Effects of #coronavirus content moderation on misinformation and anti-Asian hate on Instagram

Abstract

This study evaluated the intended and unintended effects of Instagram’s content moderation on #coronavirus for both the short- and long-term effects on misinformation and anti-Asian sentiment. We performed manual coding of images ( N = 9648), and a series of supervised machine learning methods to classify three waves of comments ( N = 22,676) published in 2020 on Instagram. Welch’s F tests were used to compare misinformation, emotions, toxicity, and identity attack across three time periods. The results showed that hashtag moderation had an intended effect in reducing misinformation, and an unintended effect in reducing anger, fear, toxicity, and identity attack. Images with people of East Asian descent were associated with more anger, fear, toxicity, and identity attack than images with people of other races. Prior to content moderation, misinformation was associated with identity attack. Stigmatization on social media, and content moderation of misinformation and hate speech are discussed.

Health activism, vaccine, and mpox discourse: BERTopic based mixed-method analyses of tweets from sexual minority men and gender diverse (SMMGD) individuals in the U.S.

“…Our findings provide implications for health communication tackling mpox and future disease outbreaks with an emphasis on fairness and equity. Corroborating previous research that found links between HIV and homophobia, 6 and between COVID-19 and anti-Asian racism, 5 this study found that mpox is not only linked to homophobia but also health activism. Our geographic analyses suggest that health activism was more likely to be motivated by oppression rather than accepting social climate in the U.S. at a census state level; future research can continue to test our geographic finding.…”

“…The preprocessed posts were then passed to the BERTopic model following six steps in order: (1) transforming documents into numerical representations using all-MiniLM-L6-v2, a sentence-transformers model capable of capturing semantic similarity between documents; (2) using UMAP 20 to reduce the dimensionality of input embeddings, preparing them for topic clustering; (3) applying HDBSCAN, 21 a hierarchical clustering algorithm, to find the natural groupings or topic structure of the documents; (4) implementing sklearn's CountVectorizer to convert the collection of documents to a matrix of token counts, namely, a bag-of-words representation; (5) adding to the BERTopic model a c-TF-IDF representation with additional BM-25 weighting to reduce frequent words in topic generation; and (6) fine-tuning the BERTopic representation by adjusting hyperparameters including n_neighbors, n_components, min_dist, min_cluster_size, and min_samples.…”

“…This occurred during the COVID-19 pandemic in the U.S., where hate crimes and violence towards people of East Asian descent surged due to perceptions attributing the pandemic to China. 5 Similarly, sexual minority men and gender diverse (SMMGD) individuals have been battling with the so-called “gay disease” stigma attached to HIV/AIDS since the 1980s when the infection and death cases were initially reported in North America to be prevalent among SMMGD. 6 The recent mpox outbreak, coupled with rising misinformation, 7 stigma, 8,9 and conspiracy theories, 10 may, like HIV, experience a stigmatization process that leads to delayed care-seeking and further marginalization of SMMGD individuals.…”