Annotation of the results obtained in 2021
PREDICTORS OF DEPRESSION We included the PHQ-8/9 questionnaire in the TREC panel (https://trec.hse.ru/en/) and obtained unique data on the severity of depression in study participants in 2018 and 2020. Also, the TREC panel provides a large amount of additional information about the participants. We analyzed this data set and identified early predictors of depression: women gender, health problems, lack of plans for the future, a large amount of time spent on the Internet / in social networks / playing computer games, financial difficulties and problems finding a job, difficulties expressing thoughts, dissatisfaction with the chosen higher educational institution, neuroticism. We found a negative association with the severity of depression symptoms in the case of the following factors: sleep duration, parental involvement, doing sports, extraversion and conscientiousness. Using the identified factors, we constructed a predictive model that explains about a quarter of the variation in the severity of depression (PHQ-9). The model can be used to assess the risks of depression. On average for the sample, the risk of depression is 16%, however, those who are in the lower quartile of the risk according to the model's estimation have an observed risk of only 3%, and those who are in the upper quartile according to the model's estimation have an observed risk of depression of 40%. An article based on the research results is being prepared for publication. In addition, we studied the relationship between the frequency of social media use and the severity of depression symptoms in schoolchildren using a data set collected in Yakutia, a region in Russia (N = 1364). Based on these data, the frequency of social media use is also positively associated with depression. However, we found that this relationship is moderated by the degree of parental involvement: in the case of high involvement, the relationship is significantly weaker. DIGITAL TRACES AND PREDICTION OF DEPRESSION We examined the relationship between changes in activity patterns on social networking site Vkontakte and changes in the severity of depression symptoms. From 2018 to 2020, the severity of depression symptoms decreased in 40% of TREC participants, did not change in 14% and increased in 46%, and increased by more than 3 points in 25%. We analyzed whether an increase can be predicted based on a) the changes in the number of friends and the density of friendships b) a shift in the time of posts to a later time c) an increase in the number of negative posts on Vkontakte. Based on the results of the analysis, we did not find any significant dependencies. LIMITATIONS OF THE NAIVE USE OF DIGITAL TRACES We continued to explore the discrepancy between the model which estimates the sentiment of posts and the model which predicts depression. The problem was that negative sentiment in posts was associated with both a user’s severity of depression symptoms and estimations of the model predicting depression, so it could be inferred that negative sentiments of posts represent roughly the same phenomena. This assumption is often made in computational social science. At the same time, when trying to look at the relationship of depression with educational achievement on a large data set, the use of the sentiment model and the model predicting depression led to opposite results. We have proposed a model to explain this paradox. The model assumes that bias in the sentiment model correlates with academic performance. That is, positive words that predict depression (model error) are used much more often by students from schools with higher educational attainment. Having estimated the parameters of the model based on empirical data, we showed that it explains the results we found. The results were presented at the International Conference on Computational Social Science (IC2S2) 2021. COMMUNICATION OF STUDENTS DURING THE COVID-19 PANDEMIC We collected public data from social networking site VKontakte about students of Russian universities and studied the impact of the COVID-19 pandemic and related lockdown on students' public communication on the social networking site. The sample included students from Russian federal universities. We analyzed changes in the number of posts, the reaction to them (number of “likes”, reposts, and views), the sentiment of posts, and their topics. We found that the volume of public communication on VKontakte and the positivity of posts decreased during the lockdown in Russia in spring 2020 compared to the previous year and returned to the expected values ​​by the summer of 2020. Despite the pandemic, students wrote little about health and the COVID-19 virus in their public posts on VKontakte; most of the posts are related to congratulations, discussion of personal achievements, and study. Apparently, public posts on VKontakte are mainly used as a tool for self-presentation and are rarely used to discuss topics related to feelings, emotions, and experiences. This may partly explain why, in general, public records have proven to be a weak signal in predicting depression. The paper is being prepared based on the results of the research. ANALYSIS OF THE REPRESENTATION OF PSYCHOLOGICAL WELL-BEING ON YOUTUBE We examined the representation of mental health and well-being on YouTube, one of the most popular social media platforms among Russian Internet users aged 18–44. We have studied the official policy of censoring video content on YouTube and also identified informal user practices to restrict the distribution of “unsafe” content. We compiled a