This is a profile of Mozilla Technology Fund awardee CheckFirst.
Search engines are the online libraries through which we discover and research new topics, feed our curiosities or fetch sassy cat memes. But behind the ranking scenes of what content appears at the top, a silent, yet critical battle for engagement is taking place aided by recommendation algorithms—the battle for your attention.
These engagement battles for top results rankings can easily be manipulated by nefarious actors to disseminate disinformation, hawk or edge out particular stories from the public eye. This is a subject area that CheckFirst, a Finnish monitoring software company has been researching. Through their CrossOver project, they have been investigating disinformation peddled through algorithmic content recommendations on platforms such as Google News, Twitter, and YouTube in eight French-speaking countries and regions. CrossOver is a recipient of a 2023 Mozilla Technology Fund award, provided to projects auditing AI systems.
Working together with journalists, CrossOver monitors a set of keywords, observing what type of content is spat out as a response. ‘Although I wouldn’t say disinformation is always the default, it's presented to users early on into search results.’ Kuster remarks.
Although I wouldn’t say disinformation is always the default, it's presented to users early on in search results. Guillaume Kuster, CheckFirst
Indeed, after the EU banned Russia Today and Sputnik from broadcasting in the wake of the Russian invasion of Ukraine, CrossOver discovered that CGTN Français, a Chinese State-run media agency, had instead risen to popularity in Belgium. ‘It was as though they were stepping up to fill this void.’ Kuster remarks.
Their investigations used Youtube’s API and contrasted it against the data collected through their Raspberry Pis, disseminated in seven different Belgian provinces. Studying content variations when the keyword “Russie” was used, they observed a stream of algorithmically recommended content majorly sprouting from CGTN Français.
Although their research does not point to any intentionality to manipulate the algorithm, Kuster explains, ‘Any bad actor would try to play any system to promote their own interests and that existed before the internet and continues to exist. Disinformation didn’t wait for the internet to exist. The only difference is the ease and scale at which these pieces of content spread, which is a cause of concern.’
But coordinated attempts to ‘play the system’ can also occur tacitly. One such case is CrossOver’s investigations of Google News and its overly positive stories about Xinjiang province in China that seemed too good to be true, ‘a seizure to control the media landscape,’ Kuster claims.
Various Chinese state-controlled media outlets pumped and swamped troves of positive stories about Xinjiang province on Google News, overshadowing critical reporting about the Uyghur community where they were facing massive human rights violations.
Kuster also spoke about other research studies which CrossOver has investigated: Why ‘Nazi’ on Twitter seemed to be frequently trending in particular areas in Belgium and how Google’s autocomplete of the search keyword Donbas (a region in Ukraine currently occupied by Russia), is plagued by disinformation. This he says is an attempt to spearhead a pro-Kremlin narrative over the war in Ukraine. ‘We are gathering all this data and pointing these case studies to regulators for better legislative reforms. At the moment, both the EU AI Act and the Digital Services Act are at critical junctures. We also co-signed Mozilla’s open letter to DSA regulators on how to grant researchers’ data access.’
However, avenues to scrutinize platforms are facing convenient paralysis from parent companies, elevating the barriers to carrying out such public research. Now Reddit is planning to have its API behind a paywall, following Twitter’s recent pricey paywall push. ‘These developments make it difficult to collect the data points to compare between our web scraping data, and the monitoring data that we collect from APIs,’ Kuster concludes.
With the support from Mozilla Technology Fund, CrossOver is expanding its pilot project and adding monitoring devices to seven additional French-speaking countries around the world: Canada (Québec), the Democratic Republic of the Congo, France, Mali, Morocco, Senegal, and Switzerland (Romandie).
About the Mozilla Technology Fund
The Mozilla Technology Fund (MTF) supports open source technologists whose work furthers promising approaches to solving pressing internet health issues. The 2023 MTF cohort will focus on an emerging, under-resourced area of tech with a real opportunity for impact: auditing tools for AI systems.