A Data Visualization Newsletter

As more states across the U.S. contemplate allowing businesses to reopen, many hope that doing so can jumpstart an economic rebound. The reality is, even a complete reopening is unlikely to have a big, immediate impact. Data from Opportunity Insights shows that consumers were already changing their behavior prior to mandatory stay-at-home orders; and public polling suggests that, even if businesses reopen, most people are still hesitant to resume normal activity.

Coronavirus antibody tests are likely to be a key component to reopening the economy, but can we trust them? This piece from Quartz explains why the tests are producing more false negatives and false positives right now than one might expect. As time goes on, however, there is hope that the results can become more accurate.

Even when you’re listening to Spotify by yourself, you may not truly be alone. Every second, more than 30,000 Spotify users press play on the same track as someone they’ve never met. This microsite leverages those connections in real time to find two people who are enjoying the same song together, serendipitously.

Imagine you’re building a model to predict whether someone is sick with a particular disease. You want the test to be sufficiently aggressive — so that it returns positive for most people who are sick — but not so aggressive that lots of healthy people are marked positive too. Now consider the fact that the disease may be more prevalent in certain parts of the population, and you have an even more vexing problem. This project from Google’s Adam Pearce considers the various tradeoffs that go into making a model “fair.”