Hide captions
View caption
Akarapon Chairman / istock
Francesca Dominici, Clarence James Gamble Professor of Biostatistics, Population and Data Science, faculty director of the Harvard Chang Kang School of Public School and Harvard Data Science Initiative and her research team are helping to develop artificial intelligence (AI) and machine learning models to help people improve resilience to environmental stress and health threats from extreme weather. They are also studying AI’s carbon footprint and developing an approach for responsible and sustainable deployment of AI. She recently spoke about a paper she co-authored in nature. This explored ways in which similar methods could communicate decision-making during an outbreak of an infectious disease.
Q: What challenges arise around infectious disease modeling during outbreaks? Also, how can AI help?
A: One of the key challenges at the onset of the epidemic is answering questions about the severity and permeability of infectious pathogens as the infection wave actually progresses.
In traditional epidemiological analysis, some of these questions can be answered from strictly controlled studies. However, despite major efforts to document, for example, what is happening during outbreaks using contact tracing, the specificity of the data ensures that true epidemic processes are incompletely observed.
The actual chain of infection events and where they occur are often vague. Individuals may visit multiple locations and meet different people.
AI can accelerate breakthroughs in answering important epidemiological questions, including data processing and analysis, speed and efficiency, increased accuracy, and integration of multiple data sources, including health records, real-time surveillance data, and environmental factors, and more comprehensive and accurate epidemic prediction.
Q: What role do data quality and availability play in the effectiveness of AI-driven epidemic models?
A: Data quality is important. AI models should be trained with representative data that can capture the characteristics of infectious diseases. The good news is that new AI approaches can work more and more with limited data. No months of initial training or terabytes of data is required to achieve cutting-edge performance.
Although it is difficult to assess human behavior during development, AI models can be explained easily if they have data. For example, data on people’s movements have been widely used and analyzed in the Covid-19 context, as well as data on vaccinations, mask use and willingness to avoid gatherings. AI can quickly learn and explain the complexity of these interaction processes.
Q: What ethical questions should we consider when using AI to inform you of infection prevention and control efforts?
A: The set of questions concerns the importance of AI tools that are equitably shared for use by public health authorities. The effectiveness of this depends on the development and sharing of expertise within a collaborative approach to monitoring and analysis.
The second set of questions concerns how AI tools are deployed in the design and implementation of public health policies. An important lesson from Covid-19 was that all policy decisions lie in core value judgments with strong ethical components regarding the limitations of privacy and freedom in the use of vaccines, for example, the distribution of vaccines, or the use of digital contact tracing. Such judgments must be subject to deliberation, justified and accountable.
This paper discusses new methodologies in AI improve the collection and merging of key data and how they are included in decision-making frameworks to improve population health. These advances must also be fair to avoid deepening health inequality.
Demonstrating the effectiveness of AI in improving policy decisions that benefit population health remains one of the biggest challenges. For AI to succeed in that respect, there is a growing need for close cooperation between researchers, policymakers and society over the next few years.
Last updated
It is featured in this article
Francesca Dominici
