After becoming the first country to ban social media for children under 16, Australia has unveiled a proposal to grant social media users the option to reset their feeds.The proposed “My Feed, My Way” initiative, part of a wider proposed Digital Duty of Care law, would allow users to choose whether their default feed includes personalised content recommended by an algorithm, or only posts from accounts they follow.“These new laws will put the onus on digital service providers to provide a safe online environment, setting minimum standards for the features and tools they use,” Prime Minister Anthony Albanese said.The government argues that recommendation systems can push users towards harmful content, including material related to body-image issues and misogyny. The proposal builds on the European Union’s Digital Services Act, which requires the largest social media platforms to offer users a recommender system not based on profiling.The question then is what changes materially for an Australian social media user?First, how does social media as we know it work today?Over the last 20 years, social media has evolved from being a largely chronological stream of posts from people and accounts users chose to follow, to becoming a hyperpersonalised, endless-scrolling feed of items that are based on a mix of their interests, online behaviour and interactions with previous content.A social media platform today has far more posts available than any user can see. Behind the scenes, recommender systems decide which posts appear first, which get buried, and which get shown even if the user never followed the account that posted it.Story continues below this adPrinceton University professor and AI Snake Oil author Arvind Narayanan illustrated three kinds of social media feeds in an essay titled Understanding Social Media Recommendation Algorithms:Subscription: A user sees content from accounts they have deliberately chosen to follow.Network: Content can travel beyond the original poster’s followers when users reshare or otherwise amplify it through their networks.Algorithmic recommendation: The platform predicts which content a user is most likely to engage with and recommends it, regardless of whether the user follows its source.As social media platforms grew, users were exposed to far more content than they could realistically consume. Recommender systems offered a way to rank that abundance by predicting what each person was most likely to engage with.Also in Explained | How India is trying to regulate children’s social media use — and why gaps remainOver time, that logic became central to the social media experience. Recommender systems learn from what users pause on, what clips they watch to the end, what they like, comment on or share, what they skip, and what keeps them scrolling. At their core, Narayanan writes, these systems are trying to predict which available post a particular user is most likely to engage with next.No major platform fits neatly into just one of these categories. Most combine them. YouTube, for instance, mixes subscriptions with algorithmic recommendations, while TikTok’s “For You” feed is heavily driven by recommendations.Story continues below this adPlatforms especially value engagement because it is measurable, increases the likelihood of users returning, and generates advertising revenue. For a user, however, what keeps them engaged may not be what leaves them satisfied, informed or feeling in control.What would the proposed law do?Australia’s proposal now seeks to reset how users choose what they see by default, instead of giving them another user experience feature to navigate.Rashina Hoda, Professor of Software Engineering at Monash University and the Director of its HumanAISE Lab, notes that opt-in and opt-out create very different experiences for the user.In an opt-out model, as is the default on major social media platforms in India, the personalised recommender system continues to run until the user opts out, or actively switches it off. More often than not, the process is convoluted and cumbersome and relies on “the user’s inertia or force of habit,” according to Hoda.Story continues below this adAn opt-in model, on the other hand, turns off recommendations by default, and users may choose to enable them.Australia’s proposal sits somewhere between these two models: users would be explicitly asked to choose which kind of feed they want as their default. Platforms would now have to notify both new and existing users and ask them to choose their default feed. The government says the choice must be “genuine and enduring”, rather than another control buried in a platform’s settings.Also Read | Meta agrees to $17bn child safety settlement: What changes for teens on Instagram and Facebook?The wider Digital Duty of Care would be enforced by Australia’s online safety regulator, eSafety, with platforms facing penalties of up to A$109.2 million for non-compliance.Either model, however, would need to be accompanied by greater digital literacy.Story continues below this ad“Children, young adults, the elderly, people with lower digital literacy, and other vulnerable cohorts such as people with mental health concerns may not fully understand the options and may not initially know what to do with their own found power of choice and control. They may need help navigating the change,” Hoda told The Indian Express.What changes when the algorithm is switched off?If the proposal were to pass, Australian users would likely revert to a social media experience that no longer defaults to a recommender system. Users who chose a “following” feed would instead see content largely from the friends, family and creators they have chosen to follow.“When personalised recommendations are turned off, a chronological feed is seen as the default,” Emilia Bellucci, Senior Lecturer in Information Systems at Deakin University, told The Indian Express. “This means they are likely to see feeds from our friends and those we have connected with, resulting in ‘boring’ content, ie unlikely to be exciting, sensational or rouse up extreme emotions.”A “following” feed would not, however, be neutral or entirely algorithm-free. What users see would still depend on who they follow, what circulates through their networks, and other platform systems that rank, moderate and organise content.Story continues below this adBellucci said giving users greater control over what appears in their feeds could also change how much autonomy they feel they have over their social media use. They may therefore spend less time on platforms, which could in turn affect the advertising model that currently underpins much of social media. Platforms such as Instagram, Facebook and TikTok derive significant revenue from advertising, and the longer users stay engaged, the more data platforms collect about their interests, moods and behaviour.“The more data points, the better the algorithm can predict what will keep us engaged and ultimately addicted to the platform,” Bellucci said. “The more addicted we are, the more advertisements we see delivered at the right time.”She said that if enough users moved away from personalised feeds, platforms could end up relying more heavily on subscriptions, contextual advertising or other revenue models. But the scale of that change would depend on how many users actually opted out of algorithmic recommendations. “If the numbers are small or easily absorbed, social media companies may do nothing,” Bellucci said.Could platforms comply without really changing the experience?Offering users a choice is not necessarily the same as making that choice meaningful.Story continues below this ad“Platforms could technically comply with the law while still nudging users back towards algorithmic feeds,” Hoda said. “For example, subtle reductions in the other services or features offered by the platform can lead to a reduced quality of user experience and tempt users to move back to algorithmic feeds.”NewsletterFollow our daily newsletter so you never miss anything important. On Wednesday, we answer readers' questions.SubscribeBellucci pointed to a similar risk. Existing options are “hidden in app settings, can be difficult to find or comprehend, and are often reset the next time you open the app,” she said.Platforms could also use “well-timed in-app suggestions” and make it easier to return to the algorithmic feed, Bellucci said, effectively nudging users back towards personalised recommendations.Australian Communications Minister Anika Wells has warned against such “malicious compliance”, where platforms technically meet the requirement while designing the experience in ways that undermine meaningful user choice.Story continues below this adThis means regulators may have to look beyond whether a platform has technically provided a choice. Hoda said compliance authorities would need to focus on “the technical nuances of how adherence to the Duty will be manifested in the software design”, while Bellucci said regulators should monitor how people actually use social media, including how long they remain on platforms.As Hoda put it, a feed limited largely to people and accounts a user follows would resemble how social media worked before recommender systems took over. “It’s funny how we’ve forgotten that is how social media used to initially work before recommender algorithms were introduced.”