Parents may be paying for artificial intelligence not because they are convinced it improves learning, but because they fear their children will fall behind classmates who use it. A new international study suggests that social pressure is becoming a powerful force in the adoption of AI for schoolwork, even while families remain deeply uncertain about the technology’s long-term educational effects.
Published in the Proceedings of the National Academy of Sciences, the study examined how parents respond when they are told that other teenagers are using AI tools to complete homework. Researchers surveyed and conducted experiments with approximately 2,000 parents in the United States, Canada and the United Kingdom. Every participant had at least one child between the ages of 13 and 18, placing the study’s findings directly within the age group most likely to encounter generative AI in secondary education.
Participants were asked how much they would be willing to pay for a three-month premium subscription to an unrestricted AI service that their child could use for schoolwork. The researchers then presented different versions of the same situation: in one, 20 per cent of young people were said to be using AI for homework; in others, the figure rose to 40, 60 or 80 per cent. This design allowed the researchers to measure how perceived peer behaviour altered parents’ demand for the technology.
The effect was striking. For every ten-percentage-point increase in the assumed share of young people using AI, parents’ average willingness to pay increased by 1.83 US dollars. When the imagined level of adoption rose from 20 per cent to 80 per cent, willingness to pay increased by more than 60 per cent, or approximately 11 dollars. The result indicates that parents’ decisions are shaped not only by their private assessment of AI’s usefulness, but also by what they believe other families are doing.
This pattern is closely related to a concept known as social learning. When people face uncertainty, they often use the behaviour of others as a source of information. If many students appear to be using AI, parents may interpret that adoption as evidence that the technology is necessary, effective or becoming unavoidable. At the same time, the behaviour of others can create a competitive concern: even if a parent believes AI may weaken learning, refusing to provide access could seem risky if classmates are gaining an advantage.
About one in five parents explicitly said they were motivated by the fear that their child would otherwise fall behind. That concern reflects a potential coordination problem. Families may support access to a technology they do not fully trust because they expect other families to adopt it. In such circumstances, individual choices can push AI into classrooms rapidly, even without broad agreement that unrestricted use is educationally desirable.
The researchers also tested whether information about the risks of AI could reduce parents’ willingness to pay. Some participants were told about evidence from another field study in which students used GPT-4 while completing mathematics exercises. In later tests completed without AI assistance, those students scored just under 20 per cent lower than students who had practised without the tool. The finding is consistent with a possible “skill substitution” effect: AI may help users produce an immediate answer while reducing the amount of mental effort invested in developing independent problem-solving abilities.
Risk information did change parents’ broader attitudes. Support for school-wide rules or bans increased from 44 per cent to 57 per cent after participants learned about the potential decline in unaided performance. Yet the information had virtually no effect on their willingness to pay for an unrestricted AI subscription for their own child. In other words, parents became more supportive of collective limits while remaining willing to purchase access individually.
That apparent contradiction is central to the study. It suggests that parents may distinguish between what they want schools to regulate for everyone and what they feel compelled to provide at home. A parent might believe that unrestricted AI use undermines learning across an entire school, while still paying for the tool to ensure that their own child is not disadvantaged. Such incentives can accelerate adoption through a self-reinforcing cycle: each family’s defensive decision makes AI use appear more widespread, increasing pressure on the next family to follow.
The authors, including Christopher Roth of the University of Cologne, Leonardo Bursztyn and Alex Imas of the University of Chicago, doctoral researcher Aaron Leonard and Rafael Jiménez-Durán of Bocconi University, argue that the findings have implications well beyond education. New technologies often spread before their long-term consequences are settled, particularly when people believe that access may affect their relative position. For schools and policymakers, the results suggest that warnings about individual risks may not be enough to slow adoption. Clear, shared rules could be more effective because they reduce the fear that opting out alone will leave a child behind.
Subject of Research: People
Article Title: Social dynamics of AI adoption in parents’ educational decisions
Web References: https://doi.org/10.1073/pnas.2533193123
References: Proceedings of the National Academy of Sciences; DOI: 10.1073/pnas.2533193123
Keywords: artificial intelligence, generative AI, education, schoolwork, parents, teenagers, social pressure, technology adoption, AI risks, educational policy
