A global study found that individuals with limited knowledge of AI use AI tools more frequently, while those with greater understanding tend to avoid them, challenging assumptions about AI adoption. This paradox highlights that familiarity with AI’s inner workings reduces its perceived magic and reliance, despite lower-knowledge users often fearing AI yet still engaging with it more.
A recent study involving thousands of participants across seven experiments worldwide has revealed a surprising relationship between AI knowledge and AI usage. Researchers administered a challenging test assessing understanding of AI concepts such as algorithms, training data, attention mechanisms, and model structures. They then tracked how frequently these individuals used AI in their daily lives. The results showed that those with limited AI knowledge used AI tools regularly, while those who scored highly on the test tended to avoid using AI.
This pattern was consistent across various demographics, including college students, everyday consumers, and even at the country level across 27 nations. The study’s findings challenge common assumptions about AI adoption, highlighting that lower AI literacy correlates with higher AI usage. This counterintuitive trend suggests that familiarity with AI’s inner workings may reduce reliance on AI technologies.
In a practical context, the researchers also surveyed 36 senior executives at a major insurance company about potential AI product markets. All executives believed that targeting knowledgeable AI users would be the best strategy. However, the study’s data contradicts this belief, indicating that the primary users of AI are those with less understanding of the technology.
The researchers draw an analogy to magic shows: before understanding the mechanics behind a trick, audiences are amazed and perceive it as magical. Once the trick is explained, the wonder fades. Similarly, people unfamiliar with AI may view it as a powerful, almost magical tool, while experts see it as a complex but ultimately mechanical process of pattern matching without true understanding or emotion.
Additionally, the study found that individuals with low AI knowledge tend to fear AI more and doubt its effectiveness, yet paradoxically, they still use it more frequently. This suggests a complex relationship between trust, fear, and usage. Ultimately, the core takeaway is that increased knowledge about AI systems tends to decrease their usage, raising important questions about how AI is perceived and adopted in society.