Instagram's algorithm ‘recommending’ antisemitic imagery - The Jewish Chronicle

Instagram is ‘recommending’ antisemitic imagery, QAnon conspiracies and Covid-19 misinformation to millions of users as a result of a newly introduced feature that automatically sends them new, unsolicited content to digest.

In a bid to extend audience engagement and boost advertising revenue, last August the social media giant introduced new “Suggested Posts” and “Explore” features which used algorithms to direct users to new material.

But an investigation conducted by the Centre for Countering Digital Hate (CCDH) into this development found that users who had previously followed leading anti-vaxxer accounts were receiving ‘recommendations’ for openly antisemitic material.

In one example from the study, published on Tuesday, four volunteers set up Instagram accounts and followed ten of the best-known users posting anti-vaxx material on the site.

The accounts – which have 6.5 million followers on the platform – were selected because they appeared to offer relatively professional advice, and some had been given verified status by Instagram.

They included accounts such as “Physicians for Informed Content”, “Children’s Health Defence” and “Robert F Kennedy Jnr”.

Damningly, the study found that of 30 ‘recommendations’ put out by Instagram’s Explore page between September and October last year, 19 contained what was deemed to be “misinformation and hatred” by the CCDH.

One used the Israeli flag to imply that Jews control the government; while the second, from a different account, used the flag to link Israel to mask-wearing, a “Plannedemic” and the “NWO”, a supposed “New World Order” directing global events.

Volunteers following this list were also recommended a range of Covid misinformation.

Numerous posts claimed that “there is no pandemic”, while other examples urged people to “stop getting tested” and “stop wearing your mask”. Others mixed Covid denial with grand conspiracy theories, with one post claiming “there is no pandemic, this is the launch of the New World Order”.

Imran Ahmed, CEO of the Centre for Countering Digital Hate, said: “It is beyond belief that as the pandemic swept the world, Instagram launched a new feature encouraging users to view conspiracy theories and lies about Covid and vaccines. This feature was created in the name of profit, to keep people scrolling so more adverts could be served to them.

"Algorithms that recommend content are the act of a publisher, making choices as to what readers see, not a neutral platform. This has serious legal and regulatory implications for social media companies and shows their liability for damage to individuals and society.

“Instagram’s actions will have created greater distrust in the Covid vaccine, costing lives and prolonging the pandemic for all of us.

“2020 proved that online lies have a cost paid in real lives. Instagram must suspend its suggested posts feature until it can be sure it is not promoting misinformation, hate and lies. They also need to rapidly remove the material - much of which contravened their community standards.”

The study also noted how a number of the posts deemed to be promoting hateful content were presented next to adverts, highlighting Instagram’s use of algorithmic ‘recommendations’ to increase the time users spend on the platform and boost ad revenues.

The volunteers also followed influential accounts linked to “wellness” and “alternative health” promotion. Again they found they were directed by Instagram’s algorithm to misleading or controversial content.

In the second phase of the investigation, which continued up until last November , volunteers turned their attention to Instagram’s “Suggested Posts” feature.

Instagram, which was acquired by Facebook in 2012, announced its “Suggested Posts” feature in August 2020, saying it was “a new way to see content you may like.”

The CCDH study, which was co-published and reaerched with the young persons development agency Restless Developement, showed that as a result of this feature, Instagram users who showed a previous interest in far-right accounts - including those backing the violent “accelerationist” ideology, which aims to bring about civil war - were ‘recommended’ new content around Covid misinformation.

The ‘Suggested Posts’ received claimed that Covid tests were “inherently faulty”, that the pandemic was being used to impose “medical authoritarianism” and that trials of the Pfizer vaccine were approved on false data.

Those choosing to follow accounts promoting QAnon conspiracy theories were also sent further material including the claim Covid-19 was a “plandemic”.

The CCDH report suggested Instagram extended the reach of its algorithm to publish “Suggested Posts” to every user in the middle of a pandemic, without doing the necessary work to ensure it would not promote dangerous misinformation about Covid or vaccines.

This was despite Facebook’s promise, made in March 2019, that it “won’t show or recommend content that contains misinformation about vaccinations on Instagram Explore or hashtag pages” first made in March 2019.

The report claims Instagram should only reinstate its algorithmic publishing features when they are proven to be safe.

This should involve a new assessment of the safety of the “Suggested Posts” feature in the light of Facebook’s decision to limit its use of such features just two years ago because of their impact on the mental health of users.

In addition to taking measures to prevent its algorithm from publishing misinformation, the CCDH research suggests that Instagram can do much more to reduce the impact and spread of misinformation on its platform.

Responding to the claims made in the CCDH Instagram study, a Facebook company spokesperson said: “We share the goal of reducing the spread of misinformation, but this research is 5 months out of date and uses an extremely small sample size of just 104 posts.

“This is in stark contrast to the 12 million pieces of harmful misinformation related to vaccines and COVID-19 we’ve removed from Facebook and Instagram since the start of the pandemic.

“We’ve been focused on connecting people to credible information, which is why, when people search for content related to COVID-19 and vaccines on our apps, we direct them to authoritative health organisations, and have now directed people to the NHS and Government websites more than 10 million times. We’re also working on improvements to Instagram Search, to make accounts that discourage vaccines harder to find.”

Facebook sources also pointed to updated policies to account for certain kinds of implicit hate speech, including removing harmful stereotypes about Jewish people.

They also said the methodology used in the CCDH report was “flawed” and that it does not produce a representative sample and does not factor in content the company have proactively identified using AI and other automated technologies.