Highlights from NICAR 2017
Last weekend the Statesman interactives crew was in Jacksonville, Florida, for the 2017 National Institute of Computer-Assisted Reporting conference. It was great! We’re watching Chrys Wu’s annual blog post of slides and repos and IRE’s list of tipsheets. Here were some of our favorite things:
Cody
#NICAR2017 was lit.
I caught a demo of Klaxon, a Marshall Project tool that monitors web pages for changes, and we stood up a newsroom instance in about five minutes. (We call it “Klaxy,” or at least I do.) I learned how using a GenericForeignKey on a subclassed model could be a useful way to structure certain kinds of Django apps. I was mentored by my rad mentee on Javascript build toolchains.
I need to read the PDFPlumber docs. I need to try GeoPandas b/c wooooooow Roberto. I am going to start using ShoutKey for all of my temporary URL shortening needs.
I saw a river dolphin! The excellent Leah Kohlenberg taught me how to draw a flower. (Yes I am very proud of it and yes it is hanging on my fridge.)
I got to ride the Jacksonville Skyway, an automated monorail once described by ABC News as the “$200 Million Ride to Nowhere” and by The Florida Times-Union as “a Jacksonville joke for a generation.” Whatever, I got to ride a free robot rollercoaster to dinner.
I watched Jeremy Bowers say ~800,000 words in three minutes. I am still unsure, Quartz bot, whether I can swim with the tiny sharks in the St. Johns River. From Aditi Bhandari I learned some good tips on “self-care for nerds, and I am currently staring off into the middle distance and gently caressing my desk.
I learned about geocod.io and Seaborn and all the crazy ways that Noah Veltman makes gifs. I met old friends and new ones and everyone was cool and swell and good and doing amazing work, and I hope to see you all next year in Chicago.
Dan
Sessions on millennial news habits and “middleform” visual stories echoed the importance of information that provides value regardless of context. I’m still unpacking all the useful wisdom Sarah Slobin shared on designing with hierarchy, whitespace, typography and intentional interactions. The three J’s of Julia Angwin, Jeremy Merrill and Jon Keegan held a fascinating panel on Facebook, and Lam Thuy Vo and Cathy Deng demo’d some super clean and handy python scripts to grab social media data. News nerds shared their open source projects in a mini session, and the list of tools built by the community is crazy impressive.
Aditi Bhandari and Tyler Fisher made this Knight Lab alum proud with their lightning talks on self care and story models. The 20th thing the New York Times knows about elections in the Jeremy Bowers presentation – “Two candidates from the same party can compete for an office in the general election” – reminded me that Allan Shivers was the Texas gubernatorial nominee for both the Republican and Democratic parties in 1952.
Inspired by sessions on natural disasters and beat reporting, I returned from NICAR ready to analyze satellite imagery, store tweets with cron jobs and FOIA all the things.
Christian
I started my week at Tapestry Conference in St. Augustine for a full day of speakers discussing storytelling with data. There were many thought-provoking talks:
Lena Groeger of Propublica talked about bias in data, and how our own point of view can unintentionally influence our storytelling. Examples like a pothole app that didn’t properly collect some geographies because there were fewer people with smartphones in some locations around the city. Or how school ratings sites contribute to segregation as consumers use test scores as a proxy for race. She referenced Parable of the Polygons, which is a news game that really makes you think about how small, unintentional decisions can have wider, even harmful implications. In the Statesman Interactives Team, we’ve begun to think more broadly about our projects audiences, and are actively looking for tools like a11y to make them more accessible.
Nathaniel Lash of Tampa Bay Times and Failure Factories talked about how they used data to point out how a group of schools became more segregated and how students suffered, but it was the simple illustration of how only six of 160 kids one school passed both the reading and math state exam that really brought home the story.
Chris Mast of Merck Research Labs talked about using insurance data to map the introduction of the rotovirus vaccine. It had me thinking about commercial data sources we haven’t tapped into, and how they might provide insights into different parts of our society that open data.
Michelle Borkin of Northeastern University shared some interesting research about what makes a data visualization memorable. Human recognizable illustrations and clear titles seemed to be the key. This wasn’t about what makes a good viz, but how people consume them and what parts of a visualization enhances recall, even after short periods.
Neil Halloran talked about bringing emotion into data storytelling, as he did with The Fallen, one of the better interactive storytelling pieces in recent years.
There were more speakers, and they were all excellent. There was also a science fair time when different folks shared research, applications and posters about interesting ways they have used data in storytelling. It is a packed day, with almost too much to think about in the end. It’s an interesting collection of people from education, government, journalism, industry and of course Tableau, the host of the conference. It’s held every year the day before the CAR Conference opens, usually in a boutique hotel within an hour or so of the NICAR host city.
And that was all before the CAR Conference, where there was four more days of data-oriented learning and thinking.
At “Year in CAR,” we were able to share how we reported on our project on Latino representation in local government, a study that had not been done before on a state level. But that was just the start in a list of great projects, from the Atlanta Journal-Constitution’s award-winning Doctors & Sex Abuse series, where they used machine learning to cull through thousands of medical sanctions documents to find those most likely to show where a doctor had a suspended license based on a patient sex abuse charge, but were still able to practice medicine. Peter Aldous of Buzzfeed talked about Spies in the Skies and how he use R to comb through airplane transponder data to study government surveillance planes and their use. There were a ton of great projects to learn from.
In the educators huddle about two dozen college professors and adjuncts discussed how we can get data journalism skills into the hands of students earlier in their studies, and how we can share our materials to help professors who may not be as comfortable with these skills to at least introduce them as part of their curriculum.
The hands-on sessions at CAR have a direct impact on our work here at the Statesman, as we learn about new tools or how to hone others in our work. I became more comfortable with Python libraries that come from the data science community, but have become vogue in Journalism because it allows us to prove our methodologies and be open about our work. This isn’t fake news, folks … this is transparent, repeatable journalism that is proven through scientific theory.
The CAR community is such a sharing group, as was on display as Matt Stiles and others talked about NPR’s “daily graphics rig” that gives developers a tool kit to create data visualizations for storytelling. In another session, journalists from NPR, Quartz, The New York Times and the Associated Press shared how they bring Agile programming methodologies into data work, so we can improve our products and react to changes and feedback to reach readers in better ways.