The goal of this project was to investigate how the number of reported cases of virus infection depends on testing policies, and why those numbers are not as reliable as they seem.
A lot of work has been done on representing the propagation of the virus over time, but these representations are limited by the available data on the coronavirus, which is dependent on the testing coverage for a given population.
Instead of focusing on those unreliable numbers, we used a popular statistical model to simulate the propagation of a disease: the SEIR model. We generated a population with a few initial cases and we observed how the disease spread over time.
Then, we defined and applied two different testing policies on this population, with different characteristics, to illustrate the testing policies of different regions of the world. Those policies led to different numbers of reported cases. To show how those different numbers can influence the discourse on fighting the pandemic, we then calculated the fatality rate associated with the numbers reported for each policy: the fatality rate associated with the policy that tested fewer people was more than three times higher than the actual fatality rate of our simulated virus.
Since those statistics are essential in responding to the pandemic, it is crucial to get those numbers right. Our visualization highlights how testing policies are decisive in that regard.
Even though our narrative here is focused on the theoretical aspect of testing and not on real-life numbers, we wanted to make it as tangible as possible for the reader. For this reason, we chose to divide the screen between visualization and narration and to always use the visualization to support our explanations. This was made possible by using the JS library scrollama, which was combined with [D3](https://d3js.org/) to implement our scrollytelling article.
To make our design more engaging, we wanted to make it as smooth as possible. For this purpose, we initially kept the annotations on the visualizations to a minimum, as we expected that the juxtaposed narrative would provide the user with all the context elements required to understand the visualizations through judicious color coding. We also spent a long time on refining the transitions from one visualization to the other: we wanted those transitions to keep the user's focus on the data so that it would be easy to see how successive visualizations were related, and we also wanted them to look interesting so that the user would want to read further, without being overly distracting.
Overall, we received positive feedback on the transitions and on the general visual impression of our visualization, which seemed to be a compelling part of our work. However, we also learned that the absence of legends and captions next to the visualizations made it hard to grasp the representations' purpose at once and that it caused some confusion for the reader. For this reason, we added some labels to our final work.
Additionally, some of our users expressed an interest in being able to manipulate the simulation's parameters themselves, in order to get a better sense of how the model worked. We had initially considered letting users modify the simulation's parameters but had deemed it too distracting from our global narrative. As a compromise, we decided to add a sandbox environment at the end of the visualization where the user was free to manipulate the model.
Overall, this was a very satisfying project to work on for both of us, as it brought to our attention many design considerations whose influence we had not immediately suspected. However, there still are several possible improvements that could add value to this work. One of those would be to tie in the numbers that are being reported in the world as of now to our visualization, and to illustrate how those numbers are influenced by different countries' testing policies. For now, our incomplete understanding of the pandemic makes it hard to precisely characterize how testing policies have influenced the virus' propagation through the world, but in the future and with more data available, such an addition could help emphasize how testing policy has a very concrete impact on people's lives.