Fiper: a Visual-based Explanation Combining Rules and Feature Importance

Visual explanations outperform text for understanding AI decisions: The paper demonstrates through user testing that combining visual representations with rules and feature importance is more effective than traditional textual explanations or raw algorithm outputs, suggesting that how we present AI reasoning matters significantly for user comprehension.

User-centered design is essential for explainable AI: The research emphasizes that explainable AI tools must be designed with actual users in mind, drawing on expertise from human-computer interaction, user interface design, and visual analytics—not just technical accuracy alone.

Feature importance paired with rules provides clearer explanations: By combining two explanation approaches (rules that show decision logic and feature importance that shows which factors mattered most), the method offers a more complete picture of why an AI system made a particular prediction.

High-stakes AI decisions require transparency tools: The paper highlights that as AI systems are increasingly used in critical domains (healthcare, criminal justice, finance, etc.), having effective explanation methods becomes a responsibility rather than an optional feature.

Visual analytics bridges the gap between black-box algorithms and human understanding: The research demonstrates that thoughtful visualization design can make complex algorithmic logic more interpretable to non-technical stakeholders, helping democratize understanding of AI systems.