Wikipedia was under siege recently from a right-wing campaign focused on its article about recessions. Isn’t it convenient, these critics said, that the online encyclopedia doesn’t clearly define a recession as two quarters of negative growth? Wikipedia must be taking its orders from the Biden administration, they claimed, which insists that though the U.S. economy has had two negative quarters, it is not in a recession because of other, more positive economic indicators, including low unemployment. The truth was that the Wikipedia article had always reflected different definitions of a recession — some, like the one Britain uses, based solely on two quarters of negative growth; others, like the United States’ preferred definition, based on economists’ assessment of a variety of factors. That didn’t stop angry readers from trying to rewrite the article.
Wikipedia’s administrators — community-elected volunteers who control how an article can be edited — rejected those readers’ demands as unsourced and then “locked down” the article so only established editors could make a change, which just helped feed the conspiracy theories. Fox News personality Sean Hannity sent out a blog post’s headline on Truth Social, Donald Trump’s social network: “Wikipedia Changes Definition of Recession and Then Locks Page.” Elon Musk tweeted at Jimmy Wales, the co-founder and public face of the project: “Wikipedia is losing its objectivity.” Wales pointed Musk to an explanation of what happened, adding, “Reading too much Twitter nonsense is making you stupid.”
This fight over whether the United States is in a recession is striking because of the lofty status it confers on Wikipedia as an objective truth-teller. Here are people who have convinced themselves that the government is lying to them, and they turn to a collaborative encyclopedia to be assured that they are right, like settling a bar bet with the Guinness World Records or checking a proposed word in Scrabble with Merriam-Webster.
Now comes a new paper from MIT and Maynooth University in Ireland offering yet more evidence of Wikipedia’s elevated status, finding that judges routinely rely on its articles not just for background information but for core legal reasoning and specific language they use in their decisions.
Perhaps it goes without saying, but exceedingly rare is the judge who openly credits Wikipedia, so how could the researchers say confidently that judges were relying on it? How to prove not just correlation (it sure seems like the judge was using Wikipedia) but causality (this decision could have been written this way only after the judge read Wikipedia)?
One way might be to introduce a subtle error that serves as a marker — if a party in a case has a strange, made-up middle name that appears only in the Wikipedia article and then appears in a judge’s decision, the evidence would seem pretty clear. (Cartographers trying to prevent their maps from being copied have been known to make up a street name to catch the guilty party.) However, researchers didn’t want to introduce an error on purpose. Maybe you could examine the judges’ computers for the sites they visited, or conduct interviews? But the legal system isn’t exactly a Petri dish, designed for close study without hindrance. Instead, researchers proved the connection through a randomized control experiment, with the judges of Ireland their unwitting test subjects.
Beginning in early 2019 and continuing through early 2020, law professors and their students at Maynooth were tasked with preparing for publication 154 Wikipedia articles on influential Irish Supreme Court decisions; fortunately, only nine such Wikipedia articles existed at the time. Each author would be assigned a pair of articles. An experienced Wikipedia editor guided half the articles (one from each pair) onto the platform, letting curious fellow editors know that the mass introduction of articles came through the university. Researchers took great pains to add formatting to the articles so that Google and other search engines would quickly notice. The rest of the articles were held back from publication. “The only difference between them is one of them gets put on Wikipedia and one of them doesn’t, and then you just wait,” said Neil C. Thompson, a researcher at MIT’s Computer Science and Artificial Intelligence Laboratory and lead author of the study.
The researchers reported that the 77 published articles instantly found an audience, receiving a total of 56,733 views through Jan. 16 of this year. Analyzing court decisions written after the new articles were published, they detected a statistically significant pattern. The Supreme Court decisions with Wikipedia articles saw a 20 percent increase in the number of times they were cited by judges, as compared to the cases whose Wikipedia articles were held back. All of that increase came from lower-level judges, whom the paper’s authors presumed were overburdened and lacking the time and resources to do formal legal research. In addition, the researchers found in decisions a similar spike in the use of certain words and phrases that first appeared in the Wikipedia articles — another way of demonstrating causality.
The paper doesn’t offer details on how cases were decided. There is no example of a decision that was particularly indebted to a certain Wikipedia article. And the authors take pains to say they didn’t find an example of a case being wrongly decided — they stand by the accuracy of the articles they published. They instead wanted to prove the ubiquity of Wikipedia in our lives, what they call “knowledge on tap.”
