Combating misleading information, hoaxes, half-truths and outright lies is a discouraging business and can feel like a losing battle. However, many
researchers and media experts continue to keep up the fight. This past October a group of technologists, academics, and platform owners got together for the second meeting of the Conference for Truth and Trust Online to discuss the current situation, ways to better understand how and why the phenomenon happens as well as technical innovations to remedy the problem.
The program committee was chaired by Emiliano De Cristofaro (University College London) and Preslav Nakov (Qatar Computing Research Institute, HBKU). Not surprisingly, much of the content and focus of this year’s conference centered on the dangers of false information related to the current COVID-19 crisis and the U.S. presidential election, which took place just three weeks after the event. In times of a crisis, quality information is especially important, but that’s exactly when bad actors spread rumors and misleading claims. Nicole Aitken from Full Fact, a U.K. fact-checking organization, reported that in times of a crisis, people actively look for information trying to understand how their lives might be impacted. She added that the stress of a crisis also means that people are less able to process complex facts, making it harder to discern real information from the noise. And, unfortunately, real-world harms can result from false information online.
The first day’s keynote speaker Meeyoung Cha, Associate Professor at the Korea Advanced Institute of Science and Technology, cited several instances of the consequences of believing wrong information about the pandemic. Some of her examples were almost comical, such as a security guard at a U.K. office building asking people to hold their breath for ten seconds as a health screening before allowing them to enter the building. Others were more dire, such as an incident at a Korean church where church members believed they would be protected from the virus by spraying salt water into each other’s mouths, leading to a superspreading event at the church. Others are even tragic. Nigeria has seen a spike in hydroxichloroquine poisoning, for example.
Cha’s broader and more important message was about how false information spreads. Comparing the spreading patterns of reliable versus false information in social graphs reveals distinct differences. A key part of the fight against bad information is understanding how and why it is disseminated to help lead to effective remedies.
Understanding How Disinformation Spreads
Indeed understanding how mis- and disinformation spreads was a recurring theme throughout the conference. Claire Wardle, the U.S. Director of First Draft, a non-profit organization dedicated to fighting disinformation online, emphasized that false information is generally not broadcast to a large audience as in traditional media, at least not when it’s getting started. “One thing I really want to stress in this talk is that too many of us who work in the quality information space still think about this in a broadcast way.” She went on to explain how misinformation propagates through clustered networks of people. “Information flows between small peer-to-peer groups. People that are trusted take information, and they twist it into a narrative that fits with their community.” She believes that those working in the quality information space need to understand that information moves in different ways now, and the old techniques need to adapt.
“One thing I really want to stress in this talk is that too many of us who work in the quality information space still think about this in a broadcast way.”
As if anticipating Wardle’s message, a group of journalists from the BBC (Rebecca Skippage, Alistair Coleman, Olga Robinson, and Shayan Sardarizadeh) covered the ways their group had to quickly adapt their approach as news of the virus originating in Wuhan, China started to break. The parallel infodemic that spread along with the virus presented new challenges to traditional reporting. Having created a disinformation observation team as early as 2018, the BBC seems to have been better prepared for the infodemic than the world was for the coronavirus pandemic.
Since disinformation normally originates and starts to spread online, traditional media companies face a dilemma. A major concern for the BBC team when reporting on false information is not to amplify it so as not to bring more attention to misleading claims. Journalists need to understand how and where bad information has been spreading. The BBC team has settled on an approach of not bringing attention to misinformation unless it is already widely spread or has the potential to create real damage, such as harmful medical advice.
Mass media professionals generally try not to bring attention to bad information, but as disinformation moves through networks and jumps from cluster to cluster reaching larger and larger numbers of people, it can get laundered along the way, so that its apparent provenance is less dubious. Once it reaches celebrities and prominent people, it’s harder for the news media to ignore it. Moreover, in this networked environment, even obscure sites and sources can originate wild claims. Existing networks like QAnon will pick up bogus information and spread it throughout their existing connections causing it to go far beyond its original, sketchy origins.
Nearly all of the presenters focused on not only the problem of disinformation, but also potential remedies. Several papers and presentations covered various aspects of fact-checking and other ways to counter disinformation campaigns.
Meeyoung Cha’s keynote included a description of the successes her team has had with their “Facts Before Rumors” program. The program attempts to predict where false information is likely to spread and then precede it with reliable information, providing a kind of inoculation to those about to be infected with false claims. If credible voices can get their message out early enough they can stanch the flow of bad information. Their research showed that fact-checks must happen before people have really absorbed bad claims. Once they have accepted a claim, exposure to the fact-check is less likely to be effective. The group has created simple and interesting infographics that debunk misleading claims.
Of course, different remedies have shortcomings, and people are justifiably sensitive to interference with free expression. Satire and sarcasm are likely to be incorrectly flagged by automated verification, and fact-checking tools are likely to make mistakes. Whether automated or manual, fact-checkers have to be considered trustworthy if they are to have any positive impact. The Director of the International Fact-Checking Network, Baybars Örsek, described his organization’s attempts to establish more trust in fact-checkers by promoting higher standards in their work.
Perversely, false fact-checking has even been used as a method to spread disinformation. The Poynter Institute has noted that fake fact-checkers are one of the latest trends in online disinformation. Sarah Samwel from Carleton University, presented details of a website devoted to denying the Armenian genocide that took place in 1915. These fake sites try to signal their integrity by aping the content of legitimate fact-checking sites. They adopt similar language and imagery to hide their true motives.
Even when information is correctly labeled, consumers have different reactions to fact-checked content. Emily Saltz (Partnership on AI and First Draft) together with Claire Leibowicz found that attitudes about fact-checking labels vary. Much like the contentious issues they try to clarify, fact-checking labels have people divided. Some consider the labels helpful, providing context to an issue while others feel the labels are paternalistic, as if platforms are spoon feeding them information. They distrust the motives behind labeling.
TTO’s Origins and What’s to Come
Due to the pandemic, this year’s conference took place online, but that didn’t interfere with its goal of providing a single gathering for all kinds of people involved in misinformation. “Lots of computer science conferences have related workshops or side events, but we thought there was a need for a single conference focused entirely on online harm,” explained Guillaume Bouchard who is one of the originators of the TTO conference and also a co-founder of [CheckStep] “We wanted a conference that would be independent and that would bring together multiple academic disciplines to address the problems holistically with real solutions rather than just talking about the situation.”
Conference content was comprised of both technical papers and talks. The talks were designed to allow scholars, activists, developers, lawyers, ethics experts, fact-checkers, public servants, journalists, and all around researchers to discuss early ideas even before they’ve been fully fleshed out from their work in the field. The virtual conference was ‘packed’ with almost 400 participants from around the world, across thirty-five countries and five continents. Over fifty volunteers helped run the conference, which took place on October 16 and 17. In addition to the topics already mentioned, papers and presentations also covered bias, polarization, propaganda, and ways to model and classify text and analyze the news.
Next year’s conference is already in the works and [CheckStep]’s own Deputy Head of Research, Isabelle Augenstein, will co-chair the program committee. No doubt much will have changed in a year’s time, but we can expect that efforts to spread disinformation will continue. She’s hoping to see some progress by the time the group convenes again. “I hope by next year, we as a community will have more insights into how fact-checking can actually work in the real world — how we can use fact-checking effectively to warn people about mis- and disinformation, which includes changing peoples’ minds.”
We’ll be keeping an eye out as the planning shapes up looking forward to an equally informative and productive conference in 2021.