health03 min read

Harnessing AI to Predict and Prevent Epidemics: A New Frontier in Network Science

Explore how Northeastern University's network scientists are developing AI tools to predict and prevent future epidemics, enhancing our understanding of complex systems.

Harnessing AI to Predict and Prevent Epidemics: A New Frontier in Network Science

In a world where the next epidemic is not a question of if, but when, researchers at Northeastern University are pioneering the development of artificial intelligence tools that could predict and potentially prevent future outbreaks. This groundbreaking work is led by Samuel Scarpino, the director of AI and life sciences at the Institute for Experiential AI, who is also a co-author of a recent paper published in Nature. The paper explores how AI can be leveraged to model future infectious disease epidemics, offering a glimpse into a future where we might outsmart the next pandemic before it even begins.

Network science, as Scarpino explains, provides a "common language" for researchers to study systems that operate on multiple scales, from cellular processes like cancer to population-level phenomena such as pandemics. This broad approach is crucial because the contextual challenges required to model events as complex as epidemics are immense. AI models often struggle with the vast amount of context needed to make accurate predictions, a challenge that Scarpino and his team are tackling head-on.

The success of AI models like ChatGPT, which handle context through advanced transformer models, highlights the potential for similar breakthroughs in epidemic modeling. However, the complexity of living systems presents a unique set of challenges. These systems, whether they are human immune systems or societal networks, are adaptive over both short and evolutionary timescales. This adaptability means that even with a deep understanding of a system's behavior, predictions can be elusive.

One of the significant hurdles in this field is interoperability—how to get diverse data types to communicate effectively. Scarpino emphasizes the need for global, interoperable data sets to create accurate models of pandemic unfoldings. He points to the recent H5N1 bird flu outbreaks as an example of the diverse data required, from USDA reports to bird migration patterns.

Despite these challenges, the potential for AI to aid in epidemic prediction is promising. While AI models currently lack the ability to explain the mechanisms behind their predictions, they can still provide valuable insights for researchers and policymakers. The goal is to shift outbreaks away from becoming full-blown epidemics by identifying and manipulating the right levers.

Scarpino likens epidemics to earthquakes—not in their predictability, but in the way they reveal missed smaller events. The challenge lies in early surveillance and understanding how frequently novel viruses with pandemic potential spill over into human populations. AI could be the key to answering these questions, providing a clearer picture of the threats we face.

In conclusion, while disease forecasting is still in its infancy compared to weather forecasting, the progress made by Scarpino and his team is significant. Their work not only enhances our understanding of complex systems but also brings us closer to a future where we can predict and prevent the next epidemic, safeguarding global health.