"Enhancing AI Explainability for Non-technical Users with LLM-Driven Narrative Gamification" explores how gamification and narrative-driven interactions, powered by Large Language Models (LLMs), can enhance AI explainability for non-technical users. Our study focuses on integrating LLMs into Explainable AI (XAI) visualization technique with the goal to improve XAI visualizations' meaningfulness and relatedness for non-technical users. This prototype introduces LLM-driven conversational NPCs that guide users through complex AI concepts and XAI visual encodings, helping them understand things like model prediction process and decision boundaries in a more intuitive way.
- Accepted by CHI'25 as a Late-Breaking Work paper!
- Integrates Large Language Models (LLMs) to create narrative-driven NPCs that explain AI models and visualizations.
- Includes interactive t-SNE projections that allow users to explore model embeddings and understand AI decision-making processes.
- Produces design implications of LLM-driven gamification in improving explainability and reducing cognitive load for non-technical AI users.
Enhancing AI Explainability for Non-technical Users with LLM-Driven Narrative Gamification (ACM DL).
Short visual overview of the system.
Full walkthrough of the interaction design.
- Python
- t-SNE
- XAI Visualization
- Large Language Models (LLMs)
- Gamification
- AI Explainability
- OpenAI API
- Javascript
- D3.js
- Human-Computer Interaction
- Information Visualization
- Explainable AI
- Gamification
- t-SNE
- Large Language Models
- XAI Visualization
Yuzhe You, Jian Zhao