list of videos (N = 740) recommended by YouTube for searches related to mental health and well-being, and then viewed and analyzed 40 of the most popular ones. YouTube tightly controls the distribution of potentially dangerous content on its platform and applies sanctions to violators from temporary blocking to account deletion. In addition, users themselves form communities to protect themselves from unsafe content, in which they post lists of unsafe channels, urge users to unsubscribe from them, and complain to YouTube support about the producers of this content. As a consequence, we were unable to find a significant number of overtly positive “romanticized” representations of mental health problems. We identified the types of users who publish videos on the topic of psychological well-being, and found that in most cases Russian-language representations of depression on YouTube are based on the medical model of depression and are medicalized: the authors operate with words and expressions from medical discourse and represent depression as a disease. It can be concluded that on the Russian-language YouTube there is no clearly expressed romanticization of mental illness in general and depression in particular. However, in some cases, depression is presented as a “temporary bad mood” or a condition that promotes creative activity, which can lead to distorted ideas in the average person. YOUTUBE’S RECOMMENDATION ALGORITHM A primary analysis of the YouTube recommendation algorithm was carried out. To do this, YouTube accounts were created on separate virtual machines to ensure that these accounts did not have any search history. Videos pertaining to psychological well-being, including romanticizing psychological / eating disorders, were watched from these accounts using a variety of strategies, and YouTube algorithm recommendations were examined. For newly created accounts, the YouTube home page contains recommendations of the most popular videos. As we watched specially selected videos about an eating disorder (anorexia), the proportion of videos about anorexia grew rapidly in the recommendations; after 2-3 videos, there were from four to six videos about anorexia in the top recommendations. As we continued watching the videos, the number of such recommendations stabilized around seven videos. When we started watching random videos (not related to eating behavior), the number of recommendations decreases but does not disappear completely. This preliminary analysis allows you to design the next stage of the study. We can parameterize user behavior through a preference for viewing videos on a specific topic, where 0 means viewing random videos from the recommended ones, 1 means viewing only videos on a specific topic; the parameter can take any intermediate values. After that, it will be possible to study how the proportion of videos on a certain topic among the recommended videos changes depending on the time and this parameter, or in other words to quantify the danger of falling into the “information bubble”. We plan to conduct such a study and publish the results. CHATBOT We studied the interaction of adolescents and young people with the Eli chatbot, developed by the UNESCO Institute for Information Technologies in Education (IITE) in collaboration with VKontakte. The chatbot is aimed at improving knowledge about psychological well-being, health, physiology, and relationships. Since these topics are sensitive, the history of communication with the chatbot is not saved. However, those user requests to which the chatbot could not find an answer are saved anonymously. Despite the limited information contained in such a dataset, it nevertheless allows one to explore which topics are of the most interest to users. We analyzed 1000 messages randomly selected from the dataset. Analysis of messages showed that chatbot users lack information on various psychological issues and on the functioning of the chatbot itself. The analysis also revealed the problem, that in the case of several sensitive topics users phrase their requests differently than the developers expect, and because of that the chatbot cannot identify them. These requests are related to relationships, sex, and psychological problems (16.4%). We also launched an experiment to study the effectiveness of a chatbot. We decided to test whether chatbot usage affects mental health literacy, namely help-seeking behavior. To conduct the experiment, students of a Russian university were sent an announcement about a study of various resources about psychological well-being. All students who agreed to participate were divided into three groups a) a group that was asked to interact with Eli's chatbot b) a group that was asked to read articles on psychological well-being in a blog maintained by UNESCO IITE experts, c) a control group without intervention. Groups a) and b) are similar in content but differ in format. Psychological health literacy was measured before and after exposure using the Attitude Toward Seeking Professional Psychological Help scale-short form (ATSPPH-SF) [1]. Unfortunately, only 72 people took part in the chatbot experiment. Given the presence of two experimental and one control group, this is not enough to analyze the effect of using a chatbot. In the initial survey, the average value on the ATSPPH-SF scale was 17.7 points, which is below the accepted cut-off of 20 points and indicates the prevalence of negative attitudes towards seeking psychological help in case of problems with psychological well-being among students. However, this finding is also limited by the small sample size.

