Online interaction turns the congeniality bias into an uncongeniality bias

Online phenomena like echo chambers and belief polarisation are believed to be driven by humans’ penchant to selectively expose themselves to attitudinally congenial content. However, if like-minded content were the only predictor of online behaviour, heated debate and flaming on the Internet would hardly occur. Research has overlooked how online behaviour changes when people are given an opportunity to reply to dissenters, potentially turning a preference for attitudinally congenial information into a preference for uncongenial information. Three main experiments consistently show that in a discussion forum setting where users can respond to earlier posts, larger conflict between user attitude and post attitude predicts higher likelihood to respond. The effect of conflict on response behaviour is shaped by the attitudinal composition of the forum, and it also predicts subsequent polarisation of users’ attitudes. These results suggest that belief polarisation on social media can be driven by conflict rather than congeniality.

Detection of Cyberbullying Through BERT and Weighted Ensemble of Classifiers

Due to the recent popularity and growth of social media platforms such as Facebook and Twitter, cyberbullying is becoming more and more prevalent. The current research on cyberbullying and the NLP techniques being used to classify this kind of online behaviour was initially studied. This paper discusses the experimentation with combined Twitter datasets by Maryland and Cornell universities using different classification approaches like classical machine learning, RNN, CNN, and pretrained transformer-based classifiers. A state of the art (SOTA) solution was achieved by optimising BERTweet on a Onecycle policy with a Decoupled weight decay optimiser (AdamW), improving the previous F1-score by up to 8.4%, resulting in 64.8% macro F1. Particle Swarm Optimisation was later used to optimise the ensemble model. The ensemble developed from the optimised BERTweet model and a collection of models with varying data representations, outperformed the standalone BERTweet model by 0.53% resulting in 65.33% macro F1 for TweetEval dataset and by 0.55% for combined datasets, resulting in 68.1% macro F1.

Detection of Cyberbullying Through BERT and Weighted Ensemble of Classifiers

Due to the recent popularity and growth of social media platforms such as Facebook and Twitter, cyberbullying is becoming more and more prevalent. The current research on cyberbullying and the NLP techniques being used to classify this kind of online behaviour was initially studied. This paper discusses the experimentation with combined Twitter datasets by Maryland and Cornell universities using different classification approaches like classical machine learning, RNN, CNN, and pretrained transformer-based classifiers. A state of the art (SOTA) solution was achieved by optimising BERTweet on a Onecycle policy with a Decoupled weight decay optimiser (AdamW), improving the previous F1-score by up to 8.4%, resulting in 64.8% macro F1. Particle Swarm Optimisation was later used to optimise the ensemble model. The ensemble developed from the optimised BERTweet model and a collection of models with varying data representations, outperformed the standalone BERTweet model by 0.53% resulting in 65.33% macro F1 for TweetEval dataset and by 0.55% for combined datasets, resulting in 68.1% macro F1.

Cyberchondria and Medical Student Syndrome

Health anxiety is a disorder that can be very distressful and cause unnecessary examinations. A doctor is expected to handle health anxiety in terms of examination, diagnosis, therapy, and counselling processes. To provide optimal patient counselling, a doctor needs confidence, empathy, and good mental health. However, the process to become a medical doctor requires a student to read through a large amount of medical information, which arguably might induce “medical student's syndrome,” or health anxiety. Contradicting research findings exist about such conditions, however, most of them use traditional measures of health anxiety and do not consider students' online behaviour. The authors hypothesized that a medical student is susceptible to cyberchondria, a form of health anxiety due to excessive internet use. Some studies have shown that there may be higher cyberchondria scores among medical students compared to the general population. Cyberchondria needs to be studied further to improve the mental health condition of medical students and to provide optimal future healthcare for patients.

