Recommendation Algorithms

Australia's bill would make social platforms let users turn off recommended content

Australia's government introduced laws this week that would force social media platforms to let users turn off recommended content, BBC News reported. The package also adds a legal duty of care for online services. Feed teams get a design task and a liability question to size.

TikTok ran a placebo safety tool on thousands of users over a staffer's objection, New York alleges

New York alleges TikTok gave thousands of users, including children, a placebo Algo Refresh tool that confirmed a reset it never made. The unsealed complaint also quotes a TikTok well-being manager's written objection, documentary support for the state's claim that the company acted knowingly.

Molly Russell's father takes his case against Meta to cinema screens before Sorkin's new film

Ian Russell will tell UK cinemagoers "Instagram helped kill my daughter" in a free 30-second advert before Aaron Sorkin's The Social Reckoning. The Molly Rose Foundation pays nothing for the slot, and it asks people watching a film about Mark Zuckerberg's company to fund its campaign against that company.

YouTube cuts Shorts reach for reposted clips that carry only template edits

YouTube changed Shorts recommendations on Oct. 2 to show fewer clips re-uploaded from other creators without additions of their own. Podcasters and streamers who rent reach through clip accounts now have to pay for original commentary or real editing to keep it.

TikTok agrees to strip the For You feed from Alabama minors in a $100 million settlement

Senators put TikTok's 15 million-user control group on a Sept. 1 clock

Blackburn and Blumenthal cite a sealed 2023 document showing TikTok withheld an anti-echo-chamber safeguard from 10% of US users, and they want a list of every other such test.

Two hundred young New Zealand women name online body ideals that contradict each other

Young women in a New Zealand survey described ideals that cannot be held at once, and the authors argue the fix belongs with platforms. The design describes what the women reported; ranking interventions is beyond it.

YouTube ships prompt-built feeds as a side tab after testers feared messing up their algorithm

Custom Feeds lets US viewers type what they want to watch and get a homepage built from it. YouTube designed the tab as a low-stakes place to experiment after testers proved unwilling to steer the real recommendation algorithm.

Washington asks Australia to write American platforms out of its child-safety duty

The Trump administration used Australia's own public consultation to ask that American platforms be exempted from the proposed under-18 duty of care, a request that lands while the exposure draft is still being written.

California prices negligent child harm at up to $1m per child

Gavin Newsom signed a package on Thursday that bans autoplay and history-based feeds for under-16s and requires companion chatbots to alert parents when a young user shows distress. Penalties are counted per child.

ISD counted 150 AI cartoon jihad videos on TikTok with 5.4 million views

ISD and Tech Against Terrorism describe cartoons and fan edits that carry no movement branding for a classifier to grab, which moves the moderation problem from known files to novel ones. TikTok disputes the methodology.

Exporting American polarization findings hands other democracies the wrong levers

A Nature Communications paper led by King's College London argues the evidence base on polarization is so US-heavy that governments elsewhere may regulate social media and AI against drivers their own countries do not have.

Meta buys ten years of defined teen-safety obligations for about $1.8bn a year

The 48-state deal pairs audited age assurance and false-positive targets with a two-hour daily cap and an overnight lockout for under-18s, and leaves autoplay and chronological feeds as things a teen has to opt into.

Senators want every TikTok test that disabled a safety feature, not just the one that hit 15 million

Blackburn and Blumenthal set a 1 September deadline for 13 questions. The broadest one asks for every US algorithmic experiment where TikTok withheld a protection, and how many minors were caught in it.