 

Publications

---

Annotation of the results obtained in 2019
DATA We decided to choose depression as the main indicator related to psychological well-being. Psychological well-being is a complex phenomenon and it cannot be reduced to depression, however, the focus on depression has several advantages: 1) Depression is well studied. It allows to put our results in a broader context and to compare them with existing literature. 2) There are established relations between depression and important life outcomes. It is not the case for some other measures, i.e. it is not always known if having a low score on some subjective well-being measure is related to objective negative live outcomes. 3) The risk of depression is determined not solely by genetic predisposition but also by the environment, including the social environment. It means that various social factors might be related to depression and it makes sense to identify and study them in order to propose interventions that could reduce the risk of depression. At the same time, depression is understudied in the Russian context and especially in the context of the Russian education system. We choose the PHQ-8/9 scale as a depression measure. This questionnaire is widely used and has psychometric properties similar to other scales (e.g. to the Beck Depression Inventory-II [Kung et al., 2013]). It also has the advantage of being shorter. We included PHQ-9 in the nationally representative panel “Trajectories in education and careers” (TrEC) (https://trec.hse.ru/en/). This study tracks more than 4000 students who participated in the PISA program [OECD, 2014] in 2012. We collected publicly available information from VK (the most popular Russian social networking site) for those participants who agreed to share their data (N = 3483). This information includes friendship networks, their interests (subscriptions to various groups) and public posts. This effort resulted in a unique data set that includes a depression measure, digital traces, and extensive longitudinal context data. We also included PHQ-9 to the “Student Life” survey. This allowed us to collect information about symptoms of depression for 4167 students from one university. This data could be combined with extensive survey and administrative data. We also designed and conduct a study in one of the Moscow schools. The sample included all 5- to 11-grade students (438) with a response rate of 58%. Participants filled the survey that included PHQ-8, questions about social connections and social integration, school climate, the use of digital technologies, parental control, family environment, and quality of sleep. Apart from the survey data, we also collected data using an experience sampling method. We developed a chat-bot for the social networking site popular among students and used it to ask questions about mood, situational anxiety and bedtime daily. The public data from VK (list of friends, subscriptions to groups, public posts, and likes) was also collected. Finally, we developed a guide for a semi-structured interview with students. The guide includes questions about relations with peers, family, school experience, educational and career expectations, etc. We conducted 10 interviews with students. SOCIAL MEDIA USE AND SYMPTOMS OF DEPRESSION We prepared a literature review “Internet Use and Depressive Symptoms in Adolescents”. The paper was published in the journal Clinical Psychology and Special Education (http://psyjournals.ru/en/psyclin/2019/n3/Bochaver_et_al.shtml). The results of empirical studies conducted in different countries are ambiguous and do not allow to speak about universal effects that apply to all children and adolescents. Adolescents who actively and emotionally use social media services often demonstrate symptoms of lower well-being: lower quality of sleep, lower self-esteem, higher levels of anxiety and depression. At the same time, many studies support the Goldilocks hypothesis, i.e. that the moderate use of technologies has positive effects on children and adolescents. We assume that the lack of unambiguous conclusions about the impact of digital technologies is due to the non-linearity and differential effects. In particular, the effects might depend on the individual characteristics, the intensity of the use of technology, socio-economic characteristics, and the child environment. This review shows the need for a differentiated approach to study the impact of digital technologies on the well-being of children. This approach would require to split a sample into smaller subgroups and could be infeasible in many cases where traditional survey methods are used. However, the use of digital traces, as proposed in our project, could help to collect information on a larger scale and, as a consequence, to study non-linear relations and relations that are