Glitter

Liwen He, Zhaowen Deng, Weibo Li, Ziheng Tang, Yixuan Li, Yutong Ren, Xin Tong

Abstract

This project introduces Glitter, an LLM-powered virtual assistant designed to assist practitioners in planning, executing, and refining ABA-integrated interventions. Through co-design iterations and a one-week field study with 11 practitioners, we evaluated Glitter’s usability, effectiveness, and potential for enhancing intervention delivery. The findings highlight Glitter as an adaptable tool while revealing key challenges and future directions. This work contributes insights into the design of virtual assistants for behavioral interventions and the integration of LLMs into practitioner support tools.

Content

The platform features a self-directed learning module and an LLM-powered agent, offering tools such as reinforcement strategies, prompting, video modeling, and the Picture Exchange Communication System (PECS). Through iterative design and user feedback, Glitter aims to enhance practitioners’ ability to deliver personalized, effective interventions in supportive, one-on-one settings.

Loading...