When More Information Makes Us Less Informed

At a time when smartphones have become small Hollywood studios, it’s never been easier for creators to bypass traditional gatekeepers and reach audiences directly. For example, a talented foodie doesn’t need to be hosted by a gourmet website to build a food channel on TikTok, Instagram and YouTube. 

But if idea sharing has become so cheap, how does that affect the quality and truthfulness of the information we absorb and retain?

Information overload might ultimately erode the quality of collective knowledge, said David McAdams, the Robert A. Bandeen Distinguished Professor of Business Administration at Duke University’s Fuqua School of Business.

In the new paper, The Economic Epidemiology of Ideas, McAdams builds a theoretical model showing that as new content inundates media platforms, people spend less time filtering information, pushing communication down a slippery slope of declining quality. 

“As social filtering erodes, society as a whole might end up a bit dumber,” McAdams said.

The paper’s model describes a plausible future scenario of over-saturated media environments, in which the abundance of unfiltered information makes it harder to discern valuable from low-quality content.

Does that mean some level of gatekeeping might not be bad, after all?

How ideas spread 

McAdams describes a marketplace of ideas that behaves like an epidemic. 

Ideas are constantly created and spread throughout digital and human networks. Some ideas come directly from producers like journalists, scientists, companies or influencers. Others reach us through friends’ recommendations, social posts and reviews.

But unlike viruses, ideas don’t spread automatically. People have agency. They decide what to pay attention to, what to retain, and what to pass along.

McAdams describes people as information processors: they evaluate what they encounter and only share the ideas they believe are valuable.

The farther ideas move through such “social transmission chains,” the more social filtering occurs, improving the average quality of ideas in circulation—in a process similar to peer review for scientific papers.

However, as idea production becomes cheaper and cheaper, the question is whether people will still devote enough attention to filter out lower-quality ideas.

The importance of social filtering 

“A wealth of information creates a poverty of attention,” wrote Nobel laureate Herbert Simon more than 50 years ago.

“What Simon meant is that the key, for any organization and for society more broadly, is not to generate all the data in the world,” McAdams said. “What’s important is getting the right data to the decision point.”

The problem is that as the volume of information rises, attention becomes fragmented and people spend less time filtering ideas, he said.

Historically, intermediaries like newspapers, academic journals, and record labels functioned as gatekeepers, allowing audiences to focus on a smaller set of content. This created more opportunities for social filtering and “positive feedback loops,” he said.

“There are positive spillovers when people evaluate ideas,” he said. “The more people spend time engaging socially, the better the quality of information in people's minds. People then want to interact socially even more, because what they hear socially is better.”

McAdams compared the process to returning to the office after COVID. When more people interact, each interaction becomes more valuable because there are more opportunities to learn from others. Information ecosystems can generate similar positive feedback loops.

An overload crash

McAdams’ model shows that information systems may suddenly shift and deteriorate, if overloaded with too much new content.

"Imagine a streaming platform releasing so many new shows that none receives much attention," McAdams said. "As conversations fragment, transmission chains become shorter and the content that people encounter socially from others falls in quality. In the long run, people are worse off because they can no longer rely on word of mouth to identify what shows to watch.”

As social transmission declines and the reach of new content drops, producers may also end up making less money—even if production costs are lower. 

“The overall welfare being created by an information system can fall off a cliff,” he said.

The role of gatekeeping

The model suggests that some forms of gatekeeping can improve collective learning, not necessarily because gatekeepers know which ideas are good and true, but simply because they reduce the number of ideas in circulation.

When fewer ideas compete for attention, each idea receives more scrutiny. 

In that sense, a prestigious academic journal, a trusted editor, or even a carefully curated platform can improve information quality simply by narrowing the field, McAdams said.

The same principle may apply to credibility signals.

McAdams points out that information often carries “tags” that help audiences assess trustworthiness. Institutional affiliations, expert endorsements, and reputation scores are all markers of credibility.

The business of filtering

McAdams said that companies like Consumer Reports, industry analysts, and professional research providers create value by helping audiences identify high-quality information in crowded environments. Their role may become even more important as information becomes cheaper to produce.

“If a reputable recipe provider takes 10,000 recipes and figures out the five best, that’s a signal that can help people identify ideas—here, recipes—that have already passed through meaningful forms of scrutiny,” McAdams said.

AI could eventually play a similar role.

AI may have the potential to filter, organize, and evaluate information, McAdams said. Whether it can help improve social learning will depend partly on how effectively people can use it.

“If AI is encountering all these ideas and evaluating them, and it gets feedback from people about what was useful, then maybe it can improve social transmission, and what you learn from AI becomes high quality,” he said. “But that all depends—at least as AI exists today—on people knowing what to ask and giving AI systems the right feedback. You will still need people to actively evaluate information.”