Human and Algorithms for Detecting and Counter-attacking Fake Medical News

Recently, an abundance of false or misleading medical information has been observed on the Web and particularly on social media, posing a considerable threat to public health while eroding trust in healthcare systems. 6 out of 10 people search for the cause of their medical condition online, and among those who found a diagnosis online, 35% does not visit a professional medical provider. The COVID-19 pandemic has exacerbated this problem by bringing forward an infodemic surrounding the coronavirus that spreads as quickly and deadly as the virus itself. As of Aug-2020, at least 800 people may have died while about 6000 people are hospitalized because of COVID-19-related misinformation. Excerpts taken out of context from the Swissmedic report regarding the PCR test have been shared and reshared in several Facebook posts leading to public distrust in the effectiveness of the test. Fake news involving alternative COVID-19 treatments have led people in Vietnam stocking on hydroxychloroquine depriving in-need patients of access to their life-saving drugs. Such incidents demonstrate the need for a novel framework to combat fake medical news that can not only detect them early and accurately using algorithmic models but also counter-attack their claims by leveraging human experts.While research on fake news has recently attracted considerable attention, research on fake medical news is still in its infancy, but it proves to be a formidable challenge. First, fake medical news poses the same challenge as general fake news as social networks with their open nature are perfect breeding grounds for fake news to be produced and propagated at an alarming rate. Second, fake medical news is more personal and highly opinionated, as it is directly connected to people's wellness. As a result, early detection is paramount in preventing the potential damage of fake medical news, as in some cases, it could be a matter of life and death. Third, in comparison with social or political fake news, fake medical news is not geographically-dependent, i.e., the same fake medical news can "infect" different populations having different socio-cultural backgrounds. This requires techniques to handle fake medical news to be multilingual. Fourth, fake medical news is usually a misinterpretation (either unintentional or malicious) of claims from scientific papers or scientists. Therefore, techniques to detect fake medical news need to be able to discern and deduce correct claims from misinterpretations.Existing techniques for detecting fake medical news in online social networks focus on building fully autonomous machine learning models. While these models excel at processing large amounts of data in a short time, they often require a huge amount of training data, and they lack a deep understanding of medical contexts. This prohibits them from detecting unseen emerging fake medical news. Human experts, on the other hand, can recognize new patterns but lack the ability to handle large amounts of data. An approach that handles this trade-off is to leverage algorithmic models for fast detection while using human experts for validating algorithmic detections and identifying unseen fake medical news. In this project, we form a positive feedback loop between human experts and algorithmic models where models assist humans in keeping up with large data while humans help models in identifying emerging fake medical news.This project aims to build a human-powered real-time system for detecting and counter-attacking fake medical news from large-scale dynamic social networks with multi-lingual, scalable, and cost-effective techniques. The system constantly monitors thousands of information diffusion channels and issues alarms on suspicious information waves. Human experts are employed to validate these emerging medical news, assisting algorithmic models to train themselves to become smarter and adaptive to unseen patterns of fake medical news. The specific aims of this project are:1) Construct a graph-based detection model capturing relationships among different actors by incorporating data from multiple sources;2) Build a seamless and cost-effective integration of human experts and algorithmic models in which the most suspicious emerging fake medical news is validated by human inputs;3) Design large-scale and dynamic protocols for timely warning of fake medical news, ensuring the system to work with high-volume and high-velocity data streams from social networks,4) Assemble a Web-based social network observatory to dispute fake medical claims and provide contextual information on fake medical news for users to disseminate, thereby, halting the spread of misinformation.This project is expected to have a significant scientific and real-world impact. From a scientific perspective, it designs and establishes a novel semi-automatic fake medical news detection model that smartly integrates human and machine. From a real-world perspective, the project results will be disseminated via various formats, including an online observatory dashboard that offers the general public not only facts and figures but also anecdotes and user stories to counter-attack claims in fake medical news. From an interdisciplinary standpoint, the research will have an immediate impact during the pandemic, while its long-lasting legacy would be in hundreds of domains that fake medical news has touched, such as political science, social science, and journalism.