Designing a Proactive Context-Aware AI Chatbot for People's Long-Term Goals
Brennan Jones, Yan Xu, Qisheng Li, Stefan Scherer
Abstract
When pursuing new complex goals such as fitness or sustainability, people often seek advice from various sources. Large language models (LLMs) such as ChatGPT have recently emerged as popular sources for information seeking, action discovery, and goal planning. However, such tools require users to provide detailed prompts, are not adaptive to the user’s personal attributes or real-time contexts, and are merely reactive to the user’s prompts rather than proactively guiding the user at opportune moments. We share the design of an LLM-based chatbot app that proactively recommends actions to the user for their goals based on context factors that can be detected or inferred by the user’s smartphone (e.g., location, time, weather) and the user’s personal profile. An early pilot field study reveals that participants enjoyed the chatbot as a personal assistant that was adaptable and flexible to their needs and kept them motivated by discovering actions toward their goals.