An Investigation of COVID-19 Spreading Factors with Explainable AI Techniques

Lockdowns and contact tracing were the most effective COVID-19 interventions: Using explainable AI analysis of data from 18 countries and regions (January-April 2020), researchers found that city lockdowns and contact tracing consistently emerged as the two most impactful public health measures for reducing disease spread.

Face masks became critical for controlling transmission: Beyond lockdowns and tracing, the study identified public mask-wearing as an important factor specifically for achieving an instantaneous reproduction number (Rt) below 1, meaning each infected person infects fewer than one other person on average.

Mass testing alone was insufficient without complementary measures: While mass testing showed limited effectiveness as a standalone intervention, the research found it could become effective when combined with other measures, suggesting that testing strategies require supporting interventions to meaningfully reduce transmission.

Environmental factors like temperature played a measurable role: The analysis revealed that warmer temperatures helped reduce COVID-19 transmission, indicating that seasonal and climate conditions were relevant factors in disease spread alongside policy interventions.

Explainable AI provided transparency in identifying effective measures: By using SHAP and ECPI techniques—methods designed to show which factors most influence AI predictions—researchers could identify which interventions mattered most, demonstrating how explainable AI can help policymakers understand the reasoning behind data-driven recommendations rather than treating them as "black boxes."