained in 2020
The goal of the project is to develop the methodology and algorithms to analyze harmful and dangerous content in the Russian-speaking segment of social networks, which are used a lot by children, teenagers and young people. The project includes: - the study of consumption and production patterns of different unsafe types of content, incl. within the areas of information epidemiology and network communications; - the study of the individual psychological foundations of the user’s predisposition to consume and distribute different unsafe (deviant) types of content; - the study and identification of the most vulnerable groups (by demographic, social, informational and individual psychological parameters), incl. for the purpose of ensuring safety and the implementation of psychological and informational prevention. The solution of problems was carried out at three levels of research organization: 1. Horizontal level - collected, structured and analyzed "per se" data: - deviant content of a social network, its classes and types, ways of existence and reproduction in the social network "VKontakte"; - individual psychological (personal, characterological, emotional) characteristics of teenagers and young people who are users of the social network “VKontakte”. 2. Vertical level - studied the relationship between the individual psychological characteristics of users and: - the fact of deviant content consumption in the social network; - the volume of deviant content consumption in the social network; - the interest specifics in deviant content in the social network (preference for different deviant content types). 3. Cross-level - predicative models were built, tested and validated based on a data set: individual psychological data and the severity level of certain individual psychological characteristics; profile data; the fact and volume of deviant content consumption on data arrays of more than 10,000 users and more than 1 million measurements. Each of levels of the study organization and implementation was based on its own methodology: - Horizontal level: algorithms and methods for searching and collecting data from the social network “VKontakte” vk.com (API), network analysis, graph method, Phyton, Gephy tools; standardized methods of psychometric and screening studies. - Vertical level: descriptive and frequency statistics, Fisher's test, contingency matrices (gamma coefficient), etc. - Cross-level: computer modeling based on a data set. The main results obtained during the second period of the project: 1. New algorithm for detecting deviant communities of the social network “VKontakte” with unsafe content (based on the use of a statistical significance test). 2. The updated classifier of unsafe content types in the social network "VKontakte" including 2849 unique deviant communities on 14 categories. 3. Adjusted methodology for primary identification of deviant communities (by updating the classifier of unsafe content using search methods for linguistic markers, snowball, an algorithm for identifying marker communities based on a statistical significance test, manual marking of communities by categories of unsafe content). 4. Description of the structure and communication strategies of deviant communities in the categories of “hate”, “psychoactive substances using”, “depressive content”, “suicidal content”, “self-harming behavior”. 5. Developed and tested model for the destructive content dissemination in the social network “VKontakte”. 6. Web application "Dangerous Content Detector" for monitoring security risk factors among students based on user data from the social network "VKontakte", hosted on the platform of the University Consortium of Big Data Researchers https://survey.opendata.university. 7. Description of the Russian specifics of deviant online communities (isolation, sparseness, low participants cohesion, lack of communication among them for the help and support) and their main function for participants (isolated personal consumption of unsafe content). Description of the community category (up to 96%) with mixed use of entertaining, humorous orientation, but hidden propaganda of psychoactive substances using, self-injurious behavior, their romanticization, presentation as norms and effective resolution of difficult life problems. 8. Description of the severity specificity of the interest in unsafe content: the predominance of interest among women (62% versus 58% among students, 71% and 56% among senior pupils, respectively), and senior pupils (85 against 56% among students). Description of unsafe content prevailing in the information consumption structure of students and senior pupils: depressive (17.8% and 27.2%), various forms of aggressive content (nationalism: 8.5% and 4.7%, hate: 8.1% and 6.7%, violence: 7.9% and 9.6%, true crime subculture: 4.6% and 6.7%), content about self-harm (6% and 5.9%). Description of the regional severity specificity of the interest in unsafe content: among the senior pupils of the Republic of Tyva, in comparison with other regions of the Siberian Federal District, there are fewer people with a high level of interest in unsafe content (1.5-2 times less). In terms of this interest intensity (the average number of subscriptions per person), women prevail over men (1.6 and 1.2) and over women in other regions (violence, hate, aesthetics of evil and death). In the Altai Republic, in comparison with other regions, women are distinguished by a higher level of interest in vulgar, depressive, suicidal content and content about self-harming behavior, men - in content about psychoactive substances, weapons and the true crime subculture. More than 50% of senior pupils living in the Siberian Federal District have communities with unsafe content in their subscriptions, and there are more such subscribers among women than among men. Male pupils are more interested in content about nationalism and weapons, while female pupils are more interested in depressive and suicidal content. 