Demystifying COVID-19 Models With Open Source Tools at Galois
Viz Responsibly is a new video series hosted by the Data Visualization Society, featuring interviews with people collecting, managing, analyzing, visualizing, and using data with a social impact. If you have a recommendation for a great story to share through this series, please send your recommendation to hello@datavisualizationsociety.org or send a note on slack to Amanda Makulec.
The COVID-19 models and associated visualizations have been held up by some policymakers as our roadmap to re-opening. But do decision-makers understand the nuances of those models?
Transparency around limitations and methods vary widely, though, and we now have amateur data scientists around the world trying to make sense of this information. The available models have also shown just how challenging it can be to make accurate forecasts based on fundamentally uncertain and incomplete data, with predictions varying widely depending on the methods used.
Galois, an employee-owned research lab in Arlington, VA, has developed an open source platform to support peer review and exploration of existing COVID-19 models. The tools enable anyone to view data on a simple dashboard and provides a more detailed interface for scientists to support peer review of COVID-19 models.
Over the last two years, Eric Davis and his team have been working on a tool to improve our ability to respond with data in a crisis situation. Their work became painfully relevant as COVID-19 emerged, and the team has stepped up to help vet existing models, make them more accessible to policymakers and domain scientists, and learn and improve their tool in real time.
Rather than focus on building new models, the team at Galois has looked to find equations and models from papers and turn them into executable code. The team aims to empower scientists and practitioners to run these models without the need to engage with software engineers. At the same time, the platform can be used by other model developers to quickly analyze code and check for errors.
The tool’s interface includes a lightweight dashboard of key metrics (including cases, deaths, social mobility, and testing), a graphic user interface for running code, and an innovative icon-driven interface that generates code as you place and maneuver icons on a screen. The team has focused on enabling users to drill down to the county level, as state aggregations can be dominated by stories from the urban centers, which has proven useful for state governments in places including Ohio and Oregon, and at the federal level in the US and abroad.
Recognizing pressing current questions around re-opening, the team has been trying to address the need for information on early warning signs of reemergence. On the dashboard, spikes in county mobility are frequently followed by a spike in new cases; by tracking these metrics in one place, decision-makers can better anticipate future spikes.
Learn more about the team’s open-source approach, see a demo of the available tools, and hear about how these tools are being used by federal, state, and local governments around the world as part of their response to COVID-19. You can explore the tools Galois is developing on their website.
