Last month, the Australian government began preliminary investigations into potential new laws that would force social media platforms to provide users with an option to switch off algorithmic recommendations, and set nonalgorithmic feeds as the default. The end result would be that users who choose to opt out of algorithmic feeds would only see updates from people and profiles they’ve explicitly chosen to follow.
Australia’s Communications Minister Annika Wells said in a recent address to the National Press Club that algorithmic incentives don’t necessarily align with what’s good or healthy for users, meaning that engagement-focused algorithms will often amplify harmful content to keep people scrolling.
“Platforms aren't passive,” Wells said. “They decide what is amplified, what captures us, what behaviors get rewarded,” Wells said. “They have been optimized for engagement and for attachment, not for well-being. Because the longer we stay, the more valuable we are.”
Requiring platforms to provide an algorithm opt-out could limit over-exposure to potentially harmful content, and it could also prevent users from being inundated with material that doesn’t benefit them or their communities.
“If someone wants their feed filled with friends and creators they follow, the social media platform must give them that choice,” Wells said. “And if you're happy to stick with your algorithm, you can do that, too. While it's true some platforms offer these choices right now, they do their best to make them unappealing and difficult to use. What changes under the Digital Duty of Care is that platforms must respect your choice. If you don't want to be thrown into an algorithm filled with people you don't follow or subscribe to, that is your choice and they must stick to it, not hide it, and not make you set it every time you open the app.”
The concept is supported by a growing body of academic research:
- A study published by The National Library of Medicine showed that social media algorithms designed to maximize engagement and screen time can lead to the accelerated development of addictive behaviors, particularly among younger users. Those behaviors are subsequently linked to depression and other detrimental health effects.
- Research published by the University of Texas in August showed that personalized feeds create a negative feedback loop, which can result in depressed users being shown more depressing content.
- Another 2025 study, published by the International Journal of Innovative Technologies in Social Science, outlined how AI-driven social media feeds alter dopamine pathways, mimicking substance addiction.
Clearly, there are established links between social media algorithms and harmful impacts on users. As such, Australia’s proposal to enable an algorithm opt-out makes sense, and could have a significant impact on mental health while also reducing social media usage time overall.
Though the platforms themselves are keen to avoid that outcome.
Instagram Chief Adam Mosseri responded to a question last month about the potential of removing engagement algorithms, and said that if engagement-aligned algorithms were removed it would lead to a worse result for creators.
As per Mosseri: “What happens [without algorithms] is people become a bit less interested in the experience and they use [Instagram] a lot less, which then means that there’s less reach to go around and people start to get more and more mad about the fact that their reach is going down because the overall pie gets smaller.”
Although his response was not specifically linked to Australia’s proposal, Mosseri reiterated this perspective over the weekend. In a post on Threads, he said: “Every major platform has had to lean into recommendations — posts from accounts you don't happen to follow — to stay relevant. Going back to a following-only feed would quickly mean a smaller Instagram, with less reach to go around for everyone.”
Mosseri’s responses seem to indicate that he believes highlighting reduced usage of Instagram and reduced feed value is something that the public will see as a negative. Perhaps Mosseri believes this will help to drive opposition against the proposals to ban algorithmic amplification.
However, based on the research, banning algorithmic feeds seems like a minor price to pay for a range of potential benefits.
Mosseri’s secondary note is that creators will get less reach, and thus less exposure potential. That is a legitimate concern, considering the impact that Instagram now has on exposure for emerging artists.
But compared to the potential harms outlined in the above-noted studies, this doesn’t seem like a strong enough counter against at least a trial of removing algorithms.
What is clear is that Meta will lose money if Australia’s proposal goes into effect. That reason alone will likely lead to Meta and others unleashing an army of lobbyists to oppose any such push.
One particular angle that’s currently being presented in opposition to the bill is that the true motivation for Australia’s proposed change is actually speech control. The opposition alleges that the government wants to limit speech through its own determination of what’s defined as harmful content.
Earlier in the month, American conservative commentator Michael Shellenberger gave a speech at a Conservative Political Action Network conference in Brisbane, in which he claimed the proposed bill could give Australia’s eSafety Commissioner “sweeping powers over online speech, including the ability to operate fake social media accounts.”
Though both of those claims are misleading.
On the first claim, a provision in an initial draft of the bill would have given the communications minister the power to determine new elements of harm to be covered by the bill via “legislative instrument.” That means the minister would be able to make this determination without passing it through Parliament. However, such acts would still need to be tabled in Parliament, where they could be questioned and disallowed by MPs.
After concerns were raised about this, Wells confirmed in her Press Club speech that the government removed this element entirely. That will ensure that any determination of what qualifies as harm under the bill will be determined by Parliamentary review.
Regarding Shellenberger’s second point, the original draft of the bill included a proposal to allow approved researchers to use fake accounts for research purposes within pre-approved and strictly controlled means.
This provision relates primarily to testing whether platforms are adhering to the laws. It was designed to stop the platforms from blocking approved research accounts, which could enable them to avoid scrutiny.
As reported by the Conversation, the proposed provisions don’t give regulators unrestricted power to “covertly manipulate public debate.”
“Instead they allow the regulator to test platforms and their claims, rather than simply take their word for it,” the Conversation said.
The provision’s specific wording will likely be revised to reflect this since the bill is still in the consultation stage.
Which is the key note. None of this has been agreed to or even proposed as a bill yet, and the government is seeking community input on the language and substance of the law.
Yet, the opposition has already begun. As social media platforms try to ensure that their business interests are not impacted, it’s likely these angles will see more discussion online.