different for different groups. USING DIGITAL TRACES TO PREDICT USERS CHARACTERISTICS To realize the potential of digital traces in well-being research it is important to be able to predict the relevant characteristics of users using digital traces. It has been already suggested that texts of posts on social media could be used to predict depression [De Choudhury et al., 2013; Eichstaedt et al, 2018]. We are going to use TrEC data to predict depression from public posts on VK. The accuracy of such model is usually rather low on the individual level because of the noise in the data. However, such models could be used on an aggregated level, e.g. on the level of schools or universities. We are going to build the model to estimate the average level of depression for various schools and universities and then to identify factors associated with this level. One limitation of this approach is that the resulting model would be hard to validate as the information on depression levels is not available for schools and universities. That is why we decided to conduct another study that would follow the same methodology but use another characteristic instead of depression. The choice of this characteristic is determined by our ability to validate it on a level of educational organization and that is why we choose academic performance. We use the TrEC data set to build a model that predicts academic performance from public posts on social media. The predictive power of the model depends on the number of posts that are available per user. If only one post is available than the correlation between real and predicted score is rather low (Pearson’s r = 0.24), however, this value increases to r = 0.55 if 20 posts per user are available. At the next step, we downloaded public posts of users from all schools of three Russian cities (Saint Petersburg, Samara, and Tomsk). We also downloaded public posts of students from 100 largest Russian universities. For each of the educational organization, we computed predicted academic performance by averaging over all users from this organization and then compared the predicted value with the USE (Unified State Examination) scores. We use average USE scores of graduates for schools and average USE scores of enrollees for university. The predicted performance was closely related to the USE scores, e.g. correlation between predicted scores and USE scores was r = 0.83 for universities. Note that the model was trained not only on a different set of users but the measure of academic performance was also different (i.e. PISA scores instead of USE scores). We also showed that the predictive power of our model is not decreased if VK posts are substituted by tweets of the same users. The results were presented at the International Conference on Computational Social Science (IC2S2), the paper is currently under review, a preprint is available at https://arxiv.org/abs/1912.00463. EMOTIONAL PULSE OF SCHOOLS An alternative approach to getting insights into the psychological well-being of students is to study the sentiment expressed in their public posts. We trained a model to predict the sentiment of users' posts (negative, neutral, or positive) and validated it using an annotated set of VK posts. The accuracy of the resulting model was relatively high (AUC = 0.76 for positive sentiment and AUC = 0.87 for negative sentiment). We applied the model to 6 million posts of users from Saint Petersburg’s schools. We find that sentiment follows expected diurnal and seasonal patterns. We also find that the sentiment of posts is negatively correlated with the educational outcomes of schools, i.e. students from high performing schools express less negative sentiments. The correlation is more pronounced for girls (Pearson’s r = -0.34) than for boys (r = -0.10). The observed difference between the top 20% of schools and the bottom 20% of schools is rather large (Cohen’s d = 0.95). This is, for instance, larger than the difference between weekday and weekend posts but lower than gender difference in sentiment. The results were presented at the International Conference on Social Informatics, the poster is available at https://ibsmirnov.com/emotional_pulse.pdf DIGITAL TRACES AND SURVEY DATA Digital traces have a large potential in research on psychological well-being. However little is known about their validity and relations to traditional survey measures. We combined survey data with data from a popular Russian social networking site VK to analyze the validity of different features of online behavior as indicators of three aspects of well-being: depression and anxiety, social connections, and sleep. We conducted a survey of high school students to measure symptoms of depression, anxiety, social connections, and sleep quality. We also used a mobile application to ask questions about mood, anxiety, and