In 2017, Thompson helped conduct a similar experiment involving chemistry to test whether Wikipedia “doesn’t just reflect the state of the scientific literature, it helps shape it.” Pairs of articles were created then, too — one to be published, one not. After an article on a certain chemistry topic was introduced, researchers noticed that journal articles mirrored the language and conclusions in the Wikipedia account. The academic research that was footnoted in the Wikipedia articles was found to be cited more often in subsequent academic publications, as well.
“I think what we learned from the first one was that scientists are like everybody else in society — we all are reading Wikipedia all the time, right?” Thompson said in an interview. “Some people had this view that you search for everyday things on Wikipedia, but then when you did science and serious stuff, then you only used textbooks and stuff like that. And that wasn’t what I saw going on around me. And so we said, okay, I think we should look at this effect.”
That judges were dependent on Wikipedia seems more serious, Thompson said, if for no other reason than that judges may hold a person’s fate in their hands. The paper’s disdain for that shortcut-taking is palpable, contrasting such behavior with Alexander Hamilton’s description of judges as members of an elite who commit to “long and laborious study to acquire a competent knowledge” of the law. What about the potential, the paper’s authors ask, for one side in a legal argument to edit the article about a relevant case to back its argument and persuade the judge? Should justice be allowed to hang on such a thin reed?
In practical terms, the more important finding in the research isn’t that judges may be vulnerable to self-interested changes to Wikipedia articles but that we now know definitively that judges rely on such articles. By proving this about judges, and scientists as well, the papers’ authors are helping demystify those priestly classes. They live in our world and use the same resources we do. “I think that we in academia are front of that line in terms of, you know, feeling two different ways about Wikipedia in sort of what we’re saying and what we’re doing,” Thompson said. “I absolutely think that it is very important that we just are upfront about the fact we do use Wikipedia.”
Armed with this fact, the solution shouldn’t be to shame those who use Wikipedia but rather to make Wikipedia as reliable and inclusive of all parts of society as we all need it to be. As it happens, the outcry over the article on recessions did lead to a revision. While the description of how the United States formally defines a recession didn’t change, the article’s introduction now gives slightly more emphasis to the two-quarter definition, noting that it is “commonly used as a practical definition of a recession.”
Among the researchers’ recommendations for Wikipedia is that experts be enlisted to create and watch over articles on their subject areas, which may seem to challenge the revolutionary premise of “the encyclopedia anyone can edit” (as I do, for example). But as long as the term “expert” isn’t tied to the number of academic degrees possessed but rather to demonstrated interest, experience and knowledge, the idea would seem to fit the way Wikipedia operates. People who care about a subject — whether warfare, medicine or LGBTQ+ issues — already watch over those articles. There are ongoing campaigns to create articles about women and minorities, who are significantly underrepresented in its pages. Intriguingly, the earlier chemistry paper has helped Wikipedia make the case for experts to get involved, Thompson said: “We showed in the first paper that if you add some content and there’s a citation to your paper, you’ll have more citations. So that was some incentive for them.”
Though the new Irish Supreme Court articles were shepherded onto the site by an experienced editor, avoiding suspicions about whether someone was trying to make inappropriate, wholesale changes to the project, at least one article was objected to and proposed for deletion. It concerned a case that was categorized as procedural — related to whether a court could review the constitutionality of a law if the legal action had been settled. The facts in the case involved allegations of domestic violence, however, and having a Wikipedia article suddenly brought to prominence events that had long been obscure. A new editor posted to Wikipedia demanding that the article be removed because it contains “in-camera and outdated information about an Irish citizen’s private life,” adding that “these publications have caused the individual particular distress and they have especially been used to blacken the person’s name in the workplace.” The Wikipedia community was not persuaded: “The decision of the Irish Supreme Court in 2004 is a matter of public record, has seemingly been so for 15+ years and it’s hard to see how the genie can now be stuffed back into the bottle. If there are genuine privacy concerns, this is not the way to go about it.”
And we are left to ponder the philosophical question of our age (forgive the hyperbole): If a tree falls in a forest and Wikipedia doesn’t write about it, did it happen?