Guidelines for designing explainable ML systems
In this section, we will discuss the recommended guidelines for designing an explainable ML system from an industry perspective while considering the open challenges of XAI, as discussed in the previous section. All of these guidelines have been carefully collated from various publications, conference keynotes, and panel discussions from various experts in the field of XAI, ML, and software systems. It is true that every ML and AI problem is unique in its own way, and so, it is hard to generalize any recommendations. But many AI organizations have adopted the following list of guidelines for designing explainable and user-friendly ML systems:
- Identify the target audience of XAI and their usability context: The definition of explainability depends on the user using the AI system. Arrieta et al., in their work Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities, and Challenges toward Responsible AI, have...