Misinformation during the Dutch covid pandemic: an informational and perceptual crisis?

In 2020, the WHO declared an ‘infodemic’ to describe the abundance of (mis)information during the covid pandemic, thus expressing its concerns about the effect misinformation had on individuals’ willingness to follow public health measures (World Health Organization, 2020). Attention to the infodemic among journalists and academics proliferated (Simon & Camargo, 2021), exemplifying the view that the world was not only suffering from a health crisis, but also a crisis of information (Hameleers & Brosius, 2022). However, in order to fully understand the consequences of misinformation, we should first examine to what extent misinformation was present during the pandemic. Concerns about misinformation may not only point to a crisis of information, but also a ‘crisis of perception’ (Hameleers & Brosius, 2022). Misinformation has become a salient issue; citizens across the world are strongly concerned about misinformation (Newman et al., 2021). Accusations of misinformation are used to discredit political opponents and news media (Egelhofer & Lecheler, 2019). Legacy media may also increase attention to misinformation by discussing it (Tsfati et al., 2020). The salience of misinformation may cause citizens to overestimate the amount of misinformation they encounter and raise general distrust towards journalism (Lecheler & Egelhofer, 2021). Concerns about misinformation are possibly especially strong during times of crisis (Hameleers & Brosius, 2022). Therefore, understanding to what extent misinformation was discussed is also important for understanding the role misinformation has played during the pandemic. A major methodological challenge has been detecting misinformation (Damstra et al., 2021). Misinformation is defined as “the inadvertent sharing of false information” (Melchior & Oliveira, 2022, p. 1505). However, since inferring the intentions of a message’s author is difficult (Damstra et al., 2021), we use misinformation to describe false information that is spread intentionally and unintentionally. Research (e.g. Paka et al., 2021) has shown that Machine Learning algorithms are able to classify misinformation on covid, reaching F1-scores of up to 0.95. However, no research has applied these algorithms to analyze the presence of misinformation during the pandemic. We contribute to the literature by using an ML algorithm to classify misinformation and references to misinformation in the Dutch public debate. Misinformation has been a salient issue in the Netherlands during the pandemic; both citizens (Hameleers & Brosius, 2022) and the government (e.g. Rijksoverheid, 2021) expressed strong concerns about misinformation. We aim to study both social and legacy media over the course of the pandemic (February 2020-February 2022). RQ1: How has the presence of misinformation in Dutch legacy and social media developed during the covid pandemic? RQ2: How has the presence of references to misinformation in Dutch legacy and social media developed during the covid pandemic? Methods We first train a BERT model to classify (references to) misinformation, topics and stylistic elements (e.g. incivility) among Dutch Tweets about covid and use this model to analyze all Dutch Tweets about covid in the specified time frame. In a second step, a BERT model is trained for classification of data from legacy media, particularly Dutch newspaper articles, and articles from alternative Dutch websites.