黑料专区

From Meta and TikTok to Amazon and Spotify, recommendation algorithms form the backbone of the modern digital economy. While they create value, their design and implementation have raised policy concerns, including the effects of social media on young users, the influence of online platforms on political discourse, and the economic power imbalances between producers and consumers.

"Recommendation algorithms can help people navigate the internet, but we need to make sure they have a positive impact on society," says Mike Luca, Block Center Director. "This includes mitigating the harms they can cause."

These concerns have also brought about increased legal and regulatory efforts seeking to address the consequences of algorithmic systems. For instance, in August 2026, Meta agreed to pay roughly with a group of states based on allegations that the platform misled the public on the potential harm posed to underaged users of its product.聽

This September, the Block Center for Technology and Society convened its Tech Policy Forum, Recommendation Algorithms, Markets, and Society in Washington, DC, to examine the evolving policy landscape surrounding these systems. "A key challenge is that the same algorithmic features that make digital products engaging can, when taken too far, contribute to polarization, market power, or digital addiction,鈥� says Ananya Sen, Associate Professor of Technology, Business, and Society at Heinz College and Conference Lead. 鈥淭his tension makes policy and platform design difficult: how do we preserve the value these systems create while limiting the harms that can emerge from the same design choices?"

We wanted to share some of the insights that came out of the convening. Here are four themes that emerged around the regulation of recommendation algorithms:

Regulators are right to be concerned about recommendation algorithms.

• Panelists generally agreed that there is reason to be concerned about potential harms arising from recommendation algorithms, pointing to examples such as social media and online marketplaces.
• Business models of platforms featured heavily in the discussion, given the potential for misaligned interests of platforms and consumers. A platform may want to increase engagement even when a user prefers to spend more time offline.
• Academic research has also documented evidence of digital addiction, as well as the ways in which platforms can steer behavior through design choices.
• Recent policy investigations similarly raised concerns about potential harms connected to issues of market power, misinformation, digital addiction, and mental health 鈥� especially among young users.

Recommendation algorithms give platforms power to influence behavior.

• There was broad agreement that platforms can play a role in shaping behavior. For instance, University of Minnesota Professor noted 鈥渢hat Spotify's decisions to put songs on their widely-followed playlists have large effects on the fortunes of artists and record labels,鈥� pointing to his joint with Professor Luis Aguiar of the University of Zurich.
• Another manifestation of the influence of algorithms is the heavy emphasis on engagement, and corresponding evidence of digital addiction. For instance, University of Illinois Professor discussed on the topic, joint with Hunt Allcott and Matt Gentzkow, which has found that roughly 31% of social media usage can be traced back to digital addiction.

Policymakers need to account for first principles, business models, and emerging empirical evidence, in addressing the power of recommendation algorithms.

• Progress in identifying and addressing potential harms has often been slow, in part because platforms control much of the data researchers need to study their effects.
• Both policymaking and academic research have struggled to keep pace with technological developments, emphasizing the need for approaches that can respond to emerging concerns as new evidence is discovered. Waldfogel noted, 鈥渆vidence on the most basic questions, such as 鈥榟ow much power do platforms have?鈥� is thin鈥� pointing to the value of more research.
• 黑料专区 Professor , former Assistant Attorney General for the Antitrust Division at the U.S. Department of Justice emphasized the value of acting on observable warning signs, noting that 鈥渟ometimes, all you need to do is stick your hand out to see if it鈥檚 raining鈥�.
• Early intervention is important, as addressing potential harms before they become entrenched may make regulation more effective.
• Policymakers must weigh competing societal values, particularly when empirical evidence is incomplete. As University of Michigan Professor observed, 鈥淲hat kind of childhood do we as a society want for our kids? Evidence-based policy is one approach. Value-based policy is another.鈥�
• Taken together, these considerations emphasize the importance of grounding policy in first principles and an understanding of platform business models, while incorporating the best empirical evidence as it is discovered.

Protecting young users involves designing effective interventions that account for real-world tradeoffs, societal values, and emerging evidence.

• Developing effective protections requires policymakers to consider issues such as addictive platform design, market incentives, how harm is defined, and who should ultimately be held accountable.
• Addressing the effects of recommendation algorithms on children will likely require a combination of complementary policy interventions. Effective protections should also account for the incentives that shape how social media platforms and other digital services are designed.
• Interventions can involve trade-offs. For instance, it may require balancing children鈥檚 desire to participate in digital environments with the need to support their well-being, safety, and growth.