9. Correspondence of trends found in the social networks analysis to the existing in society phenomena: a high level of aggressiveness of Russian men in the online environment (63% of boys and men versus 31% of girls and women); a greater tendency to depression, self-harm, suicidal mood among girls and women (64% versus 36%; 82% versus 18%; 59% versus 41%). 10. A web application for diagnostics of individual psychological characteristics of social network users, implemented in LMS MOODLE and on the platform of the University Consortium of Big Data Researchers. 11. Description of individual psychological characteristics (psychopathy, narcissism, machiavellianism, anxiety, depression, instrumental and somatic self-harm, strategies of self- and interpersonal control, problematic using of social networks) of 8849 consumers of unsafe content (7505 students and 1344 senior pupils). 12. Description of relations between individual psychological characteristics (depression, anxiety, stress, non-suicidal self-harm, etc.) and the most significant for predicting them (having a statistically greater weight) digital footprint data (profile data; subscriptions to communities with unsafe content; the number of friends, subscribers; communities to which the user is subscribed; number of photos and albums; age and sex characteristics; relationship status; status in the social network and the direction of study at the university). 13. Description of the accuracy and completeness required to predict individual psychological indicators based on the digital footprint: precision - from 50% and more, and recall - from 70% and more. 14. Identification of the minimum sample size required to predict individual psychological characteristics based on a digital footprint with the required values of precision and recall: at least 3500-4000 respondents, taking into account the data dropout during preprocessing - 7000-10,000 respondents. 15. An improved predicting model of psychological safety risk factors based on the users digital footprint in the social network "VKontakte": on a sample of 3666 respondents, the precision of predicting extraversion, kindness, openness to experience increased by 1-3%, recall - by 3-8%. The precision of predicting depression, anxiety, stress, emotional stability decreased by 1-3%, while recall increased by 5-18%. 16. Developed and tested model for predicting psychological safety risk factors based on the users digital footprint: in a sample of 3637 respondents, low precision and recall were obtained (less than 50% for psychopathy, anxiety, depression, strategies of self-control and interpersonal control, problematic social networks using, etc.). 17. Description of the revealed pattern among online community members who have unsafe content subscriptions. The higher the severity level of psychological characteristics (psychopathy, anxiety, etc.), the more pronounced the interest in unsafe content (all Ps <0.05). 18. Description of the established relations of a high level of severity of individual psychological characteristics with the unsafe content consumption of deviant communities in the social network "VKontakte" (all Ps <0.05): Psychopathy - content about self-harm, aesthetics of death, subculture true crime, violence, hate; Machiavellianism - vulgar content, content about violence, hate, self-harm, weapons, subculture true crime; Narcissism - depressive, suicidal content, content about psychoactive substances, nationalism, subculture true crime; Anxiety and depression: depressive, suicidal, violent content, content about self-harm, psychoactive substances, evil, death, subculture true crime; Somatic self-harm: depressive, suicidal content, content about eating disorders, violence, subculture true crime. 19. Databases: - individual psychological characteristics ("Short Dark Triad", "Beck Anxiety and Depression Inventory ", "Scale of self-injurious behavior" by N.A. Polskaya, questionnaire "Social networks using") and social network characteristics of 7505 students and 1344 senior pupils (gender, age, education, marital status, place of residence; results of psychodiagnostics; profile data, subscriptions, posts); - individual psychological characteristics (“Short Big Five”, “Depression Anxiety Stress Scale-21”, Young “Internet Addiction Test”) and social network characteristics of 1880 students (gender, age, education, marital status, place of residence; results of psychodiagnostics; profile data, subscriptions, posts); - socio-demographic characteristics and digital footprint of 61609 senior pupils in the Siberian Federal District (gender, region of residence; data: subscriptions). 20. Publications in scientific journals: 7 publications, of which 6 publications - in publications indexed by Web of Science or Scopus (including 3 publications in editions of the 1st quartile), 3 publications - in editions included in the Russian Science Citation Index. 21. Web application "Platform for psychological diagnostics with authorization of respondents in the social network" VKontakte "". 22. Methodological recommendations for monitoring and reducing risk factors for digital security of children, adolescents and youth. 23. Publications in electronic media (14) and an updated website about the project: http://frpss.tilda.ws

