Periodic Reporting for period 1 - ION (Information Diffusion on Networks)
Berichtszeitraum: 2019-02-01 bis 2021-01-31
The aim of this two-year research project has been to analyse the simultaneous diffusion of multiple pieces of information on (social) networks; how messages may interact, how rumours diffuse in the presence of truthful information and verification, and how, inversely, the presence of rumors affects the diffusion of truthful information. Misinformation can have severe consequences, such as the debunked myth of a link between vaccinations and autism or AIDS denialism. Political examples also abound. Most recently, the COVID19 pandemic and the various pieces of misinformation that have circulated around its severity and the effectiveness/dangers of vaccinations have highlighted just how crucial a problem is misinformation for society at large. The rise in communication through online networks is often quoted as contributing to an increased spread of rumours and misinformation. Yet, in truth the state of the art has had little to say about how alternative pieces of information interact on networks, and even less why rumours might be propagated by individuals without prior verification, even if they aim to know and spread the truth. Without an understanding of what factors might influence the diffusion of rumors, policy makers will not be able to design appropriate measures to counteract their diffusion. Therefore, this project has aimed to build rigorous economic models which can shed light on the processes and decisions involved in the diffusion of rumors, thus addressing the tangible threat that the spread of misinformation poses to society. The main outcome are three academic papers which analyse various aspects of information diffusion.
To analyse the diffusion of (mis)information on social networks, this project has spawned three academic papers that look at different aspects of information diffusion.
The first paper, "Diffusion of Multiple Information: On Information Resilience and the Power of Segregation", was published in the journal "Games and Economic Behavior" in 2019, roughly 8 months after the beginning of the project. In this paper, I analysed how two independent pieces of information interact on a social network. Individuals prefer one type of information over another, and if they communicate, constraints in communication time imply that they only share the information they prefer. I show that in general, this leads to crowding out of information, and that the final prevalence of each information depends crucially on the proportion in the population who prefers it. Furthermore, I show that segregation according to preferences implies that each individual is more likely to be informed of their preferred information, but that this leads to polarization (nobody is aware of both pieces of information) and an overall loss of information (fewer people are informed overall).
The second paper, "Debunking Rumors in Networks" (joint with Luca P. Merlino and Paolo Pin), has been conditionally accepted at the journal "AEJ: Micro" roughly a month after the end of the fellowship. I have presented it at various conferences to positive reception. In the paper, we show that even when information is verifiable and individuals are all better off knowing the truth, biases towards specific worldviews make it possible for rumors to survive in the population. We also show that communication intensity does exacerbate the prevalence of the rumor in absolute terms, but not relative to the prevalence of the truth. Furthermore, homophily, the tendency of people to interact predominantly with others who are similar to them, which may cause "echo-chambers", has intricate effects on relative rumor prevalence, and may in fact decrease it. This implies that two of the most often cited channels through which online communication is said to increase rumors may in fact not harm the truth to rumor ratio of information in the population after all. We show that the best policy to decrease both absolute and relative rumor prevalence is to incentivise individuals to verify information themselves.
Finally, the third paper, "Optimal Inspection of Rumors in Networks" (joint with Luca P. Merlino), looks at the decision a social planner would take, if they were able to choose what proportion of the population was verifying messages they receive. The idea is that policy makers may influence verification rates of information by setting policies such as, e.g. the importance of information literacy in the national curriculum. As expected, increases in verification rates reduce rumor prevalence, and high enough verification is able to fully eradicate a rumor. However, we show that in a scenario in which individuals are biased towards a certain worldview, and in which they disregard unverified information that opposes this worldview, the existence of rumors may in fact be beneficial for the prevalence of truthful information: some people become aware of the truth after verifying a rumor they heard. This implies that a policy makers whose aim it is to maximise the prevalence of truthful information may allow a rumor to circulate by setting verification rates below those that would fully eradicate the rumor. This paper has been published as a working paper, and we presently prepare it for a journal submission.
