Fake news spreading and sentiment of Italians during the first COVID-19 lockdown

Abstract The SEBCOV study is an international project which involves several countries including Italy. As a part of the project, an online survey was conducted. The last optional question of the questionnaire is open and asks for additional comments: it received a particularly high response rate. We performed a sentiment analysis dividing the textual answers into three groups: positive, negative, or neutral content. We also noticed that there were several answers that contained fake news, thus we flagged them differently. Then, to evaluate the joint effect of other survey variables on the sentiment, we fit a Bayesian hierarchical model. Young respondents and those who had an income loss due to lockdown are more likely to write a negative answer, because they suffered more than others during the lockdown restrictions. The North of Italy was the area most affected by the COVID-19 pandemic during the first lockdown, thus those living in that area reported more frequently a negative sentiment. Finally, we model the fake news spreading with a logistic model. Respondents that are more likely to report fake news in the last optional questions are those who do not use traditional (TV, newspapers) and institutional media to inform themselves and those who self-judge themselves as very capable to detect fake news.