On the Influence of Explainable AI on Automation Bias

Explainable AI may help reduce over-reliance on AI systems: The paper investigates whether providing explanations for AI decisions can mitigate "automation bias"—the tendency for humans to over-trust and blindly follow AI recommendations—suggesting that transparency in AI reasoning could be a practical tool for safer human-AI collaboration.

Automation bias is a real concern in high-stakes fields: The research highlights that automation bias poses particular risks in critical domains like medicine and banking, where uncritical acceptance of AI advice could have serious consequences, making this a key consideration for responsible AI deployment.

Current understanding of human-AI collaboration is limited: The authors note that "insights on the effective collaboration of humans and AI are still rare," indicating that organizations implementing AI systems may lack sufficient guidance on how to design systems that promote appropriate human judgment rather than blind deference to AI.

Explainability alone may not be sufficient: While the paper proposes XAI as a potential solution, it frames this as an initial investigation ("pre-test"), suggesting that simply explaining AI decisions might not fully solve automation bias and that additional safeguards or design approaches may be needed for truly safe hybrid intelligence systems.

Practical testing is essential before deployment: The authors conduct real-world experiments (using hotel review classification tasks) rather than relying on theory alone, emphasizing that responsible AI requires empirical validation of whether proposed solutions like explainability actually change how humans interact with AI systems.