Listen to the episode Science Writing as Storytelling (rebroadcast) – Ryan Kelly
Listen to the episode Science Writing as Storytelling (rebroadcast) – Ryan Kelly
Listen to the episode Cold War Ice Core Reveals Historic Glacial Melt – Andrew Christ
Listen to the episode DNA Evidence of Denisovan Interbreeding – João Teixeira
Listen to the episode The Dyatlov Pass incident – Alexander Puzrin
Listen to the episode Monkey Business – Jean-Baptiste “JB” Leca
Listen to the episode How Mosquitoes Target Us – Zhilei Zhao & Lindy McBride
Listen to the episode Epistemic Puzzles in ‘The Witness’ – Luke Cuddy
Listen to the episode Unintended Consequences of Legal Reforms – Ángela Zorro Medina
Listen to the episode Bots’ Meddling in the 2020 Presidential Election – Emilio Ferrara
0
The fields of social media, machine learning, and artificial intelligence are moving rapidly, making a challenge of keeping up with their advances. So we asked Soroush what unifying idea he thinks we should pay more attention to.
Often invisibly, computational algorithms mine and analyze our interactions with technology. While the conveniences afforded by such technologies as speech recognition and computer vision through traffic navigation and medical diagnosis provide many benefits, they can also expose us to privacy risks … and even threaten our civil liberties. Given the pace of their growth, we wanted to hear what implications Soroush predicts that such technologies will have in our future.
Publicity and press coverage of Soroush's study has been ubiquitous since its publication in early March of 2018. As of the release of this episode of Parsing Science, it ranks in the top 5% of all research publications that are monitored by Altmetric, a company which tracks the online popularity of published research, which also lists it as the second highest scoring article published ever by the journal Science over its 138-year history. We asked Soroush about his experience of all the media attention the study has received.
The widespread use of social media seems to have accelerated our capacity to isolate ourselves from others whose opinions and perspectives differ from our own. This amplification and reinforcement of media that aligns with our established beliefs and opinions has been dubbed the "echo chamber effect." We asked Soroush for his perspective into this phenomenon.
Soroush's earlier work involved creating a tool for detecting false rumors on social media platforms, and this project extended that research by examining how false and true rumors are shared. Given his expertise, Doug and I wanted to learn where his interests might take him next.
We asked Soroush for his thoughts on how the spread of false news and rumors might be best investigated, as well as what courses of action he hopes that social media platforms might adopt to mitigate them.
Unlike Facebook, tweets posted to Twitter are almost always public. The analytics data that Soroush and his team gained access to, however, are not. Since we spoke to him just as details of Cambridge Analytica's misuse of private Facebook data were first emerging, Doug and I were especially interested in learning what led Twitter to share every tweet ever made with Soroush's team.
Though it wasn’t the focus of their study, Ryan and I were interested in learning if people may have different motivations for sharing false news online, as well as what might help persuade them to reconsider doing so. Here, Soroush gives his thoughts on the question.
To learn whether there might be any linguistic clues that could explain how false rumors propogate, Soroush and his team examined the emotional content of replies to true and false news tweets. Here, he explains how they explored the ways in which people respond to both types of news.
Their findings showed that false news spreads farther, faster, further and more broadly online than does the truth. Furthermore, the algorithm Soroush trained could also predict whether other online rumors were true or false 75% of the time … even before those tweets were fact-checked. Ryan and I were curious to learn which characteristics that Soroush and his team looked at were most influential in the spread of false news online.
With social media, especially on Twitter, information is shared from one or more users to multiple other users. When such information is reshared, a “cascade” is formed. Soroush and his team selected over 126,000 tweets involving contested news stories which linked to one of six professional fact-checking websites, and could therefore be verified as true or false claims. They then examined four characteristics inherent to cascades, as he describes next.
Ryan and I began our conversation with Soroush by asking what led to his interest in studying the spread of false news and rumors online.