AI Readiness in Healthcare through Storytelling XAI

Trustworthiness remains a critical barrier to AI adoption in healthcare: The paper identifies that despite rapid AI advancement, real-world healthcare adoption is limited primarily due to concerns about model trustworthiness and domain experts' hesitation to rely on AI predictions—highlighting that technical capability alone is insufficient for responsible AI deployment.

One-size-fits-all explanations don't work across different audiences: The research emphasizes that "explainability can mean different things to people with different backgrounds, expertise, and goals," demonstrating that effective AI governance requires tailoring explanations to specific stakeholder needs rather than providing generic technical documentation.

Storytelling can make AI more interpretable and trustworthy: By combining multi-task distillation with interpretability techniques and textual justification, the authors show that narrative-based explanations can increase trust among both domain experts and machine learning specialists, suggesting that communication style matters for responsible AI implementation.

Interpretability techniques can work with complex AI models: The paper demonstrates that distillation processes enable interpretability methods to be applied to large, complex deep learning models that would otherwise be "black boxes," making it feasible to explain sophisticated AI systems used in healthcare rather than limiting adoption to simpler, less capable models.

Responsible AI requires bridging the gap between technical and domain experts: The research shows that combining model-agnostic and model-specific interpretability methods supports collaboration between different expert groups, which is essential for ensuring AI systems are both technically sound and clinically appropriate.