Not every day do we get a response from a mathematician working on computer science. In response to my column, “ >Virus in the viral content ” (February, 23, 2015), Amitabha Bagchi, Associate Professor, Computer Science and Engineering, IIT Delhi, shared a study done by his team on rumour control strategies on social networks. What made his response unique was that it provided a mathematical model to tackle the malaise of rumour and reaffirmed faith in credible news ecology.
“Yes,” answered Prof. Bagchi to my question on whether a counter-offensive could work in the case of rumours and propaganda. He wrote: “A few years ago, a PhD student of mine and I did some work in this area. We modelled rumour as a message spreading through a social network (not necessarily an online social network). Our contribution was a new idea: an “anti-rumour” which was a message that circulated through the network and had the effect of being a vaccine, i.e., when the anti-rumour reached a person in the network, it erased the person's belief in the rumour, if such a belief existed, and prevented that person from believing in the rumour in the future, if the belief did not already exist. Through mathematical analysis and simulated experiments, we found that such an “anti-rumour” can effectively contain rumour, especially if enlightened citizens who are also part of the social network, work to spread such a message.”
Leveraging trust in friends He shared the research paper, “Towards Combating Rumors in Social Networks: Models & Metrics”, published in the journal, Intelligent Data Analysis . The paper was written by Rudra M. Tripathy and Amitabha Bagchi, both from IIT Delhi, along with Sameep Mehta from IBM Research-India. In this paper, they studied different methods for combating rumours in social networks actuated by the realisation that authoritarian methods to fight rumours have largely failed. Their major insight is that in situations where populations do not answer to the same authority, it is the trust that individuals place in their friends that must be leveraged to fight rumour. In other words, rumour is best combated by something, which acts like itself, a message that spreads from one individual to another. They called such messages anti-rumours. They studied three natural anti-rumour processes to counter the rumour. They proposed several metrics to capture the properties of rumour and anti-rumour processes. The metrics were geared to capture temporal evolution as well as global properties of the processes. They evaluated their methods by simulating rumour and anti-rumour processes on a large data set of around 10 to the power of five nodes derived from the social networking site, Twitter, and on a synthetic network of the same size generated according to the Barab´asi-Albert model.
They point out that the propagation of the anti-rumour does not depend primarily on the authoritativeness of the source that issues the anti-rumour but on the trust users place in their friends in the social network. Three models — the Delayed Start Model, the Beacon Model and the Neighbourhood Model — have different trajectories and impact on our societies. In the Delayed Start Model, the local authority discovers a rumour some days after it started and decided to combat it with an anti-rumour. In the Beacon Model, the researchers assume that the social network contains a set of vigilant agents, beacons that are on the lookout for the spread of rumours. “Once a beacon receives a rumour it immediately starts spreading anti- rumours to combat the rumour. This strategy corresponds to a semi-centralized scenario where coalitions of authorities may proactively decide to seed the network with vigilant users who can both detect rumours and respond to them,” they argue. They also point out that in the Beacon model, the initial set of beacons are chosen by some authority, whereas in the Neighbourhood model, the beacons are self created with some probability during the rumour spreading process.
I am not going into the mathematical details, the matrices used by the researchers, the research methodology, and the way Twitter data and synthetic data were used to arrive at some of the conclusions. In all the three models, they observed a sharp growth in the rumour process after a slow start. Using numbers and data, they show that once the growth of rumour starts to decline, within a very short span of time the rumour is completely removed from the network. Hence, they contend that once we detect the rumour — no matter in which way — due to fast growth power of social networks we can conquer the rumour.
Based on this study, Prof. Bagchi wanted The Hindu and other responsible news organisations to take the lead by allocating resources to combat harmful rumours. He wrote: “If an individual tweets that a rumour is not true, maybe his friends will not believe it, but if an individual retweets a tweet from The Hindu ’s official Twitter handle his friends will not only believe it, but the more enlightened amongst them will actually see it as their social duty to propagate this message.”
readerseditor@thehindu.co.in
