Transparency bots catch congressional information tampering

Monitoring Wikipedia edits made from Rus- sian government addresses, an automated tool caught controversial changes in the wake of the Malaysian Airlines Flight 17 crash in Ukraine this July. Someone at a state-run TV and radio network, VGTRK, anonymously removed mentions of Russian Federation-sourced missiles, swapping in Ukrainian soldiers as the culprits.

The program, or bot, that discovered the edit then automatically posted to Twitter on its account @RUGovEdits. A similar bot, watching for changes from Boeing IP addresses, discovered edits to the article on Israel’s Iron Dome air defense system. The Boeing additions cast doubt on an analysis that found the system’s intercept rate to be low.

Both monitoring tools were created from the blueprints for @congressedits, which watches for changes made on Capitol Hill.

These simple automated alerts could be the beginning of a brighter future for journal- ist-algorithm relations. Rather than spitting out financial reports or sports recaps, technology put to this end reaches for the highest ideal of the profession – providing more transparency to the actions of government.

Perhaps more importantly, the transpar- ency bots have been built from open source code. This has allowed others to quickly adapt the tool to their aims. Although most have simply shifted from one nation’s government to another’s, other possibilities abound.

@congressedits

From a computer in a U.S. House of Repre- sentatives office, someone anonymously updated Donald Rumsfeld’s Wikipedia bio on July 15, adding that he is an “alien lizard.” In more serious historical revisionism, one anonymous staffer altered the article for JFK’s assassination that same day, changing it to read that Lee Harvey Oswald had acted on behalf of Fidel Castro. On July 22, someone added to the article for “moon landing conspiracy theories,” writing that they too were promoted by the Cuban government.

We know this because @congressedits, which sends alerts via Twitter when anonymous Wikipedia edits are made from the U.S. Con- gress, tells us.

@congressedits’ creator, Ed Summers, has freely posted the code for the bot, which matches the listings of congressional IP (Internet protocol) addresses with the metadata gen- erated when anonymous edits are made. This has helped a host of sister-bots crop up to keep tabs on governments around the world and international bodies such as the UN and NATO. Enterprising watchdogs have set up accounts to monitor Monsanto, oil companies, defense contractors, pharmaceuticals and Goldman Sachs, as well as state and local government.

Many of the changes caught by @con- gressedits are vanity projects. Michigan’s Rep. Justin Amash, for instance, went from being a “corporate lawyer” to a more neutral “attorney.”

(Most revisions, like this one, don’t last long – other editors revert pages back to original form.) Such self-serving modifications recall the first major scandal surrounding Congressional Wiki-tampering. In 2006, the Lowell Sun broke news of a Hill staffer scrubbing unflattering in- formation from the page of his boss, Rep. Marty Meehan of Massachusetts. Meehan had pledged to surrender his seat after serving eight years, which he had not; he also had the largest campaign account of any House member. In their place, favorable biographical details were added. An investigation by Wikipedia administra- tors later revealed a number of inappropriate edits to legislators’ pages made from Congressional IP addresses. Among these, the page for then-Sen.

Joe Biden had jettisoned significant criticism.

Skeletons in the closet?

The movement has also inspired retrospec- tive digging, with comprehensive lists of edits being collected for many of the organizations now being monitored in real time.

For instance, thanks to Jari Bakken, lead developer of a Norwegian parliamentary watchdog account, a database of 1,843 edits made at Pentagon IP addresses from 2004-2010 is now publically available.

Exploring this reveals Pentagon employees contributed uncivil language to the pages for John Kerry, Valerie Plame and Marion Berry. One editor put words in Keith Olbermann’s mouth, inverting his quote from “I’m not a lib- eral, I’m an American,” to “I’m not an Ameri- can, I’m a liberal.”

Regarding the “First Battle of Fallujah” article, an anonymous editor added “WHAT LIBERAL, PACIFIST, JAG-OFF MEMBER OF THE OBAMA CABINET WROTE THIS POLITICALLY SLANTED PUFF-PIECE?”

Homophobic slurs were inserted into pages for “gay pride,” “LGBT symbols,” and bizarrely, Evian water.

A global warming skeptic quibbled over the language of the article dedicated to the scientific phenomenon from his office at the Department of Defense headquarters. In general, Pentagon employees’ more notable edits addressed what they perceived as Wikipedia’s liberal bias.

The history of Congressional edits (13,269 total from 2003-2014), also now available, pro- vides evidence of more ideological alterations. In 2005, for example, someone on Capitol Hill removed a reference to the War on Terror from the page for “crusades.” In 2007, the entire page for “abortion law” was blanked.

Some sling mud: One anonymous editor called Noam Chomsky a “dangerous radical” and “undoubtedly a shifty Communist gulag- master.” Another – or possibly the same – la- beled Eugene V. Debs a “whore.”

Other changes seem more geared toward electability: Joe Lieberman’s lobby connections disappeared; Nancy Pelosi was noted as “ex- treme” and hypocritical for accepting corporate cash. Some are just silly, such as changing John Boehner’s age to 88 years old.

While it appears that most edits emanat- ing from the Beltway have been made in good faith, with a few vandals making alterations without strategic aim, there have been more deliberate changes originating from Congress than the DOD. Still, these are more likely to indicate pettiness than evil.

At the very least though, the records in- dicate a severe misuse of public time, equip- ment and ultimately tax dollars, according to Stephen Potts, chairman of the Ethics Resource Center during the Meehan embarrassment.

It’s fitting, then, that one Pentagon employee apparently spent many hours contributing to the Wiki for Joseph Heller’s classic novel, “Catch-22,” which skewers the absurdity and waste of the military and bureaucracy at large.

‘Artisanal news bots’

Of course, now we can see changes in real time. There is an instant trans- parency, even if just for one minor gov- ernment activity, provided by the tools Summers and others have developed.

Newspapers have been cutting back on investigative reporting, but these technologies try to reverse the asymme- try of surveillance between the people and the government.

Nick Diakopoulos, a Tow Fellow at Columbia University, sees even more po- tential in the movement @congressedits triggered. If we “build an open-source news bot architecture,” he says, “differ- ent news and journalistic organizations could use [it] as a scaffolding to encode their own editorial intents, newsworthiness criteria, parameters, data sets, ranking algorithms, cultures, and souls into.”

It would help expand the use of journalistic automation into tasks beyond aggregation, such as actually soliciting information. On the verge of this, an IBM research team has developed a prototype that “automatically identifies and asks targeted strangers on Twitter for desired information.” Diakopoulos says these solicitors could seek facts and photos from those likely to be nearby a breaking news event.

The diversity of news algorithms that could result from free experimen- tation would help prevent another po- tential downfall in algorithmic report- ing, according to Diakopoulos. “What I think we don’t want to end up with is the Facebook or Google of robot reporting” he says.

Rather than homogenized news generated by a few big providers, the path blazed by today’s open source trans- parency bots might lead to, as Diako- poulos imagines, “thousands of artisanal news bots serving communities and variegated audiences, each crafted to fit a particular context and perhaps with a unique editorial intent.”

Editor’s Note: This story was previously published in the Summer 2014 Print Edition of the Gateway Journalism Review.