Facebook data reveal the devastating real-world harms caused by the spread of misinformation

Twenty-one years after Facebook’s launch, Australia’s top 25 news outlets now have a combined 27.6 million followers on the platform. They rely on Facebook’s reach more than ever, posting far more stories there than in the past.

With access to Meta’s Content Library (Meta is the owner of Facebook), our big data study analysed more than three million posts from 25 Australian news publishers. We wanted to understand how content is distributed, how audiences engage with news topics, and the nature of misinformation spread.

The study^[1] enabled us to track de-identified Facebook comments and take a closer look at examples of how misinformation spreads. These included cases about election integrity, the environment (floods) and health misinformation such as hydroxychloroquine promotion during the COVID pandemic.

The data reveal misinformation’s real-world impact: it isn’t just a digital issue, it’s linked to poor health outcomes, falling public trust, and significant societal harm.

Misinformation and hydroxychloroquine … and floods

Take the example of the false claims that antimalarial drug hydroxychloroquine was a viable COVID treatment.

In Australia, as in the United States, political figures and media played leading roles in the spread of this idea. Mining billionaire and then leader of the United Australia Party, Clive Palmer, actively promoted^[2] hydroxychloroquine as a COVID treatment. In March 2020 he announced he would fund trials^[3], manufacture, and stockpile the drug.

He placed a two-page advertisement in The Australian. Federal Coalition MPs Craig Kelly and George Christensen also championed^[4] hydroxychloroquine, coauthoring an open letter advocating its use.

We examined 7,000 public comments responding to 100 hydroxychloroquine posts from the selected media outlets during the pandemic. Contrary to concerns that public debate is siloed in echo chambers, we found robust online exchanges about the drug’s effectiveness in combating COVID.

Yet, despite fact-checking efforts, we find that facts alone fail to stop the spread of misinformation and conspiracy theories^[5] about hydroxychloroquine. This misinformation targeted not only the drug, but also the government, media and “big pharma”.

To put the real-world harm in perspective, public health studies estimate hydroxychloroquine use was linked to at least 17,000 deaths^[6] worldwide, though the true toll is likely higher.

The topic modelling also highlighted the personal toll caused by this misinformation spread. These include the secondary harm of the drug’s unavailability (due to stockpiling) for legitimate treatment of non-COVID conditions such as rheumatoid arthritis and lupus, leading to distress, frustration and worsening symptoms.

In other instances, we saw how misinformation can hurt public trust in institutions and non-government organisations. Following the 2022 floods in Queensland and New South Wales, we again saw that despite fact-checking efforts, misinformation about the Red Cross charity flourished online and was amplified by political commentary.

Without repeating the falsehoods here, the misinformation led to changes in some public donation behaviour such as buying gift cards for flood victims rather than trusting the Red Cross to distribute much-needed funds. This highlights the significant harm misinformation can inflict on public trust and disaster response efforts.

Misinformation ‘stickiness’

The data also reveal the cyclical nature of misinformation. We call this misinformation’s “stickiness”, because it reappears at regular intervals such as elections. In one example, electoral administrators were targeted with false accusations^[7] that polling officials rigged the election outcome by rubbing out votes marked with pencils.

While this is an old conspiracy theory about voter fraud that predates social media and it is also not unique to Australia, the data show misinformation’s persistence online during state and federal elections including the 2023 Voice referendum.

Here, multiple debunking efforts from electoral commissioners, fact-checkers, media and social media seem to have limited levels of public engagement compared to a noisy minority. When we examined 60,000 sentences on electoral topics from the past decade, we detected just 418 sentences from informed or official sources.

Again, high-profile figures such as Palmer^[8] have played a central role in circulating this misinformation. The chart below demonstrates its stickiness.

References

1. ^{^} The study (doi.org)
2. ^{^} promoted (reutersinstitute.politics.ox.ac.uk)
3. ^{^} fund trials (www.theguardian.com)
4. ^{^} championed (www.abc.net.au)
5. ^{^} conspiracy theories (doi.org)
6. ^{^} 17,000 deaths (www.valuehealthregionalissues.com)
7. ^{^} accusations (www.aec.gov.au)
8. ^{^} Palmer (www.smh.com.au)
9. ^{^} CC BY-SA (creativecommons.org)