Govern Hidden Biases With Data
October 10, 2019 by Melissa Crowe
Photo credit: David/Flickr
Data is fodder for modern-day anthropology, revealing our values, biases, and priorities. It's the trails we leave behind, says data storyteller Ben Wellington.
"When people go and study old civilizations, they go through their garbage," he says. "How do you think people are going to study us? When people study us, they'll look at our data."
Wellington runs the blog I Quant NY, where he uses New York's open data to uncover surprising, fascinating — and sometimes frustrating — findings. He's also a quantitative analyst at Two Sigma and visiting assistant professor at Pratt Institute.
Thanks to Wellington, New Yorkers can pinpoint the location of the city's noisiest neighborhood. In many cases, his discoveries lead to change — after a deep dive into parking ticket data revealed that one poorly labeled parking place fire hydrant was generating $55,000 each year in tickets, the Department of Transportation updated nearby signage and markings.
Wellington explores what the data we share reveals about us. He offered practical advice for how governments can use their data and why it's important governments establish a way for the public to share their data-driven discoveries.
Data Can Reveal Our Biases
In one city animal control categorizes each animal response as normal, friendly, nervous, or dangerous.
"All armadillos are normal," quips Wellington. "Forty percent of cats are nervous."
Chart the data for how frequently each of the six animal control inspectors marks cats and dogs as dangerous, as Wellington does, and one gets a sense of the mix of cat and dog people in the department.
The animal control dataset can also reveal more serious findings, such as just how subjective data can be.
One point that Wellington digs from the dataset: Black cats are reported as dangerous three times as often as white cats. Although the correlation is not backed up by any scientific findings, Wellington encourages people to consider using data — even animal control responses — as laboratories to think about our own biases.
"I would challenge people who have access to this data in other areas to take a deeper look because it opens up a lot of interesting questions about how we think and our own biases," says Wellington.
Correlation Can Be Revealing
Plotting the data teaches residents about their communities.
For instance, in New York City, robberies spike at 3 p.m., right around when students leave school.
In other cities, he plotted 911 response times and found that for calls that aren't a matter of life or death, response times are slower around shift changes.
In Chicago, Wellington confirms what many public safety experts know: Crime rises with temperature. He also finds that when it gets really hot — say, 97 degrees out — crime begins to fall again. And, some crimes are more likely to correlated with rising mercury levels than others.
These insights can steer how cities manage resources, such as prioritizing after-school programs or re-examining police staff scheduling during heatwaves.
They're the kind of stories Wellington, and other civic analysts, couldn't have discovered in the past because data was locked away in static forms. For example, crime counts per precinct was updated monthly in a PDF file, Wellington says.
Governments Should Provide a Mechanism for Feedback
Public data can be used in unpredictable and surprising ways.
"There are things you'll never know people will do until you put it out there and give them that opportunity," Wellington says.
There's the mundane, such as when people start decorating for Christmas and that street addresses often skip the number 13.
As well as more impactful findings. Through the data, he discovered a single fire department at the root of $55,000 in parking tickets annually, and he spotted a nearly $800 million typo in the city budget.
It's not about "gotcha" moments, it's about encouraging governments to give residents a simple, straightforward channel to communicate feedback, ask questions, and point out issues.
"Think about that in your own agencies," Wellington says. "What do you want people to do with this information? We put it out there; give them an input."