wake-up time/bedtime daily. We used the average sentiment of posts, the proportion of negative posts, the proportion of late-night posts, the average number of “likes” per post, and the number of friends on VK as features of online behavior. We found that the strength of negative emotions expressed in posts is associated with the severity of the symptoms of depression. The severity of depression is also associated with the proportion of negative posts among all posts written by a user. The proportion of posts written late at night significantly correlates with late bedtime. The quality of sleep is not correlated with the proportion of night posts, but it is significantly lower for those who have at least one late-night post. With regard to social relations, we found that “popular” students (those who were named at least once as a popular student at school) on average had significantly more friends and “likes” on VK among the students from the same school, and “unpopular” students — fewer friends and likes. We have found that there is a correlation between indicators of well being obtained using digital traces and using survey data, and this relationship has an expected direction. However, the correlations are rather week, i.e. survey and digital trace data are not interchangeable. The results may still be of practical importance, e.g. they could be used in the management of educational organizations (for example, in monitoring the "mood" of the school). It is recommended to use such approaches only at the organization level, not at the level of individual students, and follow ethical principles (voluntary participation, obtaining informed consent, protection of personal data, etc.). SLEEP PATTERNS AND ACADEMIC PERFORMANCE Sleep is one of the key components of individual well-being, and lack of sleep and other sleep disorders make a significant impact on the physical and mental health, such as cardiovascular diseases, metabolic syndrome and depression [Simon et al., 2019; Pires et al., 2016]. In this study, we examine the relationship between academic performance and sleep patterns. This study is conducted using TrEC data. We use information about the academic achievements of students (PISA test results), as well as information on their sleep patterns. Data on sleep patterns was obtained from a survey (bedtime and wake-up time of the participants) and social media posts of the participants. We use an “online night activity” indicator which is computed as the fraction of late-night posts (1-5 a.m.) among all user posts and varies from 0 to 1. This indicator was validated using survey data. We find that respondents with higher academic performance more often go to bed after 1 a.m. (late) and less often before 1 a.m. (early). Respondents who report less sleep on weekdays also show higher performance than those who sleep more. Additionally, we find a negative correlation between sleep duration in working days and academic performance in mathematics, science and reading for women. Online night activity is positively associated with PISA results in mathematics, science, and reading, confirming the results obtained using the survey data. We fitted a series of ordinary least squares regression (OLS) models to evaluate the impact of online night activity in the PISA score for each subject. Results suggest that after taking socio-economic status and gender into account, there is still a significant association between the night online activity and PISA scores in all three subjects. However, a separate analysis of data on men and women shows that PISA scores are positively associated with high late-night online activity only for women. REFERENCES De Choudhury, M., Gamon, M., Counts, S., & Horvitz, E. (2013). Predicting depression via social media. In Seventh international AAAI conference on weblogs and social media. Eichstaedt, J. C., Smith, R. J., Merchant, R. M., Ungar, L. H., Crutchley, P., Preoţiuc-Pietro, D., ... & Schwartz, H. A. (2018). Facebook language predicts depression in medical records. Proceedings of the National Academy of Sciences, 115(44), 11203-11208. Kung, S., Alarcon, R. D., Williams, M. D., Poppe, K. A., Moore, M. J., & Frye, M. A. (2013). Comparing the Beck Depression Inventory-II (BDI-II) and Patient Health Questionnaire (PHQ-9) depression measures in an integrated mood disorders practice. Journal of affective disorders, 145(3), 341-343. OECD (2014), PISA 2012 Results: What Students Know and Can Do – Student Performance in Mathematics, Reading and Science Pires, G. N., Bezerra, A. G., Tufik, S. & Andersen, M. L. Effects of acute sleep deprivation on state anxiety levels: a systematic review and meta-analysis. Sleep. Med. 24, 109–118 (2016). Simon, E. B., Rossi, A., Harvey, A. G., & Walker, M. P. (2019). Overanxious and underslept. Nature Human Behaviour, 1-11.

