What is the context behind your study ‘The COVID-19 social media infodemic’?
In 2014 a branch of research was born that was dedicated to studying the dissemination of information, particularly false information. One year earlier, during the World Economic Forum, this topic had already emerged as one of the most critical problems for contemporary society. Until then, the topic had been addressed through analogy with epidemiological models. This was based on the suggestion that information spreads among people in the same way as a virus.
The epidemiological models applied to this field were significantly limited by their complete lack of empirical evidence. Everything changed when data from social networks became available. In 2016, new data sources enabled us to publish a study that changed the definitions of the models used to study the spread of information.
We identified one of the fundamental reasons that made the standard epidemiological models inadequate for mapping information dissemination processes: individual choice. The spread of a virus occurs independently of people’s will. The diffusion of information, however, happens by choice. We choose information. We do not choose a viral infection. This result made us reconsider the mathematical premises for studying these processes.
My colleagues and I have produced consistent scientific output on the basis of this assumption. Motivated by this new knowledge, we wanted to verify whether some of the theories on information dynamics were well founded. In particular, our goal was to assess whether it is true that fake news travels faster than other news.
The current pandemic is the ‘perfect storm’ for spreading false information because it is a new, ambiguous, and little-known topic. We have tried to understand how the information dynamics related to COVID-19 are developing on five different social platforms. This type of comparative analysis of five different social networks during a critical event is unprecedented.
What conclusions have you reached?
Not only have we estimated the diffusion process by using models that imply specific growth mechanisms, but also by using phenomenological models that emphasise the reproducibility of empirical data. Our findings show that these models actually do not capture the so called echo chamber process [when similar information, ideas, or beliefs are amplified by repetitive transmission within a homogeneous and closed environment, resulting in a lack of consideration for divergent visions and interpretations A/N]. Only the number of people ‘infected’ with false information can be observed, but not ‘by what’ they were infected.
Since a 2018 study claimed that false information spreads faster than real information, we have been comparing five social media platforms and have observed that pieces of information marked as either reliable or as questionable do not present significant differences in their spreading patterns. In other words, no substantial difference exists between how false and reliable news tend to spread.
Can we infer that ‘fake and reliable news spread with the same speed’?
The speed at which information spreads depends on the audience and the specifics of each social media platform. Of course, echo chambers play a role in this process. In our most recent study, published in Proceedings of the National Academy of Sciences of the United States of America, we explored the key differences between leading social media platforms and investigated how likely they are to influence information spreading and the formation of echo chambers.
We performed a comparative analysis on more than 100 million pieces of content concerning controversial topics from Gab, Facebook, Reddit, and Twitter, focusing on two specific aspects: homophily in the interaction networks, and bias in information diffusion between like-minded peers.
We found significant differences across platforms in terms of homophilic patterns in the network structure and we observed that the aggregation in homophilic clusters of users dominates online dynamics.
We concluded that dissemination is a condition that varies according to the target group. Information spread is driven by the interaction paradigm imposed by the specific social media and by the specific interaction patterns of groups of users engaged with the topic.
Algorithms and platforms change with time. How do you deal with the changing rules of the game?
We take this into account and use variables that are independent from the algorithmic arrangement. Right now, we are studying how the Facebook algorithm, which is changing over time, is different from the Reddit algorithm, to draw some conclusions on the different social dynamic between the two platforms in terms of information consumption. On Facebook, there is very strong polarisation that does not appear on Reddit.
Studies like yours make me think of memetics, which has been an underdeveloped discipline for many years. Memetics is the study of how culture evolves through the creation, selection, and replication or transmission of information patterns or memes – ideas, beliefs, theories, and other types of mental constructs. Data from social networks seem to have given new possibilities to deepen this field. Is there new life for memetics?
Hopefully, as memetics is a new language that has been revived through social media. It will be necessary to understand how to deal with it quantitatively because an interpretative approach does not make the grade, even though we keep on talking about it in a widespread way, which is the problem in itself. We observe the diffusion process, but we still cannot go back to the origin, to point zero. We observe the mass dynamics, we see the intensification of infections, and we see that true and false information circulates in the same way with no distinction. All of this is consistent with our previous models, highlighting how echo chambers work and how news spreads within them.
Is the goal of your studies to convey information (true or false) as effectively as possible?
Yes, that is correct. We also want to understand how algorithms play a role in the fragmentation of public opinion. We believe that understanding the social dynamics behind content consumption and social media is essential to face future global challenges. A proper understanding of the dynamics of dissemination of information will allow us to implement more efficient communication strategies in times of crisis.
What kind of interest have governments shown in your research?
It depends on the context. The World Economic Forum values our studies a lot. We bring evidence of how information contexts shape groups and tribes and how the information circulating within those tribes is consistent with their narrative, also called their ‘totem’.
Our model relativises the concept of source because the source is chosen for a fundamental reason, regardless of its reliability. It is a challenging concept for people like journalists who believe that the source is the vital factor in explaining specific dynamics. Well, it is not. These are difficult topics to define correctly. When I coordinated the Italian task force on ‘Web data and socio-economic impact’, I pointed out that it is necessary to be particularly careful with communication because in times of extreme uncertainty, encouraging tribalism only risks creating disconnection and increasing distrust in institutions.