Data Voids and Propaganda
These results are a perfect example of what researchers danah boyd and Michael Golebiewski call data voids—situations where search terms lead to limited, nonexistent, or deeply problematic information. In a 2019 paper that analyzed the concept, boyd and Golebiewski explained that data voids emerge “when obscure search queries have few results associated with them, making them ripe for exploitation by media manipulators with ideological, economic, or political agendas.” Data voids can occur naturally or through intentional manipulation, and are products of multiple factors, including information availability and the inner logic of search engines. (Understanding search engine optimization, for example, can allow malicious actors to construct problematic content around a particular term.) The paper also identified several major types of data voids. One example is a breaking news data void, which occurs when a rapidly changing news event leads to scarce or low-quality results. This void is likely to be filled over time as journalists produce more content that search engines can index. Data voids can also occur through more targeted and long-term manipulation; individuals can generate problematic information about a term to the point of dominating the corresponding search results, and then amplify public curiosity about the term to lead users to the results. As boyd and Golebiewski explain, the term “crisis actor” became the focus of such manipulation in the aftermath of the 2012 mass shooting at Sandy Hook as conspiracy theorists generated content alleging that children and parents had been paid to act as victims of the massacre. Six years later, after a horrific shooting at Marjory Stoneman Douglas High School, the concept of crisis actors entered mainstream discourse, and people who searched the term encountered “a strategically optimized information ecosystem” that had been constructed for years.
As an example of a low-quality news situation, Fort Detrick highlights a few additional aspects of data voids. First, the years-long flow of disinformation narratives surrounding the term “Fort Detrick” highlights the hybrid approaches that information manipulators can take in exploiting data voids. The first seed of the Fort Detrick disinformation campaign was planted in March 2020, when Chinese Foreign Ministry spokesperson Zhao Lijian retweeted a since-deleted article from Global Research Canada, a conspiracy site with links to Russian disinformation operations, that suggested a link between the lab and the global outbreak. This opening salvo was quickly countered by quality reporting debunking Zhao’s claim, thus preventing a breaking news data void. But over the course of the year, as Chinese state media outlets and government officials continued to generate content about Fort Detrick, search results were increasingly dominated by Chinese disinformation narratives, as evidenced by Google trends results indicating that the most common search query related to “Fort Detrick” in 2020 was “Fort Detrick coronavirus.” Therefore, when China doubled down on Fort Detrick disinformation in May, they encountered an information environment saturated with their own narratives and a western media less interested in covering a topic that had already been thoroughly debunked.
This points to a second vulnerability manipulators exploit. While there is a first-mover advantage to filling a data void, search engines are built to prioritize the “freshest” or most recent content—thus advantaging actors who not only publish early but often. Information manipulators can therefore exploit the media’s fleeting attention span or, in the case of conspiracy narratives, their responsible impulse to not amplify far-fetched rumors, even in an attempt to supplant them with truth. When credible outlets did cover the Fort Detrick rumor, as was the case in August when BBC News published analysis of the disinformation campaign, the story gained a top spot in Google and Bing search results. However, one reliable story cannot fill a data void, nor can it neutralize its dangers for news consumers. As Chinese state outlets continue to publish a steady stream of conspiratorial articles about Fort Detrick, their content again dominates news results, exploiting search engines’ emphasis on freshness.
The freshness dimension of search results can also fragment data voids in other ways. In late September, some local news from Fort Detrick—about a wastewater flooding incident and the closure of a town gate—made the top of the first page of Google News and Bing News results, overtaking older CGTN and Global Times conspiratorial articles. Chinese state media outlets nevertheless kept their foothold on the first page of both engines, and if the emphasis on fresh content persists, a new wave of content from the Chinese state propaganda machine could easily dominate results again. Bing News results for “Fort Detrick” on September 28 betrayed another vulnerability that can be exploited to spread poor-quality information. On this day, the fifth news result was a MSN link that contained a republication of a Xinhua News Agency piece about Fort Detrick based on a Chinese diplomat’s assessment of Fort Detrick’s ties to COVID-19. The news aggregator often hosts Xinhua content, and in this instance, the hosted content was part of a state-driven propaganda campaign with dire public health risks. This example hints that merely de-ranking websites would not be a complete solution to data voids;even if a state propaganda outlet were de-ranked, its news aggregator contracts would still provide an avenue for gaming search engine results.