UW Study: Children Skeptical of Generative AI Toys

Children are not passive consumers of artificial intelligence; they are skeptical evaluators who quickly lose patience with technical failures. A recent study by researchers at the University of Washington's KidsTeam UW found that while children are initially curious about AI-powered toys, that interest often sours into frustration when the technology fails to meet their conversational expectations.

The study involved eight children between the ages of six and eleven who interacted with AI toys from the company Curio, including characters such as Gabbo and Ballerina Cappuccina. These toys utilize AI models designed to remember previous interactions and tailor their responses to the user over time. However, the participants frequently struggled with the toys' inability to listen or comprehend complex dialogue. One study participant highlighted this friction, stating, "It didn’t listen to me like 26 million times."

A Shift in Play

Historically, talking toys—from the 1960s-era Chatty Cathy to the 1990s Furby—operated on simple programmed languages or pre-recorded lines. In those cases, children projected their own imagination onto the object. The introduction of generative AI changes this dynamic by providing what Jason Yip, a UW associate professor and director of KidsTeam UW, describes as an "imitation of imagination." According to Yip, the industry has entered uncharted territory because the toy now attempts to simulate genuine conversation and memory.

The Uncanny Valley of Plush

This shift has created a psychological tension for young users. Aayushi Dangol, a researcher and co-lead author of the study, noted that the juxtaposition of a soft plushie with signs of intelligence was "both interesting and disturbing for the kids." This unsettling combination, paired with technical glitches, led some children to describe the toys as "ugly" and "evil," with some even joking about throwing the devices into the ocean.

Implications for Development

The findings, presented June 25 at the Interaction Design and Children conference in Brighton, UK, underscore a significant gap in the understanding of how AI influences emotional development and play. Researchers warn that AI toys carry risks, including the potential to provide incorrect information, excessively flatter children, or manipulate them into unhealthy emotional attachments.

What Remains

As generative AI continues to integrate into consumer products for minors, the UW study suggests that developers must address the gap between AI's perceived capabilities and its actual performance. The research leaves open the question of how these "imitations of imagination" will affect long-term cognitive development and whether the inherent unpredictability of LLMs can ever truly align with the needs of a child's play environment.