Abstract
The SARS-CoV-2 virus, the COVID-19 disease, and the resulting pandemic have reshaped the entire world in an unprecedented manner. Massive efforts have been made by AI communities to combat the pandemic. What roles has AI played in tackling COVID-19? How has AI performed in the battle against COVID-19? Where are the gaps and opportunities? What lessons can we learn to enhance the ability of AI to battle future pandemics? These questions, despite being fundamental, are yet to be answered in full or systematically. They need to be addressed by AI communities as a priority despite the easing of the omicron infectiousness and threat. This article reviews these issues with reflections on global AI research and the literature on tackling COVID-19. It is envisaged that the demand and priority of developing “pandemic AI” will increase over time, with smart global epidemic early warning systems to be developed by a global collaborative AI effort.
Introduction
When the COVID-19 pandemic emerged, the global artificial intelligence (AI) community responded with unprecedented speed and volume. Thousands of models were developed for tasks ranging from X-ray diagnosis to spread prediction. However, two years into the crisis, a critical gap remains between the sheer volume of AI research and its actual impact on pandemic management. This paper addresses this disconnect by performing a systematic "post-mortem" of the global AI research landscape, identifying why many efforts remained reactive and siloed, and proposing a roadmap for a more robust "Pandemic AI."
Methodology and Data Corpus
To map the AI research landscape, a comprehensive meta-analysis was conducted using Natural Language Processing (NLP) to categorize and analyze global research output. The study analyzed a massive corpus of data extracted between January 2020 and December 2021, including 281,718 formal publications and 24,271 preprints sourced from Web of Science, Scopus, PubMed, and various preprint servers. The analysis focused specifically on a subset of 38,730 AI-related studies to compare disciplinary approaches and regional contributions.
Global Contributions and Research Trends
The global response was led by the USA in terms of publication volume, followed by China, Italy, the UK, and India. While volume was high, publication quality (measured by H5-index and CiteScore) was particularly notable in the UK, Germany, and France among G20 nations.
Research priorities shifted significantly as the pandemic evolved. Early efforts were dominated by case and transmission prediction. However, by late 2021, the focus transitioned toward vaccines, mental health, and mortality prediction, reflecting the changing needs of global public health.
Key Insights: The Disciplinary Divide and Complex Systems
A significant finding is the "disciplinary divide" between computer science and medical researchers. Word cloud analysis revealed that computer scientists focused heavily on "Simulation," "Deep Learning," and "Neural Networks," while the medical community favored "Regression Models" and "Statistics." This lack of cross-pollination suggests that medical researchers prioritized interpretability, while AI researchers focused on raw predictive power, often at the expense of clinical utility.
Furthermore, the author argues that COVID-19 must be viewed as an Open Complex System. Most current AI models are reactive and fail to account for the complex interactions between biological evolution, social hierarchy, and economic variables. Much of the research is characterized as "shallow and observational," often producing "me-too" models (such as redundant X-ray classifiers) rather than addressing harder problems like de novo drug discovery or the socio-economic impacts of lockdowns.
Future Directions: Toward "Pandemic AI"
To better prepare for future "Gray Rhino" (probable but ignored) and "Black Swan" (unpredictable) events, the paper proposes the development of a dedicated sub-field: Pandemic AI. Key pillars of this roadmap include:
- Global Collaborative Early Warning Systems: Moving beyond national silos to create integrated global monitoring.
- Ethical Management: Developing AI that can detect privacy-violating or anti-humanity policies.
- Global Benchmarking: Establishing standards to ensure models developed in one region are reliable and applicable in another.
While AI served as a useful "soldier" during COVID-19, it has yet to become the strategic "general" required to lead the defense against future global health crises.
Summary updated