PNAS: Peer Take-Up Lifts Parent AI Spend 60%, Warnings Don't

Something more interesting than the usual will-AI-ruin-kids panel debate is buried in a new PNAS paper from a group of Chicago Booth, Cologne, and Bocconi economists: parents are not really choosing AI subscriptions on the merits. They are choosing them because other kids have them.

In the study, published in PNAS, Leonardo Bursztyn, Alex Imas, Rafael Jiménez-Durán, Aaron Leonard, and Christopher Roth ran incentive-compatible experiments on roughly 2,000 parents of 13-to-18-year-olds in the US, Canada, and UK. They elicited maximum willingness to pay for a three-month premium subscription to a leading general-purpose AI while varying how many of the child's peers were said to be using it. Willingness to pay rose $1.83 for every 10-percentage-point increase in perceived peer take-up, and shifting the scenario from 20% to 80% lifted spending by more than 60%, or roughly $11.

The follow-up experiment is the one that should worry anyone counting on information campaigns. When parents were told about a field experiment showing GPT-4 users scored 'just under 20% lower' on later math tests taken without AI, their beliefs updated and their appetite for a schoolwide ban jumped from 44% to 57%, per the Chicago Booth Review write-up. Willingness to pay barely moved. About 20% of parents who backed a ban still permitted AI use anyway, citing fear their kid would fall behind, and over 70% agreed AI use helps in the short run even if it may hurt long term.

A few things the coverage does not settle. The retrieved write-ups do not report an income split, do not say whether hypothetical willingness to pay translated into real subscription purchases, and cover only three high-income Western countries. Christopher Roth's stated read is blunt: 'Social pressure can accelerate the spread of new technologies even before their long-term consequences are fully understood.'

For anyone in ed-tech, the sales motion aimed at anxious individual parents is the easy one. The lever the authors themselves point to, coordinated school-level adoption of structured tools, is where policy actually bends the curve.