ExplainabilityAccountabilityTransparencycs.ET
IoXT: The Internet of Explainable Things. Why Explainability in IoT Requires a New System-Level Paradigm and Protocol Design
Michael Georgiades
August 9, 2026
Key Takeaways
-
Explainability requires end-to-end system design, not just better AI models: The paper argues that existing explainable AI (XAI) methods can only explain individual model predictions, but cannot explain the complete path from what sensors observe to what physical actions occur in the real world. This means organizations cannot achieve true explainability by simply upgrading their AI algorithms alone—they need to redesign their entire IoT systems.
-
"No Orphan Actuation" principle ensures accountability for real-world consequences: IoXT introduces a requirement that every physical action taken by an IoT system must remain traceable back to the decision that authorized it and the sensor evidence that informed it, or be explicitly marked as degraded or non-conformant. This prevents situations where systems take consequential actions in the physical world without anyone being able to explain why.
-
Explainability must be built into system architecture from the start: Rather than treating explainability as an add-on feature, IoXT proposes making it an "architectural property" of intelligent IoT systems. This means explainability considerations should guide fundamental design choices about how sensors, communication networks, AI systems, and actuators are connected and interact.
-
Cross-layer traceability and temporal accuracy are critical: The framework emphasizes the need for "bidirectional traceability" and "temporal fidelity" across all system layers, meaning organizations must be able to track information flowing both forward (from sensors to actions) and backward (from actions to their origins), while maintaining precise timestamps throughout.
-
Security and privacy must be integrated with explainability: IoXT identifies security and privacy as core design principles alongside explainability, recognizing that making systems more transparent cannot come at the cost of exposing sensitive data or creating security vulnerabilities.