JMIR Publications

ABSTRACT

Background:

The use of General Purpose Conversational Agents (GPCAs) for social and emotional purposes has become commonplace among adolescents and young adults (AYAs, ages 12–25), yet a significant gap remains in understanding how AYAs themselves perceive and experience its harms and benefits.

Objective:

This qualitative study addresses this gap by centering AYAs’ voices to better understand their use of AI for social and emotional purposes and to provide a first version of an AYA-centered overview of harms and benefits, to inform families, providers and policymakers.

Methods:

We conducted semi-structured interviews and focus groups with 48 AYAs from diverse backgrounds (ages 14–22 years) living in the US.

Results:

Findings revealed a wide range of use cases, including emotional support, social skills guidance, relationship advice, and health guidance. Perceived benefits included privacy, accessibility, directness, and personalized advice, often used as a substitute for unavailable peers/adults or due to barriers to professional care. Key risks identified by AYAs were becoming socially, emotionally, and cognitively dependent on AI, receiving unhelpful, inaccurate, or unempathetic responses, and the erosion of human-to-human relationships, particularly related to AI’s tendency toward sycophancy. While some AYAs disclosed AI use freely, the majority kept social and emotional AI use secret from parents, peers, and teachers, suggesting it remains a taboo topic.

Conclusions:

These findings demonstrate the complex ways AYAs are navigating these technologies and the need for future AI model- and product policies to be co-developed with AYAs to ensure responsible innovation and meaningful support.