Company using foreign workers botches u.s. senate campaign finance records center for public integrity

But similar provisions have consistently faltered, since lawmakers began floating them in the early 2000s. Reform advocates say this is in large part because of opposition by a small group of Senate Republicans, most notably Senate Majority Leader Mitch McConnell, R-Ky. McConnell couldn’t be reached for comment, and, in the past, has failed to respond to repeated reporters’ requests for comment on this issue.

But the process of converting paper records to digital makes pre-existing errors worse and introduces new ones. A Center for Public Integrity analysis of comparable data from paper filings and data from electronic filings found 20 percent of digitized paper filings had at least one significant error, compared to 2 percent of regular electronic filings.

Captricity turns these images into machine-readable text through what its website calls a “groundbreaking collaboration between humans and computers.” The firm recruits workers on the Internet to verify those automatic transcriptions, or to type out the contents of many of the image files. According to Captricity’s website, every piece of data is verified by at least one person.

FEC data posted before the spring of 2016 was digitized by traditional data-entry clerks, who would type up the records page by page. Captricity touted the FEC subcontract in a blog post in 2016, after the commission transitioned into the current system. The company wrote: “Captricity’s technology has allowed the FEC to cut turnaround time by over 90 percent, eliminate the manual data processing of thousands of pages monthly, and increase the number of digitized pages per month.” The FEC says the project has cost $3,376,743 over the last 4 years.

But the results of Captricity’s work are sometimes inaccurate. The Center for Public Integrity sent Captricity’s CEO, Nowell Outlaw, some of the errors and asked him to explain. Outlaw declined to be interviewed, saying that he would need the FEC’s approval to speak. AuroTech, the company that subcontracts Captricity, did not return requests for comment. Error name

More than 2,300 contributions in FEC records are not easily traceable to their source because of missing name data. Moreover, nearly 2,400 spending records are missing the name of the recipient. The Center for Public Integrity’s analysis indicates the overwhelming majority of missing names seem to be missing because of an error in digitization.

Even when paper conversion works right, the FEC itself sometimes errs. In the course of investigating, the Center also found 591 paper filings that incorrectly displayed $0 totals or were missing from the agency’s website entirely. The Center for Public Integrity notified the FEC of the problem in March, and the agency began fixing it in April.

In a statement, the FEC told the Center for Public Integrity it was aware of errors introduced by digitization and said its contractors have worked to reduce such errors. Several of the erroneous records counted above had already been partially corrected by the FEC. In many cases the correction process has resulted in two, side-by-side records: one wrong and the other correct.

Captricity administers this kind of work through Mechanical Turk, an Amazon-owned online labor marketplace. On Mechanical Turk, workers earn money by completing hard-to-automate tasks, such as typing out information contained on scanned images of U.S. Senate campaign finance reports. Most tasks only take a few minutes, and they pay a few pennies.

Captricity appears to be connected to a Mechanical Turk account named “p9r.” Shortly before an April 2018 FEC filing deadline, p9r posted thousands of tasks asking workers to transcribe data from federal campaign finance reports, including FEC committee names. Workers completing other p9r tasks have connected this account to Captricity’s other clients. And example code posted publicly by Captricity uses the username “p9r” to sign in to a database.

These conditions could be contributing to U.S. Senate campaign finance data problems. Workers have no incentive to correct or report flawed images. If they type in anything other than what’s in the box, they risk their work being “rejected”: p9r could flag their work as against instructions, and refuse to pay them. Mechanical Turk workers can dispute rejections, but to do so with p9r is to argue over pennies.

“The fact that the Federal Election Commission is now having folks all over the globe manually enter campaign finance information into a database is irresponsible,” Jon Tester, the Democratic Senator from Montana, wrote in 2015. “It’s long past time the Senate enter the 21st Century and file these reports electronically. We’d save time and protect taxpayers’ money and privacy, while improving election transparency. This isn’t rocket science, we need to pass my e-file bill."