Publications

1. Evseev V.D., Peshkovskaya A.G., Matsuta V.V., Mandel A.I. Несуицидальные самоповреждения (NSSI) и их связь с цифровыми данными социальной сети Академический журнал Западной Сибири, Том: 16. №3 (86). С. 38-40 (year - 2020)

2. Evseev V.D., Peshkovskaya A.G., Matsuta V.V., Mandel A.I., Bokhan N.A. Взаимосвязь цифровых маркеров онлайн-активности и социально-демографических характеристик лиц призывного возраста с несуицидальными формами самоповреждающего поведения Суицидология, Вып. 11. №3 (40). С. 72-83 (year - 2020) https://doi.org/10.32878/suiciderus.20-11-03(40)-72-83

3. Peshkovskaya A., Evseev V., Matsuta V., Myagkov M. Social media content preferences and non-suicidal self-injuries in youth European Neuropsychopharmacology, Volume 40, Supplement 1, S388 (year - 2020) https://doi.org/10.1016/j.euroneuro.2020.09.503

4. Peshkovskaya A., Matsuta V. How Social Media Big Data Can Improve Population-based Suicide Prevention European Neuropsychopharmacology, - (year - 2021)

5. Peshkovskaya A., Matsuta V., Evseev V. Time spent online on social networks linked with anxiety and personality traits European Neuropsychopharmacology, Volume 40, Supplement 1, S387–S388 (year - 2020) https://doi.org/10.1016/j.euroneuro.2020.09.502

6. Serbina G.N., Matsuta V.V., Goiko V.L. Анализ связи психологических характеристик пользователей социальной сети «ВКонтакте» с подписками на сообщества с девиантным контентом Вестник Томского государственного университета, - (year - 2021)

7. Sergey Chudinov, Galina Serbina, Valeria Matsuta, Yuliya Mundrievskaya, and Anastasia Peshkovskaya Dangerous social media: social network analysis of virtual communities related to school shooting and suicide Computers in Human Behavior, - (year - 2021)

8. - Платформа психологической диагностики с авторизацией респондентов в социальной сети «ВКонтакте» -, № 2021612933 (year - )

9. - Доклад сотрудника НИИ психического здоровья стал одним из самых обсуждаемых на авторитетной конференции в Лондоне Сайт Томского НИМЦ РАН, Доклад сотрудника НИИ психического здоровья стал одним из самых обсуждаемых на авторитетной конференции в Лондоне // Томский НИМЦ РАН, сентябрь 2020 (year - )

10. - Алгоритм ТГУ выявляет пользователей VK, подписанных на опасные группы Сайт Томского государственного университета, Алгоритм ТГУ выявляет пользователей VK, подписанных на опасные группы // Томский государственный университет, 8 июня 2020 (year - )

11. - Ученые ТГУ узнали сколько подростков читают "опасный" контент РИА Томск, Ученые ТГУ узнали сколько подростков читают "опасный" контент // РИА Томск, 8 июня 2020 (year - )

12. - Учёные ТГУ разработали алгоритм для выявления студентов, которые подписаны на опасный контент Агентство новостей ТВ2, Учёные ТГУ разработали алгоритм для выявления студентов, которые подписаны на опасный контент // Агентство новостей ТВ2, 8 июня 2020 (year - )

13. - Более трети подростков подписаны на опасные для психики группы — ученые РФ Общероссийская общественная организация защиты семьи, Более трети подростков подписаны на опасные для психики группы — ученые РФ // Общероссийская общественная организация защиты семьи, 8 июня 2020 (year - )

14. - 25 марта – онлайн-семинар по цифровой безопасности Сайт Томского государственного университета, 25 марта – онлайн-семинар по цифровой безопасности // Томский государственный университет, 23 марта 2021 (year - )

15. - Ученые ТГУ выявили у молодежи рост интереса к опасному контенту в VK Сайт Томского государственного университета, Ученые ТГУ выявили у молодежи рост интереса к опасному контенту в VK // Томский государственный университет, 17 мая 2021 (year - )

16. - Томские ученые выявили у молодежи рост интереса к опасному контенту в VK Мультимедийный портал ПОИСК, Томские ученые выявили у молодежи рост интереса к опасному контенту в VK // Мультимедийный портал Поиск, 17 мая 2021 (year - )

17. - Томские ученые выявили у молодежи рост интереса к опасному контенту в VK Портал Social media news, Томские ученые выявили у молодежи рост интереса к опасному контенту в VK // Портал Social media news, 17 мая 2021 (year - )

18. - Сибирская молодёжь, по мнению томских учёных, тянется к небезопасному контенту в соцсетях Портал ЧС Инфо, Сибирская молодёжь, по мнению томских учёных, тянется к небезопасному контенту в соцсетях // Портал ЧС Инфо, 17 мая 2021 (year - )

19. - Ученые ТГУ: интерес молодежи СФО к опасному контенту VK вырос в 2 раза РИА Томск, Ученые ТГУ: интерес молодежи СФО к опасному контенту VK вырос в 2 раза // РИА Томск, 17 мая 2021 (year - )

20. - Молодежь стала больше интересоваться опасным контентом в VK - томские ученые Томский обзор, Молодежь стала больше интересоваться опасным контентом в VK - томские ученые // Томский обзор, 17 мая 2021 (year - )

21. - Интерес сибирской молодежи к опасному контенту вырос вдвое Портал Sibnet.ru, Интерес сибирской молодежи к опасному контенту вырос вдвое // Портал Sibnet.ru, 17 мая 2021 (year - )

22. - Цветы зла: томские ученые выявили за год у молодежи Сибири рост интереса к опасному контенту АО ИД «Комсомольская правда», Цветы зла: томские ученые выявили за год у молодежи Сибири рост интереса к опасному контенту // АО ИД «Комсомольская правда», 17 мая 2021 (year - )