Written by: Illah Nourbakhsh
When coronavirus became a household word practically overnight, the idea of dangerous particles traveling in the air suddenly became very real for people around the world as the threat of COVID-19 reached into our communities, our homes, and our nightmares. How—and how far—can the virus particles travel? How long can they stay? How can I stay safe? For most people, those questions were a completely new concept.
Here at Carnegie Mellon’s CREATE Lab, however, the concept of dangerous airborne particulates was already very familiar territory. While COVID-19 added a new virus to the equation, for years much of our research has been focused on creating methods to effectively gather, visualize, and interactively evaluate data related to airborne particulates on a massive scale. One of our key missions has been to use this data, combined with the power of AI and deep machine learning, to invert the power relationship between governments and citizens to influence public policy.
In early July, the World Health Organization (WHO) confirmed the presence of emerging evidence that the novel coronavirus may be transmitted via airborne particles. The announcement came after an international plea by more than 200 scientists who signed a petition urging the WHO to update their guidance—a move that could finally move the needle on public policy in the US and around the world. As someone who has spent much of my career studying the impact of airborne particles on the health of our communities, as well as the inequities created by how governments respond to volatile chemicals in the air we breathe, my response is simple: It’s about time—because clean air needs to be a human right.
For those of us in the scientific community, the potential of airborne transmission of COVID-19 is not a new idea. While world leaders, the WHO, and the CDC have continued to focus on transmission through large droplets that are expelled when an infected person coughs, sneezes, or breathes within close range of others for an extended period of time, scientists have been researching the potential for transmission via fine particulates since the earliest days of the pandemic. In April, Nature published the article Aerodynamic analysis of SARS-CoV-2 in two Wuhan hospitals. In May, Scientific American published the article How Coronavirus Spreads through the Air: What We Know So Far. In June, a team led by Joshua Santarpia, a microbiologist at the University of Nebraska Medical Center, released a pre-print, non-peer-reviewed paper titled Aerosol and Surface Transmission Potential of SARS-CoV-2. Each of these papers and many others point to the very real possibility of aerosol transmission of the virus.
When COVID-19 first emerged as a threat, our team at the CREATE Lab immediately paused our research and turned our attention to adding to the body of scientific knowledge that could help address the challenges of COVID-19. Using our existing knowledge and sophisticated AI and deep machine learning tools, we shifted our research toward exploring two critical areas: how the virus travels through the air, and how the pandemic is impacting marginalized communities.
Identifying And Measuring Airborne Particulates
To understand how identifying and measuring airborne particulates can effect change, we can look at the example of the Shenango Coke Works on Neville Island, not far from our lab here at Carnegie Mellon in Pittsburgh. In 2011, many of the 70,000 residents living in communities downwind from the Shenango Coke Works were complaining about a multitude of issues that seemed to stem from pollutants from the factory that baked coal to produce coke for steelmaking. Their porches were covered in soot. Many had a consistent bad taste in their mouths. Childhood asthma levels in the local school district had become the worst in the state. But without hard evidence that the pollutants causing the problems were coming directly from Shenango, forcing change had, so far, been impossible.
To gather the evidence needed to convince the health department and the governor of the reality of the problem, we enlisted the help of community scientists to feed our Breathe Cam with images to record what was actually happening. The Breathe Cam enabled local citizens to use cameras installed on their own properties to capture evidence of the illegal furtive emissions from the factory. By consolidating the thousands of collected images, our team was able to apply the power of AI to effectively measure the levels of dangerous airborne particles, and then visualize the pollutants to help tell the story. We empowered the local citizens to present that compelling data, combined with volumes of personal anecdotes from the affected communities, to the county health department and local officials. This undeniable evidence shone light on the problem and served as a catalyst for change: within six months, the Shenango Coke Works was shuttered for good.
In a society where business interests too often use high-powered legal teams and seemingly bottomless coffers to disarm the people, arming citizens with AI technology can be the key to gathering real, quantitative data to support personal, emotionally compelling narratives. This powerful combination can create the undeniable evidence needed to shed light on important issues. In my experience, it is access to scientific data and the ability to communicate that information visually that creates real change.
