Participatory Development and Pilot Testing of an Adolescent Health Promotion Chatbot

Editorial reading guide · Collaboration

If an adolescent helps choose a health chatbot, how can their priorities shape its language, boundaries, and human support?

Demonstrates both the promise and limits of inviting adolescents into chatbot design. Participants wanted confidential, youth-relevant conversation, small talk, and support broader than health instructions. Despite iterative development, misunderstandings and stalled conversations limited engagement. Their preferences are evidence about design fit; they do not establish a health-behavior benefit.

Study context

Dutch secondary-school adolescents: focus groups with 36, log-based pretesting with six, and a pilot analysis involving 73 students

Age / stage: Adolescents 12–15 years

Read with three questions

1. Which adolescent preferences changed the chatbot beyond the researchers’ initial health-promotion aims?
2. What did conversation failures reveal about the gap between natural-language expectations and the bot’s actual capabilities?
3. Which engagement and health outcomes would need longer and more diverse testing?

Keep the limits in view

Only three of 20 approached schools participated, and school convenience recruitment and small qualitative samples limit transfer. The pilot was short and did not establish durable behavior change. Intent-based NLP frequently failed to understand input, so the findings should not be described as a trial of current LLM conversation.