The first paper, "Diffusion of Multiple Information: On Information Resilience and the Power of Segregation", was published in the journal "Games and Economic Behavior" in 2019, roughly 8 months after the beginning of the project. In this paper, I analysed how two independent pieces of information interact on a social network. Individuals prefer one type of information over another, and if they communicate, constraints in communication time imply that they only share the information they prefer. I show that in general, this leads to crowding out of information, and that the final prevalence of each information depends crucially on the proportion in the population who prefers it. Furthermore, I show that segregation according to preferences implies that each individual is more likely to be informed of their preferred information, but that this leads to polarization (nobody is aware of both pieces of information) and an overall loss of information (fewer people are informed overall).
The second paper, "Debunking Rumors in Networks" (joint with Luca P. Merlino and Paolo Pin), has been conditionally accepted at the journal "AEJ: Micro" roughly a month after the end of the fellowship. I have presented it at various conferences to positive reception. In the paper, we show that even when information is verifiable and individuals are all better off knowing the truth, biases towards specific worldviews make it possible for rumors to survive in the population. We also show that communication intensity does exacerbate the prevalence of the rumor in absolute terms, but not relative to the prevalence of the truth. Furthermore, homophily, the tendency of people to interact predominantly with others who are similar to them, which may cause "echo-chambers", has intricate effects on relative rumor prevalence, and may in fact decrease it. This implies that two of the most often cited channels through which online communication is said to increase rumors may in fact not harm the truth to rumor ratio of information in the population after all. We show that the best policy to decrease both absolute and relative rumor prevalence is to incentivise individuals to verify information themselves.
Finally, the third paper, "Optimal Inspection of Rumors in Networks" (joint with Luca P. Merlino), looks at the decision a social planner would take, if they were able to choose what proportion of the population was verifying messages they receive. The idea is that policy makers may influence verification rates of information by setting policies such as, e.g. the importance of information literacy in the national curriculum. As expected, increases in verification rates reduce rumor prevalence, and high enough verification is able to fully eradicate a rumor. However, we show that in a scenario in which individuals are biased towards a certain worldview, and in which they disregard unverified information that opposes this worldview, the existence of rumors may in fact be beneficial for the prevalence of truthful information: some people become aware of the truth after verifying a rumor they heard. This implies that a policy makers whose aim it is to maximise the prevalence of truthful information may allow a rumor to circulate by setting verification rates below those that would fully eradicate the rumor. This paper has been published as a working paper, and we presently prepare it for a journal submission.
The research of this project has pushed the modelling frontier in a way that allows the inclusion of optimising behaviour in the otherwise well-developed SIS (Susceptible-Infected-Susceptible) framework. This combination means that the resulting model is tractable, while is also allows meaningful questions to be asked (and answered) about how individuals may react to the diffusion of rumors.
The expected results are the publications of three papers, (i), "Diffusion of Multiple Information: On Information Resilience and the Power of Segregation", (ii), "Debunking Rumors in Networks", and (iii) "Optimal Inspection of Rumors in Networks". The first of these has been published during the duration of the project, the second has been accepted for publication about a month after the end of the project, and the third was published as a working paper about two months after the end of the project. This last paper will be submitted for publication soon after the end of the project.
Direct impact so far has been predominantly on the research community, by the innovation in the modelling frontier. It is hoped that wider socio-economic impact and societal implications will follow through an improved understanding of how rumors may propagate, and how best to fight them. The results of this project highlight indicate that no other policy variable may be as successful at reducing rumors as private verification of information by individuals. It is this area that the project proposes policy should be focused on.
The expected results are the publications of three papers, (i), "Diffusion of Multiple Information: On Information Resilience and the Power of Segregation", (ii), "Debunking Rumors in Networks", and (iii) "Optimal Inspection of Rumors in Networks". The first of these has been published during the duration of the project, the second has been accepted for publication about a month after the end of the project, and the third was published as a working paper about two months after the end of the project. This last paper will be submitted for publication soon after the end of the project.
Direct impact so far has been predominantly on the research community, by the innovation in the modelling frontier. It is hoped that wider socio-economic impact and societal implications will follow through an improved understanding of how rumors may propagate, and how best to fight them. The results of this project highlight indicate that no other policy variable may be as successful at reducing rumors as private verification of information by individuals. It is this area that the project proposes policy should be focused on.