In the face of COVID-19, we are putting these tools to work to measure particulates using existing technology we developed to measure VOCs (Volatile Organic Chemicals) in the air at a localized level. Using a device the size of a human hand, we are able to place chips that detect VOCs wherever we want to track pollution. These can be used as locally as within a single room to detect humidity, temperature, and the presence of mold spores (a true life saver for people with compromised lungs and other ailments); or as broadly as across an entire neighborhood using a dozen or more devices to measure specific pollutants.
As researchers continue to investigate the potential for aerosol transmission of COVID-19 virus particles, this type of AI technology may be the key to providing the data needed to shift this possibility from theory to demonstrated phenomenon. As I wrote in my March article for ROBO Global, To avoid a COVID-19 replay, robotics & AI technologies will come to the rescue, “Applying these innovations toward pandemic-related data analytics and the efficient dissemination of this vital information is a critical next step.” Again, once such a phenomenon is demonstrated and visualized, change is truly possible.
Addressing The Impact On Vulnerable Communities
Understanding how the virus is being transmitted is a critical goal, but just as important is measuring the impact of the pandemic on the people and communities that have been most affected by the lockdown. Understanding that impact—in all its facets—is the key to driving public policies that help protect marginalized individuals and families.
One of our most recent successes is a project that measured eviction trends from neighborhood to neighborhood across the Pittsburgh metropolitan area. Throughout June and into early July, our team worked in deep cooperation with the community to gather information on pending housing evictions and foreclosures. In collaboration with the United Way, the UrbanKind Institute, and other community organizations that were introduced to us through the Heinz Foundation, we looked at factors such as race, demography, and real-time housing information to explore how these factors intersected. In short, we used AI to visualize injustice. This mission became extremely urgent as Pennsylvania’s July 10 deadline for the moratorium on evictions loomed, with some members of our team working 90-hour weeks in order to cull the data and argue the case for an extension. On July 9, just a day short of the deadline, Gov. Wolf extended the statewide moratorium until August 31.
While one month may not offer much breathing room for those who remain unemployed, we are continuing to work with the Governor’s office, providing AI data to help policymakers see and understand pandemic-related issues that are creating problems in our communities—and to make wise decisions based on that information. We’re fortunate to have a state government that considers this important. Early in the pandemic, Gov. Wolf’s office reached out to Carnegie Mellon’s Heinz College of Information Systems and Public Policy for guidance and research to help drive its decisions regarding the re-opening of the economy. Then and now, that includes not only looking at how to increase GDP and improve other financial factors, but also understanding how to minimize the negative impact on our most vulnerable communities. For example, while the decision to re-open restaurants may appear to be vital to keeping small businesses afloat and re-employing lower-income restaurant staff, it is vital to carefully evaluate the availability of local, affordable childcare for these workers, many of whom are not able to return to work without care for their children. When we encouraged our government to consider inequalities such as these in their decision-making process, they welcomed this additional input. Since then, they have been using the data we provide to look carefully at multi-week trends and help guide their decisions about when and how to re-open Pennsylvania’s economy.
Using AI And Visualization Tools To Create A Better Tomorrow
The COVID-19 pandemic has been an important reminder of the importance of our ongoing work in AI, machine learning, and visualization. While few question the value of data for fueling better, more informed decisions—about public policy or anything else—the fact is that there is simply too much data available for our human brains to gather, digest, and make sense of it all. That’s why we are continuing to apply AI whenever and wherever possible to use real data to uncover real answers. Today, we are using web cams around the world to gather images and determine the relationship between how many people in a community are wearing masks and the real rates of new COVID-19 cases and deaths. Here, and everywhere we look, data is the key to understanding. And using AI and deep machine learning to make that data to inform and educate decision-makers around the globe is the key, hopefully, to driving meaningful change for decades to come.