Towards Improved XAI-Based Epidemiological Research into the Next Potential Pandemic

Area of Science:

• Epidemiology
• Artificial Intelligence
• Public Health

Background:

• Artificial intelligence (AI) has aided SARS-CoV-2 pandemic control through data analysis.
• Epidemiological machine learning models for SARS-CoV-2 are prevalent but often lack transparency.
• The 'black box' nature of AI hinders trust and confident reliance on policy recommendations.

Purpose of the Study:

• To systematically review interpretable AI-based epidemiological models for SARS-CoV-2.
• To propose a conceptual framework for AI pipelines in SARS-CoV-2 epidemiology.
• To identify research gaps and suggest improvements for AI in pandemic policy support.

Main Methods:

• Systematic literature review of studies combining AI, SARS-CoV-2 epidemiology, and explainable AI (XAI).
• Development of a conceptual framework to synthesize AI methodologies in SARS-CoV-2 studies.
• Analysis of selected epidemiological studies to identify trends and gaps.

Main Results:

• Identified a growing body of research on AI in SARS-CoV-2 epidemiology.
• Highlighted the critical need for explainability in AI models for public health policy.
• Proposed a framework to understand and advance interpretable AI in this domain.

Conclusions:

• Interpretable AI is essential for building trust in epidemiological models for pandemic response.
• Further research is needed to fill identified gaps in AI toolboxes for enhanced policy support.
• Developing explainable AI approaches will strengthen preparedness for future pandemics.