Third-party cyber evaluations involving OpenAI models
openai.com·13h ago
AI Research
arXiv CS.AI·Jun 9·5 min read
Self-Explainability in Self-Adaptive and Self-Organising Systems: Status and Research Directions
TL;DR
Self-adaptive and self-organising systems are becoming complex, making them hard to understand and trust. This research presents a systematic literature review on Self-Explainability (SX), defining it and introducing a taxonomy and levels of SX.
✦ Why It Matters
Engineers can leverage the defined taxonomy and levels of Self-Explainability to guide future system designs.
Key Takeaways
- Explore existing self-explainability frameworks for your projects.
- Assess the current state of SX in your AI systems.
- Identify gaps in practical implementations of SX.
- Advocate for the establishment of evaluation standards in SX.
How It Works
The review introduces a taxonomy of Self-Explainability (SX) that categorizes approaches based on their domains and targets. It also defines Levels of Self-Explainability, which provide a structured way to evaluate and position current and future research efforts in this area.