Concerns Around Opposition to the Green Pass in Italy: Social Listening Analysis by Using a Mixed Methods Approach (Preprint)

Abstract

BACKGROUND The recent introduction of COVID-19 certificates in several countries, including the introduction of a European Green Pass, has been met with protests and concerns by a fraction of the population. In Italy, the Green Pass has been used as a nudging measure to incentivize vaccinations, since unvaccinated people are not allowed to enter restaurants and bars, museums, or stadiums. OBJECTIVE This study aims to understand and describe the concerns of anti-green pass individuals in Italy, the main arguments of discussion, and their characterization. METHODS We collected data from Telegram chats and analyzed with a mixed-methods approach the arguments and the concerns that were raised by the users. RESULTS Most individuals opposing the green pass share anti-vaccine views, but that doubts and concerns about vaccines are not often among the arguments raised to oppose the green pass. Instead, the discussion revolves around legal aspects and the definition of personal freedom. Further, we explain the nature of the dichotomy and similarity between anti-vaccine and anti-green pass discourse, and we discuss the ethical ramifications of our research, focusing on the use of Telegram chats as a social listening tool for public health. CONCLUSIONS A large fraction of anti-green pass individuals share anti-vaccine views. We suggest public health and political institutions to provide a legal explanation and a context for the use of the green pass, as well as to continue focusing on vaccine communication to inform hesitant individuals. Further work is needed to define a consensual ethical framework for social listening for public health.

AI model GPT-3 (dis)informs us better than humans

“…In line with previous research (21), we found that both respondents and GPT-3 were not able to distinguish whether a tweet was organic or synthetic (data on GPT-3's assessment are available in the study's repository) (15). It might be possible to develop specific training courses to improve humans' recognition of synthetic text, based on linguistic markers, grammatical structure, and syntax.…”

“…An effective and commonly used strategy involves impersonation. When GPT-3 declines to generate output that may breach content policies, users simply request it to impersonate a character, for which content policies apparently do not apply (22)(23)(24). With this approach, even more credible swathes of disinformation could be produced by first asking GPT-3 to generate fake profiles of people to impersonate and, in a second iteration, to generate tweets that these profiles could write.…”

“…The output is then exported as a .xlsx file containing three columns: the text, the reason for termination, and the category. All the codes, available in this study's preregistration repository, are organized in commented Jupyter lab notebooks for scrutiny and replication (15). The prompts and the output are available in the same repository.…”

“…Using Twitter's advanced search, we collected a random sample of recent organic tweets on the topics listed above, including both true and false tweets. The tweets are available in the study's repository (15).…”

“…We collected a set of real tweets written by users on the same topics and programmed a survey in which we asked respondents to classify whether randomly selected synthetic tweets (i.e., written by GPT-3) and organic tweets (i.e., written by humans) were true or false (i.e., whether they contained accurate information or disinformation) and whether they were written by a real Twitter user or by an AI. Note that this study has been preregistered on the Open Science Framework (OSF) (15), and we have conducted a power analysis based on the findings of a pilot study, as described in Materials and Methods.…”