A 2015 study from Yale revealed that individuals who sought explanations online felt intellectually enhanced afterward, even when their searches yielded no real answers. Initially, this seems like a strange quirk of self-deception, a peculiar flaw in how people assess their comprehension. However, it’s more insightful to consider it as a sign of a product functioning exactly as intended, fulfilling the purpose for which it was designed.
Insights from the experiments
The study, conducted by Matthew Fisher, then a doctoral candidate, alongside senior author Frank Keil and co-author Mariel Goddu, involved over 1,000 subjects participating in nine distinct experiments. In the fundamental design, one group searched online for answers to a question while a different control group either did not search or received the same information without needing to look it up. Afterward, each participant evaluated how well they believed they could explain unrelated subjects that were not connected to the search. The search-engaged participants rated their general knowledge higher across the board, not just on the material they investigated. As Fisher noted, “those who seek information often blend easily accessible knowledge with their own personal knowledge.”
The aspect that complicates the dismissal of these findings as mere coincidence is what transpired when searches yielded no results. Even when queries provided no valuable information, participants maintained an exaggerated sense of knowing. Keil, a professor of psychology and linguistics at Yale who directed the lab, characterized the effect as vivid enough to withstand its own failures: “the cognitive repercussions of being in ‘search mode’ online might be so potent that individuals continue to feel smarter, even when their online inquiries yield nothing.” The confidence did not align with the answer’s quality; it was linked to the act of searching itself.
The researchers excluded a few clear alternative explanations during their investigation. It was not merely that searchers were generally more self-assured, since a different group of participants who didn’t engage in searching rated their own knowledge lower on the same unrelated questions. The results were not influenced by the content being flattering or reassuring, as the effect was observed even for topics untouched by the search, and when searches returned no results. What remained, across nine iterations of the same basic setup, resembled a contagion effect: the sense of having trustworthy information nearby transferred to a personal sense of knowing the information, regardless of actual transfer.
The distinction between feeling smart and being smart
This aspect deserves more consideration beyond the headline observation. A search engine’s business model does not rely on users leaving with more accurate beliefs about the world; it relies on users returning. A tool that consistently made users feel less confident after use, even if that uncertainty was more accurate, would prove less effective based on one of the key metrics determining a search company’s commercial viability: repeat usage. A tool that consistently leaves users feeling assured, informed, and marginally smarter, even if the fundamental queries remain unanswered, is commercially superior.
In simple terms, the confidence boost is not merely a side effect; it is a significant emotional byproduct intricately linked to the service provided. For a search-centric business, it aligns more closely with the core offering, irrespective of how the product is marketed. The information obtained is nearly secondary to the sensation of having retrieved it, which is a surprising realization regarding a tool that many view as strictly informational.
This also redefines what “engagement,” the metric that most of these products are ultimately tailored around, actually measures. It is not an indicator of the tool’s usefulness. Instead, it serves as a measure of how reliably the tool generates the feeling of being useful, which can fluctuate independently from whether anyone’s actual understanding has improved. A product team observing engagement increase lacks an inherent signal indicating whether that rise reflects improved answers or just a more convincing impression of receiving one.
The significance of this in today’s context compared to 2015
The original study predates the current surge of AI chat tools that respond in articulate, confident prose rather than providing a list of links for users to sort through. Keil did not evaluate chatbots directly, as the study occurred before their emergence, but the mechanism he illustrated doesn’t necessarily depend on search results appearing as lists of links. If merely entering “search mode” could elevate a sense of personal knowledge, a tool that entirely bypasses the sorting process and