Usability Analysis and Clustering Model in e-Learning from the User Experience Perspective

Online learning offers opportunities responding to their different individual and group learning needs by leaving digital traces that allow tracking their experiences at the user level. This study aims to examine the perceived usability of the gamified educational platform called (ELORS) in relation to online behaviour. As well as analyse the clustering models in terms of their high and low level of engagement through their interaction metrics. A quantitative, descriptive correlational approach and an educational data analysis design was adopted through the K-means algorithm. The participants were 51 students in mathematics in the second year of secondary education. An instrument was used to evaluate usability and behavioural metrics, analysing 1065 interactions with 57 activities. The results showed advantages in usability and grouping. The level of usability achieved depends on the interaction of the users with the different learning objects and their moderate relationship in their interactions. In relation to the centroids, two groups are evidenced by number of attempts and interactions, identifying students with low levels of participation in the minority. A significant finding is given in relation to the preference of redeeming virtual values in gold from the diamonds collected. The perspective of the analysis allows identifying the potential of the gamified platform to work online in the formation of mathematical competence according to the current educational curriculum.

Digital citizenship among Appalachian middle schoolers: The common sense digital citizenship curriculum

Background: Nearly 60% of teenagers in the USA have experienced abusive online behaviour. Identifying effective programmes to address these behaviours and promote digital citizenship is a research priority to reduce the rate of occurrence and consequential harmful effects of abusive online behaviour. Purpose: To evaluate the effectiveness of a Digital Citizenship Curriculum in increasing knowledge of digital citizenship and reducing cyberbullying and online aggression among middle-schoolers in an underserved community using a free curriculum. Method: Middle-schoolers participated in pilot implementation of a Digital Citizenship Curriculum (DCC) to evaluate its effectiveness in increasing knowledge of digital citizenship and reducing cyberbullying and online aggression. Follow up interviews were conducted to explore participants’ perceptions of the curriculum. Results: Participants demonstrated a statistically significant increase in their knowledge of digital citizenship with an increase of 2.96 in the mean score ( p < .001). Paired t-tests by gender demonstrated a significant difference in pre-post assessment mean scores for girls ( p < .001). Post-intervention perceptions indicate the curriculum was positively received and informative. Conclusion: Identifying cost-effective and resource-friendly programmes that support social-emotional learning and promote digital citizenship is crucial for underserved populations. Regions such as Appalachian Ohio often lack the resources to fund costly curriculum aimed at online aggression prevention. This study supports the implementation of the DCC and indicates the need for future research on the long-term effects of the curriculum on middle school participants.

Modeling E-Behaviour, Personality and Academic Performance with Machine Learning

The analysis of student performance involves data modelling that enables the formulation of hypotheses and insights about student behaviour and personality. We extract online behaviours as proxies to Extraversion and Conscientiousness, which have been proven to correlate with academic performance. The proxies of personalities we obtain yield significant (p<0.05) population correlation coefficients for traits against grade—0.846 for Extraversion and 0.319 for Conscientiousness. Furthermore, we demonstrate that a student’s e-behaviour and personality can be used with deep learning (LSTM) to predict and forecast whether a student is at risk of failing the year. Machine learning procedures followed in this report provide a methodology to timeously identify students who are likely to become at risk of poor academic performance. Using engineered online behaviour and personality features, we obtain a classification accuracy (κ) of students at risk of 0.51. Lastly, we show that we can design an intervention process using machine learning that supplements the existing performance analysis and intervention methods. The methodology presented in this article provides metrics that measure the factors that affect student performance and complement the existing performance evaluation and intervention systems in education.

Examining the Gendered Impacts of Technology-Facilitated Sexual Violence: A Mixed Methods Approach

This study employed a mixed methods approach, integrating quantitative online survey data (N = 333; Mage = 33.91; 63% women) with qualitative interview data (N = 10; Agerange = 24-46; 50% women) to gain a more comprehensive understanding of TFSV. We found that victims of TFSV experienced anxiety, stress, depression, loss of control, mistrust, multiple victimizations, poor academic/occupation functioning, problematic alcohol consumption, embarrassment, and online behaviour changes (e.g., limiting personal information online) due to TFSV victimization. Individuals who experienced online image-based abuse reported greater distress on items of depression, anxiety, and occupational/academic functioning than did victims of other types of TFSV. The current study provides partial support for the gender similarities hypothesis that TFSV is not exclusively a gender-based harm; our findings suggest that women and men’s TFSV experiences are similar for most TFSV types. Overall, the present study demonstrates the negative impact TFSV has for both women and men and highlights the need for greater awareness and increased support for all victims of this form of sexual violence.