 

Publications

---

Annotation of the results obtained in 2020
PUBLISHED WORK Estimating educational outcomes from students' short texts on social media (https://epjdatascience.springeropen.com/articles/10.1140/epjds/s13688-020-00245-8), Measuring Adolescents' Well-Being: Correspondence of Naïve Digital Traces to Survey Data (https://link.springer.com/chapter/10.1007/978-3-030-60975-7_26) SENTIMENT ANALYSIS AND PREDICTION OF DEPRESSION Before building a model to predict depression from the social media posts, we estimated the predictive power of the sentiment of posts in relation to depression, using information about the posts of 1,637 VK users for whom the results of the PHQ-8 were known. The relationship between the proportion of negative posts and the severity of depression symptoms varies depending on the number of available posts and the time they were published in relation to the time of the survey. If a sufficient number of relatively recent posts (two months before the survey) is available then the correlation between the proportion of negative posts and PHQ-8 is rather high (r = 0.65). Using the methodology proposed in the paper “Estimating educational outcomes from students’ short texts on social media ”, we also built a model to predict depression from social media posts. The model's accuracy is comparable to the results obtained for the sentiment (r = 0.6). The model was then applied to the posts of students from 600 high schools in one city and the 100 largest Russian universities. We obtained the results contradicting those that were previously obtained using the sentiment of posts: in high-performing educational organizations there are fewer negative posts, but more users are predicted to have depression (r = -0.34 vs r = 0.31). This is especially surprising given that the “negativity” and “depressiveness” of posts are strongly correlated and it would be natural to assume that these are similar if not interchangeable characteristics. To explore this discrepancy in more detail, we compared the results obtained by two models for individual words. We found a large number of negative and “depressive” words (infuriate, depression, idiot, tired, hate, loneliness, die, etc.). There was also a large number of positive and anti-depressive words. These are mainly words related to family, friendship, walking, and travel. There are practically no words that would be both strongly negative and strongly anti-depressive. These three observations are in line with expectations and explain the high agreement between the sentiment model and the model to predict depression. However, the remaining category is an exception, namely, there is a large number of strongly positive, but at the same time, strongly depressive words (inspiration, dream, wisdom, freedom, happiness, etc.) are found. This may explain why we obtain contradicting results for the relationship between academic performance and posts’ sentiment on one hand and depressiveness of posts on another hand. This result might have implications outside our research program since an analytical strategy similar to ours is used in many computational social science papers. ONLINE SOCIAL NETWORK STRUCTURE AND SYMPTOMS OF DEPRESSION Using TrEC data, we studied the relationship between the structure of the personal online network and the severity of depressive symptoms. A negative association was found between network size and depression. However, this holds true only for the size of the network core; the size of the periphery (isolated nodes) is not associated with depression. This is also true only for relatively small networks (up to ~ 150 nodes, which, curiously, correspond to the Dunbar number – the estimated maximum number of permanent social connections that a person can maintain). These results indicate that not only offline but also online social connections can act as a protective factor against depression. It also indicates that not all online connections are equally important and that users may use different strategies when forming an online friendship network. It is important to consider these factors when analyzing online social networks. ONLINE SOCIAL INTEGRATION AND SYMPTOMS OF DEPRESSION (HIGH-SCHOOL STUDENTS) We also examined the relationship between online social integration and the severity of depression using data on high school students from one school. We found that the number of friends on VK is negatively associated with the severity of depression, however, this is only true for friends from the same school, i.e. the number of friends who do not study at the same school is not associated with depression. Thus, as in the previous study, we have found that not all online social connections are equally important. We also found that the number of received “likes” correlates with the severity of depression: posts of adolescents with more pronounced symptoms of depression receive, on average, fewer “likes” per post than adolescents with less severe symptoms. Also, the number of posts positively correlates with the severity of depression. The severity of depression is not correlated with the total number of sent likes, however, it is positively correlated with the number of likes sent to popular students (those were mentioned as popular by other students in a survey). Consistent with theories regarding social media comparison processes, this finding suggests that interactions with popular accounts may increase the risk of depression. EDUCATIONAL ACHIEVEMENTS AND DEPRESSION We examined the relationship between educational outcomes and depression using data from the TrEC panel. The prevalence of depression in this sample (PHQ-9> = 10) is 12% among men and 20% among women. We found no association between educational outcomes and depression for women (Pearson's r = 0.02), but we found a positive association for men (r = 0.15). This corresponds to an increase in the risk of depression