Explainable Artificial Intelligence (XAI) for Internet of Things: A Survey

Black-box AI models create trust and comprehension problems: The paper identifies that traditional AI models often don't allow users to understand how decisions are made, which is particularly problematic in critical applications where understanding the decision-making process is as important as the results themselves.

XAI is essential for high-stakes domains: The survey documents that Explainable AI has already been applied across critical sectors including healthcare, military, energy, financial, and industrial domains, indicating that interpretability is recognized as necessary where decisions significantly impact people's lives and safety.

IoT integration with XAI remains underdeveloped: Despite growing attention to XAI in other fields, the paper notes that "its integration into the IoT domain has not yet been fully defined," suggesting this is an emerging area where responsible AI practices need further development as IoT devices become more prevalent.

Systematic categorization is needed for responsible deployment: The survey organizes XAI approaches by methodology and application area, highlighting that developers and researchers need structured frameworks to guide responsible implementation of AI in IoT systems.

Open challenges remain for future work: The paper emphasizes there are still "challenging problems and open issues" in applying XAI to IoT, indicating that responsible AI in this domain requires ongoing research and development before widespread deployment.