Beyond ChatBots: ExploreLLM for Structured Thoughts and Personalized Model Responses (2312.00763v1)

Abstract: LLM powered chatbots are primarily text-based today, and impose a large interactional cognitive load, especially for exploratory or sensemaking tasks such as planning a trip or learning about a new city. Because the interaction is textual, users have little scaffolding in the way of structure, informational "scent", or ability to specify high-level preferences or goals. We introduce ExploreLLM that allows users to structure thoughts, help explore different options, navigate through the choices and recommendations, and to more easily steer models to generate more personalized responses. We conduct a user study and show that users find it helpful to use ExploreLLM for exploratory or planning tasks, because it provides a useful schema-like structure to the task, and guides users in planning. The study also suggests that users can more easily personalize responses with high-level preferences with ExploreLLM. Together, ExploreLLM points to a future where users interact with LLMs beyond the form of chatbots, and instead designed to support complex user tasks with a tighter integration between natural language and graphical user interfaces.

Introduction

LLMs powering chatbots have revolutionized AI system interactions, enhancing user experiences with natural language processing capabilities. However, existing interfaces for these chatbots typically present linear, text-based conversations that can hinder users as they attempt to navigate intricate tasks like planning trips or researching new topics.

Task Decomposition and User Interface

To address these challenges, a system called ExploreLLM has been developed, embracing a novel interaction pattern that combines LLMs with a graphical user interface (GUI) designed around schema-like structures. This approach automatically breaks down users' complex tasks into more manageable sub-tasks, which are then presented in an interactive, structured format. For instance, when a user wants to plan a trip to Tokyo, ExploreLLM segments the task into items like date selection, hotel booking, and travel document checks, thereby simplifying the process and reducing cognitive load.

Personalization and Interaction

Personalization is another critical component of ExploreLLM. Users are encouraged to provide high-level preferences and personal context, which are globally taken into account by the system when generating responses and options for each sub-task. This allows for more tailored recommendations and outcomes. ExploreLLM facilitates user engagement with the options through an interactive UI, where they can select their preferences to steer the model towards more customized solutions.

User Study Findings

A user paper involving eight participants provides insights into ExploreLLM's effectiveness in structured guidance for planning tasks compared to traditional chatbots. The paper underscores the system's ability to organize users' thoughts and present personalized options, enhancing the planning experience. Users expressed appreciation for this structured guidance, which contrasts the more general and verbose responses from text-based chatbots. Despite some technical and usability limitations, the paper indicates a positive user reception for ExploreLLM's approach to complex task interaction and emphasizes the potential benefits of integrating LLMs with more traditional GUI elements.