Recent reports suggest that as many as 50% of Americans are now regularly using AI in their daily lives, but the actual number could be even higher.
A research paper from the University of Chicago suggests that a significant number of individuals are actively concealing their reliance on artificial intelligence due to social pressures. The study explores social desirability bias—the tendency for survey respondents to provide answers that look favorable to outside observers rather than telling the truth. This behavioral pattern is causing a major underreporting of how deeply tools like ChatGPT are embedded in daily routines.
The researchers, Yier Ling, Alex Kale, and Alex Imas, surveyed 338 undergraduate students to examine this phenomenon using a psychological technique known as indirect questioning. Instead of only asking participants about their own habits, the survey required them to report on the AI usage of their immediate peers. The results revealed a massive divide: while roughly 60% of students admitted to using large language models themselves, they estimated that nearly 90% of their peers were regularly doing the same. This gap appeared across almost every metric tracked, including the frequency and overall reliance on the technology.
To understand the mechanics behind this statistical discrepancy, the authors conducted a follow-up study with a separate group of students. When presented with the findings, 79% of respondents stated that the gap exists because people are underreporting their personal usage. The overwhelming majority—70%—selected a forced-choice explanation stating that individuals are simply embarrassed to admit they rely on these models, though they feel perfectly fine acknowledging that their friends do.
The data is wrong.
Relying on skewed self-reported surveys prevents institutions from accurately assessing how technology reshapes workflows. Qualitative feedback from the participants shed light on the cultural pressures driving this behavior, pointing directly to “AI shaming”. This is an emerging social norm where relying on machine-generated output is stigmatized as lazy, deceitful, or lacking in individual capability. In academic and creative settings, declaring the use of an algorithm is frequently conflated with a lack of personal skill or a violation of integrity, forcing users to keep their efficiency gains private.
This discrepancy introduces substantial challenges for organizations trying to build proper operational strategies. In environments where tech adoption is heavily criticized, self-reported data becomes unreliable for leadership. Interestingly, the paper notes that the opposite pattern can occur in corporate environments. In workplaces where there is immense pressure to look innovative, workers might engage in “AI washing” by overreporting their usage to satisfy managerial expectations and appear tech-savvy.
Whether driven by a culture of shame or a culture of hype, failure to account for these cultural biases can lead to a severe mis-measurement of technology. For organizations aiming to implement effective oversight, understanding the hidden social dynamics of AI tools is becoming just as critical as the capabilities of the technology itself.