Search for: All records

Anti-Asian prejudice increased during the COVID-19 pandemic, evidenced by a rise in physical attacks on individuals of Asian descent. Concurrently, as many governments enacted stay-at-home mandates, the spread of anti-Asian content increased in online spaces, including social media platforms such as Twitter. In the present study, we investigated temporal and geographic patterns in the prevalence of social media content relevant to anti-Asian prejudice within the U.S. and worldwide. Specifically, we used the Twitter Data Collection API to query over 13 million tweets posted during the first 15 months of the pandemic (i.e., from January 30, 2020 to April 30, 2021), for both negative (e.g., #kungflu) and positive (e.g., #stopAAPIhate) hashtags and keywords related to anti-Asian prejudice. Results of a range of exploratory and descriptive analyses offer novel insights. For instance, in the U.S., results from a burst analysis indicated that the prevalence of negative (anti-Asian) and positive (counter-hate) messages fluctuated over time in patterns that largely mirrored salient events relevant to COVID-19 (e.g., political tweets, highly-visible hate crimes targeting Asians). Other representative findings include geographic differences in the frequency of negative and positive keywords that shed light on the regions within the U.S. and the countries worldwide in which negative and positive messages were most frequent. Additional analyses revealed informative patterns in the prevalence of original tweets versus retweets, the co-occurrence of negative and positive content within a tweet, and fluctuations in content in relation to the number of new COVID-19 cases and reported COVID-related deaths. Together, 

Social media has revolutionized communication, allowing people worldwide to connect and interact instantly. However, it has also led to increases in cyberbullying, which poses a significant threat to children and adolescents globally, affecting their mental health and well-being. It is critical to accurately detect the roles of individuals involved in cyberbullying incidents to effectively address the issue on a large scale. This study explores the use of machine learning models to detect the roles involved in cyberbullying interactions. After examining the AMiCA dataset and addressing class imbalance issues, we evaluate the performance of various models built with four underlying LLMs (i.e. BERT, RoBERTa, T5, and GPT-2) for role detection. Our analysis shows that oversampling techniques help improve model performance. The best model, a fine-tuned RoBERTa using oversampled data, achieved an overall F1 score of 83.5%, increasing to 89.3% after applying a prediction threshold. The top-2 F1 score without thresholding was 95.7%. Our method outperforms previously proposed models. After investigating the per-class model performance and confidence scores, we show that the models perform well in classes with more samples and less contextual confusion (e.g. Bystander Other), but struggle with classes with fewer samples (e.g. Bystander Assistant) and more contextual ambiguity (e.g. Harasser and Victim). This work highlights current strengths and limitations in the development of accurate models with limited data and complex scenarios.