Explainable AI for Next-Generation Wireless Physical Layer: Basics, State-of-the-Art, and Open Challenges

AI is becoming central to wireless networks, but we can't trust what we don't understand: The paper emphasizes that as neural networks become embedded throughout wireless communication systems to improve speed and efficiency, the "black box" nature of these AI models creates serious concerns about reliability, safety, and privacy—especially when network conditions are unpredictable and constantly changing.

Explainability requires clear responsibility goals, not just technical fixes: Rather than applying explanation techniques randomly, the authors argue that wireless systems need to first establish "responsibility-oriented goals" for XAI, meaning organizations should define upfront what they need to explain and why, based on safety, privacy, and reliability concerns specific to their use case.

There's a real tradeoff between AI performance and explainability: The paper identifies a key open challenge: making AI systems more explainable often comes at the cost of reduced performance. This means deploying trustworthy AI in wireless systems requires difficult decisions about acceptable performance losses in exchange for transparency.

Different communication problems need different explanation approaches: The survey demonstrates that explanation techniques must be customized to wireless-specific structures and challenges rather than applying generic AI explanation methods, suggesting that responsible AI deployment requires domain expertise and tailored solutions.

Emerging AI systems (like large language models) create new explanation challenges: The paper flags that next-generation approaches using large language models and autonomous AI agents will introduce additional complexity in explaining decisions, requiring the field to develop new explanation methods beyond current techniques.