Case Study
Empowering children with chronic illness to express their emotions
At the University of Michigan School of Information, we collaborated with Naver to develop a pediatric chatbot system that supports the communication between children with cancer and their caregivers.
Product
Chatbot framework
Skills
User Research
Artificial Intelligence
My Role
Research Assistant
Timeline
August 2022 - April 2025
Collaborators
Woosuk Seo, Sun Young Park, Mark Ackerman, Young-Ho Kim, Ji Eun Kim
Overview
Bridging the gap between stakeholders in pediatric care
Children with chronic illness often struggle to talk about their feelings during doctor visits. I co-designed a chatbot system that helps young patients share their emotions more openly, making healthcare experiences less intimidating, more empathetic, and more collaborative. Parents can also use the chatbot to gain advice and reassurance about their communication approaches and techniques. The chatbot can also serve as a tool for assessment for health providers.
Problem
Effective communication is critical in healthcare, but children require a different approach.
In 2021, our team interviewed 20+ children with cancer and their caregivers and discovered three main communication challenges between the two stakeholders.
Problem #1 — Different perspectives on life with cancer
Parents feel endless anxiety about their child’s health, causing them to implement many restrictions on the child. The child does not fully understand or realize the importance of the restrictions. This leads to disagreements about major life choices such as attending school and playing sports.
Problem #3 — Children hiding their true feelings
Children tend to conceal physical and emotion pain in order to avoid being treated differently by their parents or prevent their parents from being worried about them.
Research Question
User Interviews
Learning from child-parent dyadic interviews
We recruited 12 child-parent pairs to interview concurrently. The child participants were between the ages 6-12 and have received medical treatment for a chronic illness such as cancer and their parents identified as a primary caregiver. Each session was about 60 minutes.
Each interview was broken down into three phases
Caregiver interview phase — We first interviewed the parent to learn about their experiences, perspectives, and responsibilities as the primary caregiver of the child.
Child interview phase — We then provided each child with three comic-strip stories that demonstrated communication challenges in pediatric healthcare. We used scenarios as prompts to explore how each dyad navigates the specific contexts and how they would mitigate the communication challenge.
Design workshop — We provided the kids with colored pencils and paper to draw out their ideal chatbot friend. After, we initiated a collaborative discussion for the pairs to describe their ideal chatbot features.
Thematic Analysis
Understanding the attitudes and behaviors of caregivers and children
After completing the interviews, we coded each transcript. We then sorted the code across all 12 interviews into distinct thematic groups.
Theme #1 — An understanding friend
Children expect the chatbot to take on the persona of a friend that is about the same age and has also undergone cancer treatment. Parents believe that this allows the child to better relate to and connect with the chatbot.
Theme #2 — An emotional outlet
Children want someone who listens to them. Parents want them to have a safe space where they can share their feelings without judgement.
Theme #3 — Be appropriate
Parents do not expect the chatbot to be completely honest and instead redirect serious conversations, such as questions about cancer death rate, to a trusted adult.
Theme #4 — The middleman
Children and parents both expect the chatbot to act as their communication middleman that can relay information between the two, especially when the child is too nervous or scared to speak to their parent about a specific topic.
Theme #5 — Confidentiality
Children and parents expect the chatbot to ask for consent before sharing private conversations in order to maintain trust between the chatbot and user.
Language Model
Meet ChaCha!
In order to create the chatbot system, we used the chatbot model called ChaCha that combines a state machine and large language models (LLMs). It was developed by research scientists at Naver Cloud AI Lab in collaboration with members from our team. ChaCha is trained to carry free form conversations and support children when they share their emotions.
Chatbot System
A dual-chatbot architecture powered by ChaCha
Child Bot — This chatbot engages directly with children. It acts as a friendly companion, guiding them through scenario-based conversations to help express feelings and understand their health in an age-appropriate way. It balances empathy and factual information.
Expert Bot — Aimed at caregivers, this chatbot offers contextual support and health communication strategies, helping adults better understand their child’s needs and emotional state.
By combining structured flows (via state machines) with the adaptability of LLMs, the system provides personalized, emotionally sensitive conversations tailored to pediatric oncology care.
