Recently, news sources from the New York Times to Google have suggested the best way for readers to fix their news feed is to filter their sources of information. I’ve created products and experiences for billions of people over my career and that advice is flat out wrong for news.

News via Feeds

Feeds are how the vast majority of Americans get their news. At their core, feeds use a technique called attention reinforcement to predict what to show readers, optimizing for likes, shares, and comments. Unfortunately, when applied to news, attention reinforcement inevitably leads to two failure modes.

The first is that the natural endpoint of unconstrained personalization is to put the reader in a filter bubble. Filter bubbles create feeds where readers expect to only see content that matches their existing beliefs. To accomplish this, the feed becomes a stream of ads and content unique to each reader, removing any sense of shared experience. This custom feed is also uniquely vulnerable to the second effect.

Outrage capture is the second, more insidious effect. Outrage capture happens because once readers are wrapped in filter bubbles, the easier and most cost-effective way to generate increased engagement is by generating outrage. Worse, outrage is profitable. Whether the goal is growing a personal brand, selling ads, or spreading disinformation, hijacking attention reinforcement algorithms with outrage is the winning strategy. And it’s easy when a reader is primed to expect content they agree with.

Filter bubbles and outrage lead to long-term drops in news trust, engagement, and rising news avoidance. Suggesting that readers filter their sources only accelerates both of these failure modes. It’s the wrong direction for news.

Rethinking Attention Now

From video games and virtual worlds to Meta and Google, I’ve spent a lifetime building experiences that capture people’s attention, connect them to each other, and deliver news and information. I feel a deep sense of responsibility to take this problem head-on, particularly with the new technologies available to reinvent how we think about feeds and news. A healthy news ecosystem requires products that respect the attention of readers, the work of journalists, and reward a high-quality, shared view of news. It’s why I joined SmartNews to create NewsArc.

This is a very challenging problem, particularly in an environment where attention reinforcement is viewed as the only viable way to deliver news. While there is justifiable concern around Large Language Models (LLMs) and their capacity to flood the Internet with garbage, LLMs’ remarkable language capabilities also unlock entirely new ways to deliver the news.

LLMs enable product designers and editors to build systems that help us deeply understand the news and make effective predictions about what would matter most to a reader. Rather than optimizing an infinite feed of attention-grabbing headlines, LLMs can understand the importance, quality, and benefits of a given article. Instead of focusing on generating a click, it is now possible to prioritize news that someone would regret not knowing. Whether that news is deeply serious, health and safety reporting, or the latest celebrity gossip, LLMs make it possible to shift the focus from a personalized feed – with all its defects, vulnerabilities, and incentives – to shared news experiences that reward quality.

To me, this is the electric car moment for news. Rather than carrying forward all the challenges of old AI models, we can build technologies that use LLMs to enable journalists and editors to scale their expertise. It’s an incredibly exciting shift.

This isn’t how most products are using LLMs today. Most people are experiencing LLM-created synopses or bullet points. While these are certainly solid use cases, I believe the more important question for news products is what will help readers better discover and enjoy the great work of journalists and publishers? And what business models will ensure creators are compensated for their work?

Inventing the News Future We Want

News and journalism matter. Shared understanding of what’s happening around us is a key element of democracy and community. NewsArc is built on these ideas. From conversations with publishers, journalists, and readers, we know there’s excitement for a news experience that respects the entire ecosystem. News readers want the best, most complete view of the story and we finally have the tools and technology to do it.

Rather than offering recommendations that reinforce the worst properties of feeds, it’s time for platforms to build the solutions that will create a better news future.