Home Data Engineering Data News Informing Pandemic Policy With Innovative Ways To Use AI

Informing Pandemic Policy With Innovative Ways To Use AI

It’s been almost one year since the Covid-19 pandemic started. Data scientists worldwide have been analyzing data gathered during the pandemic to inform policies. As we have seen, policymaking has not been straight forward. During this time of social isolation, it’s been a great opportunity for policymakers to figure out the right approach to making sense of the data to gain flexibility in community-based policy decisions.

A Challenge That’s Worth The Effort

On Nov 17th, 2020, XPrize and Cognizant announced their Pandemic Response Challenge. As the COVID-19 vaccine may be on the horizon for many communities worldwide, it’s a great time to follow the vaccination rate to consider policies of safely reopening communities.

Leveraging Cognizant’s foundational Evolutionary AI™ research, as the first competition from XPRIZE’s AI and Data for Good Alliance, this competition has the potential to rally data science teams across multiple areas of expertise.

Babak Hodjat, VP of Evolutionary AI at Cognizant, says, “The competition does not recommend any specific type of algorithm to build the models and, judging by the expertise of the various teams, we expect a diversity of approaches to be used. What the competition does define is the API and expected input/output of the models. This is important, as we’ve designed the competition to allow for a superior Pandemic Response System to emerge from the collection of the best submissions.”

Consider Different Approaches

Since the competition does not specify a particular approach, different types of innovative approaches are encouraged. The collaborative efforts can generate new ideas for both research and further exploration of pandemic policy.

From the beginning of the pandemic, there’s been new awareness about agent-based models. These models can simulate community spread with a lot more nuance. SingularityNET’s COVID-19 Simulation Summit demonstrated that simple epidemiological models do not account for people’s behavior patterns or social interactions that are vital to policy.

As we have seen during the pandemic, as cities such as Los Angeles and New York City reopened during the summer, different types of community spread occurred. It was difficult to quantify how much “herd immunity” has already been achieved in London, Paris, Rome, Los Angeles, New York City, etc.

From SingularityNET’s simulation using agent-based modeling, it turns out that cultural norms such as “cliquishness” of people matters. Achieving the dynamic equilibrium of herd immunity requires different lockdown policies depending on the cultural nuance of “cliquishness.”

One of SingularityNET’s findings is that if social interactions’ clumpiness is above a certain threshold, then it may be perfectly fine for policymakers to be a little laxer regarding triggering and duration of lockdowns.

Above a certain threshold of clumpiness, it does not seem to matter how many cases trigger a lockdown or how long the lockdown lasts.

Policymakers can make policies that as much as possible guide people’s social interactions into large-discrete “clumps” or subgroups.

Lockdown rules based on clumpiness or cultural factors

While agent-based modeling is appropriate for simulating nuanced pandemic data, using Cognizant’s Evolutionary AI framework, more social variables can be examined using other types of models.

Hodjat says, “The Evolutionary AI approach consists of three models: Predictor, Prescriptor, and Certainty. The approach is algo-agnostic when it comes to the Predictor, and we have used algorithms ranging from Random Forest and XGBoost, to agent-based simulations, to Neural Networks, Deep Learning, Evolved Deep Networks, and Genetic Programming. For the Prescriptor, in most cases, the algorithm we use is to evolve neural networks. For the Certainty Models, which are not required but highly recommended, we use a Gaussian technique called RIO (Residual Input/Output).”

Hodjat says, “In early 2020, our Evolutionary AI team worked with Oxford and John Hopkins University data sources to apply artificial intelligence to publicly-available COVID-19 data. Our goal was to accurately model virus growth and create predictions and prescriptions for managing pandemic spread. This project laid the groundwork for the Pandemic Response Challenge and represents a step forward for the deployment of Evolutionary AI to manage virus spread on a global scale. Whether the objective is to reduce cases, minimize economic impact, or improve safety, AI-powered data can deliver better outcomes and inform new ways of thinking.”

An Opportunity To Affect Change

The pandemic caused temporary and permanent shifts in many parts of our society. As data scientists, this time has provided unprecedented opportunities to observe the changes and help policymakers devise strategies to manage future pandemics.

Organizations such as XPRIZE, Cognizant, and SingularityNET offer new opportunities for data scientists to participate in group innovation use their critical skills to help policymakers.

This article has been published from a wire agency feed without modifications to the text. Only the headline has been changed.

Source link

- Advertisment -

Most Popular

- Advertisment -