Social Media ‘Engagement’ Raises Risk of Manipulation and Misinformation

Facebook Begins Reducing Political Content in News Feeds, Acknowledging Algorithm Issues

Facebook has initiated a quiet experiment to reduce the amount of political content appearing in users’ news feeds, tacitly recognizing fundamental problems with how its algorithms function. The social media giant’s shift highlights the growing tension between content that provokes user responses and content that people genuinely want to see.

At the core of this issue lies the distinction between engagement and quality. Social media algorithms heavily rely on user behavior to determine what content to display, particularly focusing on posts that generate likes, comments, and shares. While this approach appears logical on the surface, it contains inherent flaws that can lead to problematic outcomes.

The principle behind these algorithms resembles the concept of “wisdom of crowds,” where collective signals from other people’s actions and preferences guide decision-making. This evolutionary concept has been hardwired into human brains through cognitive biases like the familiarity, mere-exposure, and bandwagon effects. When everyone around you starts running, your instinct is to run too—someone might have spotted a predator, and asking questions later is the safer approach.

Our brains naturally collect environmental cues, particularly from peers, and translate them into decisions using simple rules: follow the majority, copy neighbors, go with winners. These heuristics typically work well because they’re founded on reasonable assumptions that people act rationally, the majority is usually correct, and past behavior predicts future outcomes.

Modern technology, however, dramatically expands this dynamic by allowing individuals to access signals from vast numbers of strangers. AI applications heavily leverage these popularity signals across platforms—from search engine results to music recommendations, and crucially, to ranking social media posts.

Research shows that virtually all web platforms exhibit strong popularity bias. When systems are driven by engagement metrics rather than explicit user queries, this bias can produce harmful unintended consequences. Social media platforms like Facebook, Instagram, Twitter, YouTube, and TikTok rely extensively on algorithms that process user engagement signals to rank and recommend content, aiming to maximize further engagement.

The assumption that this approach helps high-quality content rise to prominence has been tested and found wanting. Studies examining algorithms that combine quality and popularity metrics reveal that popularity bias typically reduces overall content quality. The fundamental problem occurs because engagement provides unreliable quality indicators when few people have seen an item. Initial noise gets amplified, and once low-quality content gains sufficient popularity, the algorithm continues promoting it.

The problem extends beyond algorithms to affect human behavior. Information spreads through “complex contagion”—the more someone encounters an idea online, the more likely they are to adopt and share it. When social media indicates something is going viral, cognitive biases create a nearly irresistible urge to engage with it.

An experiment using Fakey, a news literacy app that simulates social media feeds, demonstrated that users are more likely to share content from low-credibility sources when they can see that many others have already engaged with it. Exposure to these engagement metrics creates a vulnerability in users’ judgment.

The wisdom of crowds fails in social media contexts because it assumes the crowd consists of diverse, independent sources. In reality, people tend to associate with similar individuals, creating homogeneous communities or echo chambers. Additionally, interconnected friend networks influence each other’s preferences, distorting independent judgment.

Furthermore, popularity signals can be manipulated. While search engines have developed sophisticated techniques to counter manipulation schemes, social media platforms are only beginning to address their vulnerabilities. Bad actors have created fake accounts and networks to create the illusion that conspiracy theories or certain political candidates have widespread support, simultaneously exploiting both platform algorithms and human cognitive biases.

In response to these challenges, social media companies have taken defensive measures, particularly during elections, by removing fake accounts and harmful misinformation. However, these efforts often resemble a game of whack-a-mole.

A more preventative approach would involve adding friction to slow information spread. Implementing CAPTCHA tests or fees for high-frequency behaviors like automated liking and sharing could reduce manipulation opportunities while allowing users to pay closer attention to content. Perhaps most importantly, social media platforms could adjust their algorithms to rely less heavily on engagement metrics when determining what content to serve users.

Facebook’s current experiment with reducing political content represents an important first step toward addressing these complex, interrelated issues at the heart of social media’s influence on public discourse.