Research map / Large language models / AI and society
Using Explainable AI and Hierarchical Planning for Outreach with Robots
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Presents JEDAI.Ed, an open-source web platform that teaches novices robot planning via a Blockly visual interface, explainable failure feedback (HELM/VAL with GPT-3.5 translation), and adaptive curriculum generation, validated by a 42-student user study.
Problem
Teaching non-experts such as K-12 students and novices how robots plan and execute high-level tasks is difficult because motion planning and robot programming require substantial technical knowledge.
Approach
Combines a block-based high-level instruction interface with integrated hierarchical planning and simulation, formal explanations of plan failures (HELM, VAL) rendered in natural language by an LLM acting only as a translator, and an adaptive curriculum that adjusts action costs from user performance to generate targeted problems via BFS.
Key results
- In a user study with 42 novice university students, JEDAI.Ed users solved the test task in 235 seconds, about 20% faster than JEDAI.
- 95% of JEDAI.Ed users agreed the system made robot programming easier, versus 71% positive sentiment for JEDAI.
- 66% of users found JEDAI.Ed's UI intuitive compared with 28% positive sentiment for JEDAI.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use JEDAI.Ed instead of JEDAI | JEDAI.Ed · User study, Q5 curiosity | Likert mean: 3.00 $\pm$ 0.89 -> 3.47$\pm$ 0.60 | Helped |
| Use JEDAI.Ed instead of JEDAI | JEDAI.Ed · User study, Q7 easier programming | Likert mean: 3.04 $\pm$ 0.80 -> 3.47 $\pm$ 0.60 | Helped |
| Use JEDAI.Ed instead of JEDAI | JEDAI.Ed · User study, Q8 improved understanding | Likert mean: 2.90 $\pm$ 0.76 -> 3.23 $\pm$ 0.62 | Helped |
| Use JEDAI.Ed's explanations instead of JEDAI's | JEDAI.Ed · User study, Q2 helpful explanations | Likert mean: 2.42 $\pm$ 1.20 -> 3.38 $\pm$ 0.86 | Helped |
| Use JEDAI.Ed's interface instead of JEDAI's | JEDAI.Ed · User study, Q4 intuitive UI | Likert mean: 2.19 $\pm$ 0.87 -> 2.71 $\pm$ 0.71 | Helped |
| Use JEDAI.Ed instead of JEDAI | JEDAI.Ed · User study, Q6 programming confidence | Likert mean: 2.33 $\pm$ 1.06 -> 2.85 $\pm$ 0.79 | Helped |
| Use JEDAI.Ed instead of JEDAI | JEDAI.Ed · Coffee Shop test task | solve time: 20% faster | Helped |
| Replace adaptive problem generation with randomly generated training problems (ablation) | JEDAI.Ed · Test task after Coffee Shop training | test-task performance | Hurt |
Surprising findings
- Without adaptive problem generation (random training problems) in an ablation, solve times on the test task were similar to JEDAI, implying the explanations of JEDAI already provided much of the benefit.
Limitations
- Evaluation is limited to 42 university students with a 45-minute session, and comparison is only against its predecessor JEDAI.
- Cloud deployment introduced network latency that disrupted video streaming during the high-school pilot.
- Not accessible on devices without keyboard and mouse; loops and conditionals not yet supported.
Experimental setup
Metrics: Likert sentiment (one-sample t-test vs neutral) task solve time (seconds) percent positive sentiment
Baselines: JEDAI
Builds on: JEDAI HELM VAL Blockly FF
Closest papers by meaning
- “AlphAI”: Teaching AI Algorithms to K12 by Training Learning Robots and Visualizing How It Works (AAAI 2025 · similarity 0.87)
- Build Your Own Robot Friend: An Open-Source Learning Module for Accessible and Engaging AI Education (AAAI 2024 · similarity 0.86)
- Foundations of Autonomous Vehicles: A Curriculum Model for Developing Competencies in Artificial Intelligence and the Internet of Things for Grades 7–10 (AAAI 2024 · similarity 0.83)
- Hi Robot: Open-Ended Instruction Following with Hierarchical Vision-Language-Action Models (ICML 2025 · similarity 0.83)
- Personalized Learning in Action: Exploring AI and Robotics for Early Childhood Education (AAAI 2025 · similarity 0.82)
- Smart Motor: A Low-Cost Hardware and Software Toolkit for Introducing Supervised Machine Learning to Elementary School Students (AAAI 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.