A flaw identified in Facebook’s ranking algorithm caused users to receive fake news and other harmful content directly in their feeds for months.
According to The Vergewho had access to internal company documents, the error was described by big tech officials as a “massive classification failure” that went uncorrected for months and affected “at least half of all News Feed views.”
During the period in which the flaw was active, “Facebook’s systems were unable to adequately suppress [conteúdos de] nudity, violence and even Russian state media that the social network recently promised to stop recommending in response to the invasion of Ukraine. That way, instead of preventing content from circulating through the platform, the algorithm allowed it to circulate.
The flaw was identified when experts at the company noticed last October that an increasing number of inappropriate content began to appear unexpectedly in the social network’s news feed – the increase in such posts was 30% globally.
The issue, which came to be categorized as “serious”, was corrected in early March when the company’s cause experts were able to identify the root cause.
The incident was confirmed by Meta spokesperson Joe Osborne at The Verge. The executive said the company “detected rating inconsistencies on five separate occasions, which correlated with small, temporary increases in internal metrics.”
Osborne also stated that “we tracked down the root cause of the software bug and applied the necessary fixes” and that the flaw “had no significant, long-term impact on our metrics”.
The Facebook Algorithm and the (Lack of) Transparency
The not always clear way on how the company’s algorithms work has already been the subject of discussions and was even one of the questions raised by Frances Haugen in the Facebook Papers.
The former employee of the company – responsible for leaking information from Facebook that exposed flaws in the security of users – argued that it is problems like these that highlight the need for the company to make its algorithms transparent and open, so that external researchers and independent experts can understand its operation and contribute to the evolution of technology.