Volunteers in DC are developing platforms to map climate risk and detect bots

Outside of their busy days writing code and building platforms, some DMV developers are taking time out of their schedules to assist nonprofits with technical issues. 

Local engineers, policy experts and nonprofit leaders met in downtown DC on Wednesday for Civic Tech DC’s inaugural demo night, where developers showed off tech they’re creating to address different civic problems. That included a platform to detect bots on social media and a program using photos of streets to map the likelihood of debilitating heat waves. 

These presentations are for people to get feedback on their progress and offer inspiration for possible solutions, per Mike Deeb, the head of the group.  

“Our mission is to teach civic tech to people, to teach people what it means to help people with technology,” Deeb told the crowd. 

Civic Tech DC, which was known as Code for DC before rebranding two years ago, is a completely volunteer-led nonprofit. Leaders work with various local and national organizations, including the Campaign Legal Center and the Atlantic Council Digital Forensic Research Lab, to help them source possible solutions among Civic Tech DC members.. 

Sometimes organizations will come in with issues that aren’t necessarily solved with technology, said Deeb. But helping them come to that conclusion is still helpful. 

“Not everything is solved with an app,” he said. 

Detecting bad actors on social media 

CIB Mango Tree is a library of open source programs to test datasets of social media activity for inauthentic behavior, such as bots, troll farms or humans meddling in online discourse. CIB stands for “coordinated inauthentic behavior,” and the majority of software’s users are nontechnical researchers.  

The platform tracks certain words or phrases, and accounts linked to said content. Repetition of specific phrases can reveal bad actors, explained Ben Sando, one of CIB Mango Tree’s leaders. The symbol of the mango is intentional because detecting this content is “low-hanging fruit.”

It’s also relatively easy to scrape data from social media sites because they follow a similar format in an Excel spreadsheet or CSV file, he said. CIB Mango Tree will then analyze the data and flag repeated phrases and users who excessively post. 

The goal is to strengthen the data researchers use since manually reviewing a spreadsheet of hateful messages online is very time-consuming. Software offers a way to automate this process, per Sando. 

He noted the tech has been helpful to researchers, including several that found a network of accounts in China posting propaganda — but only during business hours. The data even found that a significant number of account-holders took a lunch break. 

Mapping climate change risks

Street View Green View uses street-level imagery to map vegetation and identify areas susceptible to disaster during heat waves and other climate-related phenomena. It also assigns this risk using a percentage-based score.

Dan Joseph, a full-time data tech manager for the American Red Cross, is working with the Indonesian Red Cross on this project. He acknowledged that research like this commonly relies on satellite imagery, which is very costly. 

Equipment to capture street level imagery can instead be as simple as a durable digital camera with the ability to automatically tag GPS coordinates, and many options exist to do that. He’s not relying on Google Maps imagery for this work, so those interested in expanding it to a specific region could ask a local government to add a camera to the front of a garbage truck. 

Joseph sees a lot of other use cases outside of mapping vegetation. He’s expanding the tool to be employed in disaster relief, possibly to collect imagery to assess buildings in a region after a severe weather event. 

“Instead of having individual people go around with digital survey tools house by house, assess things,” he explained to the group, they could “just drive around the city real quick, collect imagery and then have a rough idea of which areas might be more damaged.”

Automating petition signature validation

Mobolaji Williams is trying to automate the petition process for ballots through the Ballot Initiative project. His inspiration came from witnessing the process of getting I-83, a measure that permits ranked-choice voting in the city and opens primaries, on the ballot in DC.

Right now, ballot validation is lengthy and expensive. Williams is streamlining the process of checking signatures for organizations that bring such measures forward during elections.

Users can upload PDF files of signatures with printed names and addresses to see if pledges are connected to registered voters, which makes the petition eligible for approval by an elections board. Williams requested the database of registered voters from the DC Board of Elections to do this work. 

In testing, he found that the detection is not entirely accurate. But it does capture 90% of valid signatures, he said, and about the same percentage of the predicted valid signatures are actually considered certrifiable. He’s giving the program a B+ at its current stage.

He plans to start reaching out to organizations outside of DC because he’s found need across the country.

“It’s very expensive and costly to organizations to submit signatures and then have measures not passed,” Williams presented. “That’s where tech fits in.”