OnGoal: Tracking and Visualizing Conversational Goals in Multi-Turn Dialogue with Large Language ...
summary
OnGoal is a chat interface that automatically infers, merges, evaluates conversational goals from each user message, generates LLM explanations and extractive examples, and shows three temporal visualizations plus a goal history view. A 20 participant study on a writing task finds users spend less time and effort and explore new prompting strategies. The system is built as an LLM pipeline but some participants want manual goal editing and feedback mechanisms.
AI-written summary · about
chapters (8)
- 0:00Introduction to OnGoal
- 1:05Limitations of linear chat interfaces
- 2:28Design opportunities for OnGoal
- 3:09Goal communication pipeline
- 4:13Visualizing goal evaluations
- 6:18Goal alignment and history view
- 8:04User study insights and future work
- 8:47Q&A session
related talks
32mEvaluating multi-turn conversations
01h 04mBuilding Expertise Into Generated Conversations
16mWhere can my career take me?: harnessing dialogue for interactive career goal recommendations
01h 04mBuilding Expertise Into Generated Conversations
03mInterview
25mBuilding the shopping journeys of the future - ChatGPT Apps, Adaptive Interfaces and more...
51mChatGPT as a developer's personal servant
51mA ChatGPT on your own data: how does it work?