Internet Gaming Disorder, Risky Online Behaviour, and Mental Health in Hong Kong Adolescents: The Beneficial Role of Psychological Resilience

In the present cyber age, Internet gaming disorder (IGD) and risky online behaviour are prevalent, and adolescents are especially vulnerable to such emergent problems. Few studies have explored the protective factors that mitigate harm caused by IGD and various common risky online behaviours. This study examined the prevalence of IGD and risky online behaviour, their hypothesised associations with depressive symptoms, and the beneficial role of psychological resilience as an underlying psychological mechanism. The participants included 1,099 Chinese junior secondary school students (33% boys, mean age = 13.5 years, age range = 10–17 years) who completed a battery of validated self-report questionnaires at their schools. The results revealed that 4% of the participants were at high risk of IGD and 6% were at an overall risk level of IGD. Depressive symptoms were positively associated with IGD and risky online behaviour, and psychological resilience mediated both of these associations. These results imply that clinicians and teachers should incorporate psychological resilience training into intervention approaches to mitigate IGD and risky online behaviour.

An exploration of trolling behaviours in Australian adolescents: an online survey (Preprint)

BACKGROUND Internet trolling (i.e., “trolling”) is an intentional, disruptive antisocial online behaviour, where an individual posts provocative and inflammatory content intended to distress and provoke their targets. Unique characteristics of trolling, such as meaningless disruption, distinguish the behaviour from cyberbullying. To understand why people “troll”, researchers have explored a range of individual differences including personality traits, social reward, and empathy. To date, these studies have primarily been conducted in adult samples. This is despite adolescents being highlighted as a particularly vulnerable group with regards to both experiencing and perpetrating trolling. Given the significant psychosocial impact of experiencing trolling, there is urgency to understand the experience of adolescents. Further, an understanding of why adolescents perpetrate trolling would inform development of effective management and prevention of the antisocial online behaviour. OBJECTIVE There are two primary objectives of the current study. First, we aimed to explore adolescents experience of trolling, by documenting how often they experience and perpetrate trolling and to explore the social media platforms on which this behaviour occurs. Second, we aimed to replicate adult research that has constructed a psychological profile of the Internet troll by exploring the utility of personality traits (i.e., psychopathy and sadism), self-esteem, empathy (cognitive and affective), and social rewards (negative social potency) and to predict perpetration of trolling in a sample of Australian adolescents. METHODS A sample of 209 Australian adolescents (59.1% male, 40.9% female, 0.5% non-binary) aged between 13 and 18 years of age (M = 15.87, SD = 1.60) completed the Adolescent Measure of Empathy and Sympathy (AMES), Rosenberg Self-Esteem Scale (RSES), Youth Psychopathy Traits Inventory – Short Version (YPI-S), Social Rewards Questionnaire (SRQ), and Short Sadistic Impulse Scale (SSIS). The experience of being trolled and perpetration of trolling was measured via a series of questions. RESULTS Results indicated 34.4% of Australian adolescents reported they had been targeting by trolling in the previous year and 18.2% reported they had perpetrated trolling in the previous year. Experiencing trolling was most likely to occur on Tumblr (44.8%) and Twitter (39.7%) and perpetrating trolling was most likely to occur on WordPress (30.4%) and Twitter (20.5%). Psychopathy, sadism, self-esteem, cognitive empathy, affective empathy, and negative social potency explained 42.4% of variance in adolescents’ perpetration of trolling (p<.001). High negative social potency (β=.15, p=.020), high sadism (β=.29, p=001), high affective psychopathy (β=.17, p=.033) and low cognitive empathy (β=.-.28, p=001) were predictive of trolling. Boys were more likely than girls to troll (β = -.11, p=041). CONCLUSIONS The findings indicate personality and psychological traits important to trolling in adults also play a significant role in perpetrating trolling in adolescence. Future research should continue to examine adolescent trolling behaviour to develop targeted interventions to prevent or manage trolling in adolescence.