Using the Internet has become one of the most popular leisure activities among postsecondary students in China. Concern about the large number of students using the Internet has led to an increase in research on the influencing factors of Internet addiction and the negative consequences caused by it. This short-term longitudinal study examined the associations among three dimensions of social support [objective support (OS), subjective support (SS), and support utilization (SU)], loneliness, and the four dimensions of Internet addiction (compulsive Internet use [CIU] & withdrawal from Internet addiction [WIA], tolerance of Internet addiction [TIA], time-management problems [TMPs], and interpersonal and health problems [IHPs]) in a Chinese sample. A total of 169 postsecondary first-year students (88 girls and 81 boys; mean age = 18.31 years) participated in the study. The questionnaire measurements were taken at the beginning of the school year (T1), 6 months later (T2), and 1 year later (T3). Cross-lagged and structural equation modeling analyses indicated that (a) OS (T1) and SU (T1) negatively predicted loneliness (T2); and loneliness (T2) negatively predicted OS (T3) and SU(T3); (b) CIU & WIA (T1) and TMPs (T1) positively predicted loneliness (T2); and loneliness (T2) positively predicted CIU & WIA (T3), TIA (T3), TMP (T3), and IHP (T3); (c) SS (T1) directly affected TIA (T3) and TMP (T3); and (d) loneliness (T2) played a mediating role in the relationships between OS (T1) and CIU (T3), OS (T1) and TMP (T3), OS (T1) and IHP (T3), and SU (T1) and IHP (T3). Finally, interventions for Internet addiction and implications for future studies were discussed.
13 October 2019 in Scandinavian Journal of Psychology
Recently, combined quantitative and qualitative analysis has become popular for research. In studying careers, subjective and objective information are ideal for assessing individual career development and are relevant in career counseling. This paper measures career adaptability by combining text mining and item response theory (IRT), with college students' self-reported career adaptability as a subjective measure and responses to questionnaire items as an objective measure. The two are combined under a Bayesian framework. Additionally, the validity of text categorization and IRT, combined with model measurement, were explored; text categorization results were used as prior information when estimating IRT capability parameters to test whether adding prior information can improve accuracy. This study draws the following conclusions: (1) The text classification method had the highest sensitivity in 300-person samples; however, the text-IRT method had the best predictive effect, high reliability, and unique advantages in accuracy. (2) In 600-person samples, the text classification method had the best predictive effect. The effect was relatively good, with unique advantages in identifying low career adaptability. However, this must be selected according to actual needs. If the accuracy requirement is high and sensitivity can be sacrificed, the text-IRT method is more appropriate. (3) The text-IRT method is more suitable for 900 subjects when accuracy, sensitivity, and specificity need to be considered, and text classification is best when identifying low career adaptability. (4) Sample size influenced accuracy, specificity, and the negative predictive values of text classification, as well as the sensitivity of IRT and text-IRT methods.
The prevailing “996” overtime phenomenon in China has raised extensive consideration and discussion towards the topic of work-life balance. Following this trend, this study focused on the topic of work recovery experience. Based on Lens Model, we aimed to construct prediction models of weekend recovery experience with individuals’ social media footprints, which include their social media posts, behavioral information, and demographic information. We acquired Weibo data and Recovery Experience Questionnaire results from 493 participants and extracted Weibo data features for model training through two methods. As a result, two types of model were constructed: regression models which applied Ridge Regression, LASSO Regression, and Elastic Net; classification models which applied Gradient Boosting Decision Tree, Logistic Regression and Support Vector Machine. For the results of regression models, Pearson correlation coefficients between predicted values and self-reported scores ranged from 0.40 to 0.84; for classification models, F1-score ranged from 0.49 to 0.78. The results showed that individuals’ recovery experience on weekends could be predicted by their social media footprints. What is more, the methodology proposed in this study could help organizations to evaluate large groups of employees’ work recovery in real-time, which will have further implications for both theoretical and practical purposes.