from 4% for low performing men to 17% for high performing men. This result holds after we control for the socioeconomic status of the participants, the fact of studying at the university, and the selectivity of the university. These results might seem to contradict existing work on the relationship between academic performance and depression. Most of the studies have found a negative relationship that can be explained by the negative effect of depression on cognitive functioning or by the fact that low academic performance can lead to stress that increases the risk of depression. However, almost all studies use relatively homogeneous data, such as data on students from one school or one university. In the studies that use data on students from different educational institutions, academic performance is usually measured via GPA. As GPA is a relative measure, it is impossible to directly compare students from different educational organizations in these cases too. Overall, our results suggest that the relationship between depression and academic achievement might be more complex than is commonly assumed. STUDENT SATISFACTION AND DEPRESSION We also studied data on students from one university. In this case, it makes less sense to study socio-demographic factors, so we decided to focus on those factors that are in principle could be controlled by a university, and analyzed the relationship between the severity of depressive symptoms and student satisfaction with the educational process and various services provided by the university. We found a negative relationship between the level of depression and student satisfaction, including factors ranging from satisfaction with their educational program to satisfaction with canteens (more than 20 indicators). Correlation ​​ranges from -0.1 to -0.34. In addition to individual indicators, we also considered the integral satisfaction indicator. This indicator is negatively associated with depression after controlling for the respondent's gender, year of study, mode of study, and their financial difficulties. We also found that depression was associated not only with satisfaction at the time of the survey but also with satisfaction a year ago. That demonstrates the predictive power of this characteristic in relation to depression in the future. DEPRESSION AND PARENTAL INVOLVEMENT We studied the relationship between various forms of parental involvement and depressive symptoms in one of the regions of Russia (Yakutia), using data from a survey of 1364 schoolchildren from 5-11 grades. Factor analysis suggests that there are various types of parental involvement: “Help and support” is associated with the most resource-intensive positive parenting practices: helping the child with difficult school subjects and joint activities with the child; “Friendship” combines positive parenting practices that are supposedly less time-consuming for a parent and are associated with maintaining friendly relations with the child, and being interested in their life; “Control” is most strongly associated with the controlling and regulating activities of a child, e.g. checking on homework, regulating the use of gadgets, etc. Independently, all types of parental involvement are protective factors against depression. However, when these factors are considered together, controlling practices are not associated with depressive symptoms. Their association with the severity of depressive symptoms is explained by the fact that positively involved parents also tend to exhibit higher levels of control. DISCOURSE ON THE WELL-BEING OF STUDENTS We conducted a study on the “quality of education” construct to analyze how the notion of the students’ well-being has been incorporated into it over the past 20 years. We analyzed reports on the seven waves of the PISA project. Focusing on a single international project makes it possible to trace the evolution of educational discourse. We have found that over the successive waves of the PISA, there is a shift in research interest from a simple recording of academic achievements in different countries to inclusion of various indicators of the quality of the educational environment and then to a discussion of the students’ well-being. Initially, the characteristics of the school environment were considered exclusively as factors contributing to academic achievements, but then the quality of life and the quality of the school environment started to be considered as important indicators per se. The meaning of “education” has changed: it is now considered not only as a step needed to prepare for the future life but also as an important part of the life cycle, valuable in itself. SUBJECTIVE WELL-BEING OF STUDENTS Based on the interviews with high school students, we have analyzed how the process of self-determination, associated with the necessity to choose the future life and educational trajectory, affects perceived stress levels and subjective well-being of students. The results of the study showed that the age-related task of self-determination, intensified by the situation of leaving school, plays a key role in the subjective well-being of students by their own assessment. Elevated levels of anxiety and stress are caused by the choices that students are forced to make by the educational system: the choice of a university, the choice of subjects for the Unified State Examination, the necessity to prioritize activities in a situation of time deficit, and also the necessity to choose the image of their own future. It follows from the interviews that the stress from these problems is higher for students who have higher educational aspirations and/or study in schools with a highly competitive environment. These results allow us to hypothesize that school status plays an important role in the subjective well-being of adolescents, this hypothesis requires confirmation by quantitative methods.

 

Publications

---