by Michelle Porter
In 2016, the Arizona Center for Investigative Reporting (AZCIR) developed what is now known as the “AZ Dark Money Bot,” or @AZDarkMoneyBot, a Twitter bot that enables the automation of “dark money” expenditure reporting. This was created after rampant occurrences of dark money spending or the non-disclosure of donors spending millions on campaigns of certain Arizona candidates in 2014. The bot reported dark money expenditures during the 2016 electoral cycle for Arizona’s all state-level candidates in near real-time, and it has garnered public attention in the Twittersphere. AZ Dark Money Bot’s tweets include information such as the names of the groups, the candidates receiving dark money, and the amount of money spent on campaigning for or against each candidate.
Project Goals: The major goal for this pioneering project was to enlighten the public, particularly Twitter users, to the gravity and frequency of “Dark Money” campaign expenditures in the 2016 Arizona elections. The project also let audiences directly engage with the reporting process, as the near real-time reporting of the Twitter Bot lets the public react as it is happening, going so far as tagging and directly tweeting at those involved. The Bot’s co-creator, Evan Wyloge, mentions that the bot allowed AZCIR to iron out conceptual challenges, as well.
Tools and Technology: A bot is a software program that runs automated tasks over the Internet. In order to create its own bot to track down expenditures, AZCIR used IFTTT, Google Docs, custom Python code (with a variety of different libraries, including Selenium), and Twitter API. The only technology that cost any money was the hosting of executable code from a cloud. He estimates a monthly cost of around $5 for that service.
Impact: The Dark Money Twitter Bot brought light to the issue of Dark Money and enabled the public to follow Dark Money issues directly on Twitter. Although the project team did not establish measurements for success prior to launch, ultimately, the goal of raising awareness of an important topic was met. Quantitatively, he Dark Money Twitter Bot garnered 457 followers.
Organization Background: Arizona Center for Investigative Reporting is a nonprofit, independent organization that accomplishes unbiased investigative journalism through data-driven methods. AZCIR’s major sources of revenue largely from The Ethics & Excellence in Journalism Foundation (60-75%, as estimated by Wyloge). This organization is now moving over to trying to find local foundations that want to support them and build an audience of people who will provide individual contributions
Project Resources: Aside from the monthly cost of $5 for the hosting platform, the project required two staff members (Wyloge and another staff member, Justin Price) to create and run the project. Wyloge estimates a total of 150 working hours spent spread out over about two or three months.
Here’s what worked
1. Divide and conquer
When pioneering a project, a rough start is almost inevitable, but Wyloge says that it helped his team of two to divide the project into smaller chunks of work. Of course, adjustments had to be made along the way, but clearly identifying project components helped the team push forward in a focused way.
2. Keeping open communication is key
Despite the individualistic nature of writing code, the team behind the Dark Money Twitter Bot made sure that open dialogue was kept, especially when major decisions had to be made. For example, multiple times each week staff would review how the code was working and discuss next steps.
3. Staying close even while remote
Wyloge shares the simple, but very effective solution that is a shared work environment, both physically and electronically. He adds that a big part of this process exists in a Google Spreadsheet.
Here’s what could work have worked better
1. Asking for cooperation from the government
The biggest challenge was turning responses to the bot’s inquiries into usable data. Particularly, turning PDFs into data that could be parsed was difficult. To address this, the AZCIR team wrote custom Python code to go through all of the characters on those PDF reports and looked for specific patterns and use regular expressions or “regex” to identify the important parts. Wyloge says that the solution is to convince government agencies to provide data in a much more useful way.
2. Reporting across platforms
Having a Facebook Chat Bot would have “taken a whole new approach to journalism” where it prompts you with a lot of questions, so you’re actually “chatting,” according to Wyloge. He calls it “conversational journalism” and describes it as very unique.
However, he says that they don’t necessarily have plans for transferring what they’ve done to Dark Money Twitter Bot into Facebook because it’s a whole new platform with different engagement levels. It would be structurally exhaustive.
3. Grabbing fundraising opportunities
Wyloge admits that they could have tried to fundraise around the project, which is always a consideration for nonprofit news organizations like AZCIR. Wyloge says that in 2018, they are trying to find ways to build products that generate revenue. They are working on the creation of a subscription-type service, probably in the form of an app, using a freemium model, that would facilitate campaign finance research. As for what he calls, “Super Users,” or very politically-engaged people such as lobbyists, they can track, get notifications, and even share lists with clients. Wyloge adds that they are already working with a developer and they believe that there is a market out there for this product.
To learn more about this project, you can send Evan Wyloge an email at email@example.com.