Healthcare Explainable Artificial Intelligence- Possibilities, Challenges, and a Fresh Perspective on the Problem Space

computer Engineering

Despite the potential that XAI can offer to the application of AI in this business, explainable artificial intelligence (XAI) is still in the early stages of acceptance in the healthcare sector. Standards for explanations, the level of interaction between various stakeholders and the models, the implementation of quality and performance metrics, the agreement on standards for safety and accountability, its integration into clinical workflows, and IT infrastructure are just a few of the issues that still need to be resolved. There are two goals for this Paper. The first one involves summarizing the findings of a literature review and highlighting the current state of explain ability, including any gaps, difficulties, and opportunities for XAI in the healthcare sector. We advise using a combined taxonomy to group explain-ability methodologies in order to facilitate understanding and onboarding into this field of study. The second goal is to determine whether using a novel strategy to analyze the explainability problem space through a particular problem or domain lens and automating that method in an AutoML-like manner would help reduce the issues outlined above. The literature has a propensity to view the explainability of AI through a model-first lens, which ignores real-world issues and domains. For instance, the explainability of a patient's survival model is handled similarly to the calculation of a hospital's procedure cost. We can (semi-)automatically find appropriate models, optimize their parameters and their explanations, metrics, stakeholders, safety/accountability level, and suggest ways of integrating them into clinical workflow when the problem or domain to which XAI should be applied is clearly defined.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.