ChaCha: Leveraging Large Language Models to Prompt Children to Share Their Emotions about Personal Events

TL;DR

This work tackles the challenge of helping children express emotions about personal events by introducing ChaCha, an LLM-driven chatbot that uses a state-machine dialogue flow to guide open-ended narratives. Informed by formative interviews with child mental health professionals, ChaCha adopts a peer-like persona, empathizes with emotions, offers emotion word options, and encourages sharing with parents, all while maintaining safety via a backup Help phase. An exploratory study with 20 Korean children demonstrates ChaCha's ability to elicit event-centered emotional disclosures, revealing that children perceive the system as a close friend and that the phase-based design effectively structures the conversation toward labeling emotions, exploring solutions, and diary-style recording. The findings illuminate the potential and caveats of deploying LLM-driven chatbots for children's emotional development, highlighting design considerations around empathy, consistency, long-term use, and parental involvement to balance support with child privacy.

Abstract

Children typically learn to identify and express emotions through sharing their stories and feelings with others, particularly their family. However, it is challenging for parents or siblings to have emotional communication with children since children are still developing their communication skills. We present ChaCha, a chatbot that encourages and guides children to share personal events and associated emotions. ChaCha combines a state machine and large language models (LLMs) to keep the dialogue on track while carrying on free-form conversations. Through an exploratory study with 20 children (aged 8-12), we examine how ChaCha prompts children to share personal events and guides them to describe associated emotions. Participants perceived ChaCha as a close friend and shared their stories on various topics, such as family trips and personal achievements. Based on the findings, we discuss opportunities for leveraging LLMs to design child-friendly chatbots to support children in sharing emotions.

Figures (4)

• Figure 1: Photos taken in one interviewee's office: (a) The interview settings with a book that the interviewee showed and a tablet of the experimenter to demonstrate a video prototype, and (b) emotion cards that the interviewee uses when guiding children to describe their emotions. The emotion keywords on the envelopes are anger, anticipation, joy, sadness, fear, and surprise (from left to right).
• Figure 2: The overview of conversational phases of the ChaCha dialogue system and transition rules among them. Each time the user enters a message, the system inspects the entire dialogue history by performing a test corresponding to the current phase to decide whether to proceed to another phase or stay.
• Figure 3: An example case for the mechanism of ChaCha response generation in the [patternparam, background-color=labelcolor]Label phase, especially how the LLM is prompted dynamically. Receiving the child's message, (1) the conversation analyzer (Ⓐ) analyzes the current dialogue (Ⓑ) and extracts a structured summary (Ⓒ) of what emotions are identified and whether ChaCha has acknowledged them. Combining the incomplete piece of the summary (Ⓓ) as well as the summary data from the previous phase (Ⓔ), (2) the system formulates a new instruction (Ⓕ) for the response generation. (3) That way, the LLM (Ⓖ) generates a response (Ⓗ) explicitly steered to empathize with the child's regretful event.
• Figure 4: The frequency of the eight asking-expression categories by conversational phase. The orange bars indicate the average turn ratio for ChaCha's turns, whereas the green bars indicate those for participants' turns. Bars with a hatching pattern indicate turns coded with the expression codes.