Nieman Foundation at Harvard
HOME
          
LATEST STORY
From shrimp Jesus to fake self-portraits, AI-generated images have become the latest form of social media spam
ABOUT                    SUBSCRIBE
Aug. 26, 2019, 9:56 a.m.
Aggregation & Discovery

Every crime map needs context. This USC data journalism project aims to scale it

“When we do a story looking at 18 different commuting routes, we’ve created 18 different stories, in a way — we’ve created a story with 18 distinct audiences.”

A bunch of money — check. A bunch of local data — check. Brains galore — check. Future of local news? TBD.

When University of Southern California professor and Wall Street Journal alum Gabriel Kahn was given part of a grant made to USC’s journalism school, he sat down with a professor in the computer science department. He wanted to know what that department was seeing and learning and how it could be retooled with journalism — specifically, local news.

The CS folks had been contracted with LA’s traffic department for five years to analyze its data, and a giant load of data (for instance, all the signals from road directors that can be used to calculate traffic density and speed) had been stockpiled. The data just kept on coming, but could also be broken down into smaller sets for different areas and — the magic word! — localized. “Could we make these large datasets into hundreds of locally relevant datasets and then turn that into something like a local news feed?” Kahn wondered.

It was a big ask, but Kahn — along with computer science department chair Cyrus Shahabi — had big resources to work on it: The grant provides around $120,000 per year for five years, and the team now includes part-time data scientists to crunch the numbers, designers to visualize the trends, and journalists to put it all into context. This is Crosstown LA, a nonprofit news project based at USC, but it has the potential to expand to other regions.

Crosstown uses publicly available data on LA’s traffic, air quality, and crime to analyze trends and report local — really local — stories from that. Kahn compared it to the police commissioner sharing updates on the city’s yearly crime statistics. How are the numbers impacting each neighborhood? That’s the wedge Crosstown is trying to get at, though it’s still in the developmental phase (with a plan for future monetization). Some recent stories:

  1. So far, LA crime in 2019 is dipping: “Sharp increases in housing prices in neighborhoods like West Adams have led to changes in both demographics and public safety. The neighborhood saw a 13 percent increase in crime during the first half of 2019. However, it is unclear whether this increase was due to actual higher crime or more vigilant crime monitoring by residents.”
  2. Recent quakes trigger alarm for retrofitting: “Almost four years after the ordinance passed…only 18 percent of the buildings are earthquake ready. Nearly 9,700 building owners have filed at least the initial permits to begin work. About 2,200 have hired contractors and have projects underway. And 269 owners, according to Building and Safety, haven’t done anything at all.”
  3. Suspect wore a hoodie: “In the City of Los Angeles, the hoodie appears in Los Angeles Police Department crime data to describe what suspect or suspects were wearing at the time of a reported crime: ‘Suspect wore hood/hoodie.’ The department has been cataloging this clothing descriptor for suspects since at least 2010, when it started making its data publicly available. But in the year following Martin’s death, the number of crimes reported with ‘Suspect wore hood/hoodie’ skyrocketed. In 2013, there were 1,243 reports, a 92 percent increase from 2012.”

Kahn hopes that as the team develops its system for analyzing and formulating the data (it’s automated at a minor level at this point, with, for instance, a slackbot that tracks reports of hate crimes), it can be scaled to more datasets and more locations. “When we do a story looking at 18 different commuting routes, we’ve created 18 different stories, in a way — we’ve created a story with 18 distinct audiences,” he said.

One wrinkle: The data isn’t always completely comparable; reporting differs across police departments, and certain types of theft might be broken down differently in different areas, for instance. And officers don’t always record the time of a crime, in which case it’s automatically filled in as midnight. How do you scale nuance? “We learned a lot of the idiosyncrasies of police data. You need to get close to the data and understand its flaws and limitations,” said Kahn. With practice, “we feel much more confident about what we can report on.”

So far, Crosstown has focused primarily on crime because data on it is widely available. But “we don’t want to be in the crime [news] business, we want to be in the data business,” Kahn said. (See Ring and its system for invoking fear into potential customers through journalism.) Stories on traffic and air quality are waiting in the wings.

In the case of the health and air quality stories, Crosstown consults with public health experts on how to responsibly interpret and share the data. Readers email Kahn for advice about the environmental safety of certain areas, he says, and the team wants to make sure people can draw the right conclusions from the work. It’s one reason they include a “How we did it” section at the end of each article explaining the data sources and interpretations:

KPCC and LAist have worked with Crosstown to publish reporting on traffic delays, and a Facebook Community Network grant of $25,000 will help Crosstown envision more ways to share its data and collaborate with more local news outlets in southern California. The team has also compiled its findings in one-pagers and brought them to neighborhood council meetings. In the future, Khan is thinking about developing hyperlocal newsletters and creating a membership program.

“If we can take regularly occurring data about crime, real estate, and public education and package it into a weekly newsletter,” he said, “that would be 110 different newsletters” across the city.

Screenshot of Crosstown’s safety map.

POSTED     Aug. 26, 2019, 9:56 a.m.
SEE MORE ON Aggregation & Discovery
Show tags
 
Join the 60,000 who get the freshest future-of-journalism news in our daily email.
From shrimp Jesus to fake self-portraits, AI-generated images have become the latest form of social media spam
Within days of visiting the pages — and without commenting on, liking, or following any of the material — Facebook’s algorithm recommended reams of other AI-generated content.
What journalists and independent creators can learn from each other
“The question is not about the topics but how you approach the topics.”
Deepfake detection improves when using algorithms that are more aware of demographic diversity
“Our research addresses deepfake detection algorithms’ fairness, rather than just attempting to balance the data. It offers a new approach to algorithm design that considers demographic fairness as a core aspect.”