Are Decision-Making "Experts" Immuned from Potential xAI Risks?

This project addresses the prevalent protocol for human-AI collaboration in decision-making processes, in which the AI system is responsible for making decisions and providing explanations via eXplainable AI, while the human partner assumes a monitoring role.

We critique this protocol due to potential issues: humans may find it challenging to remain vigilant in a monitoring capacity, and the known limitations of eXplainable AI may impede the detection of AI failures.

The objective of this research is to develop and evaluate an alternative protocol for human-AI collaboration mitigating the limitations to existing solutions.