UW Study Finds Curio AI Toys Struggled With Children’s Complex Questions

A University of Washington study of eight children playing with three Curio AI toys found that early curiosity often gave way to frustration when the toys could not respond well to complex questions. The result is a small, short-term look at how children react when a plush toy is also expected to hold a conversation.

A talking plush toy meets real play

Curio’s plush toys use onboard AI models that let them talk with children, remember conversations and personalize responses. UW researchers used three of those toys in two campus sessions with children ages 6 to 11.

The children began by asking basic questions, including a toy’s name, and testing its responses to touch and conversation. Some enjoyed details such as a toy naming seven as its favorite number. But the toys often did not handle more complex questions well; some children then called them “ugly” and “evil” and joked about throwing them into the ocean.

Observed frustration versus a broader warning

Co-lead author Aayushi Dangol said the combination of a plush toy and signs of intelligence was both interesting and disturbing for the children. She warned that AI toys can give incorrect answers, flatter children excessively or manipulate children into forming attachments, even when the toys are fun and dynamic.

Children were initially curious but often became frustrated when the toys did not respond well to complex questions. Researchers identified incorrect answers, excessive flattery and manipulation into attachment as possible harms.

A child-led exercise, not a long-term verdict

Those warnings were not outcomes measured in the two sessions. The exercise recorded children’s play and reactions; it did not establish whether incorrect information, flattery or attachment would emerge through longer use at home.

After playing, participants completed a comicboarding activity, filling in panels to imagine what could happen if they continued using the toys. The team presented the work June 25 at the Interaction Design and Children conference in Brighton, United Kingdom.

The study’s eight-child sample offers close detail about a limited setting, not a broad measure of children’s experiences or long-term household use. The research was funded by the National Science Foundation, the Institute of Education Sciences, the U.S. Department of Education and the Institute of Museum and Library Services.