JEDAI: A System for Skill-Aligned Explainable Robot Planning
Project Overview
The document presents JEDAI, a generative AI system designed to improve educational outreach for non-experts in artificial intelligence by facilitating their interaction with robots. JEDAI combines task and motion planning with explainable AI, enabling users to generate executable plans for robots while enhancing their comprehension of AI's functionalities and constraints. By offering personalized explanations of errors and an intuitive user interface, JEDAI helps demystify the complexities of AI, making it accessible to a wider audience. Key applications include fostering user engagement with robotics and improving learning outcomes by tailoring the educational experience to individual needs. Findings suggest that such systems can significantly enhance understanding and interaction with AI technologies, ultimately contributing to a more informed and capable society in the context of rapid advancements in artificial intelligence.
Key Applications
JEDAI - an AI system for educational outreach
Context: Designed for non-AI experts to learn about AI planning and robotics
Implementation: Users create high-level plans using a drag-and-drop interface with Blockly, which are then validated and converted into low-level motion plans for robots.
Outcomes: Improves user understanding of AI capabilities, enhances learning through personalized explanations, and facilitates interaction with robotic systems.
Challenges: Users may initially struggle with understanding task requirements and robot capabilities; requires effective explanations tailored to user knowledge levels.
Implementation Barriers
Usability Barrier
Non-experts may find technical language and domain model syntax challenging to understand.
Proposed Solutions: JEDAI uses natural language templates for generating user-friendly descriptions of goals, actions, and explanations.
Project Team
Naman Shah
Researcher
Pulkit Verma
Researcher
Trevor Angle
Researcher
Siddharth Srivastava
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Naman Shah, Pulkit Verma, Trevor Angle, Siddharth Srivastava
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai