Integrating Intrinsic and Extrinsic Explainability: The Relevance of Understanding Neural Networks for Human-Robot Interaction

Two-way explanations improve human-robot collaboration: The paper argues that explanations benefit not just humans understanding robots, but also robots understanding feedback from humans, creating a bidirectional learning process that leads to "better and more successful collaborations."

Standard XAI methods don't directly transfer to robots: Most explainability approaches focus on explaining deep neural networks in isolation, but these methods "are not applicable to humanoid robots," requiring new approaches tailored to robotic systems that must operate in physical environments.

Combining internal and external explanations is key: The research demonstrates that "intrinsic explanations by the robot itself" combined with "extrinsic explanations provided by the environment" work together to enable more efficient robotic behavior, suggesting a holistic approach to robot transparency.

Explainability builds trust in autonomous systems: The paper emphasizes that XAI can "help foster trust in and acceptance of intelligent and autonomous systems," which is particularly important for robots that work alongside humans in shared spaces.

Open-source platforms advance responsible robotics: The introduction of NICO, an open-source humanoid robot platform, suggests that making explainable robotics research accessible and reproducible is important for developing responsible AI systems in physical domains.