“He has been found guilty for…”, you already know the rest of the phrase, hopefully only from having heard it in hundreds of movies in which there is a trial involved. That verdict is always accompanied by a sentence, always documented. But now comes the surprise: What if we told you that on many occasions the source of it is on Wikipedia?

It’s no joke: someone could end up face-first behind bars or be fined financially, with this harsh fate, based in part on this popular online database. How did we know this chilling fact? It has been the prestigious Massachusetts Institute of Technology (MIT, for its acronym in English) who has demonstrated it in a study.

Using a completely randomized field study led by Neil Thompson of the Computer Science and Artificial Intelligence Laboratory (CSAIL), this institution has empirically demonstrated a correlation between articles published on Wikipedia and the direction of sentences.

To do this, this team prepared 150 articles from law students on the well-known page, all of them related to judicial information and sentences, a kind of jurisprudence as bait. Half of them were published on Wikipedia; the other half remained unpublished.

This done, the show began: an algorithm was dedicated to analyzing the published sentences in search of keywords entered in the Wikipedia texts. Come on, like the famous “pixel” of advertisers, but on the hunt for clueless judges.

The aim was to compare the sentences one by one and look for semantic similarities with the articles published by this control group of law students. Of course, no judge suspects that there is no one capable of reading all of Wikipedia and all the sentences, and they are right… In part. A human is not capable; a bot, yes.

The study showed that citations to information published on Wikipedia increased sentences by 20% compared to the same information that was not published online. Or what is the same: the judges took a look at this database to argue part of their reasoning.

You understand the problem, right? Anyone can publish on Wikipedia and this network carries its penance in sin: it is not known if what it has published is true or not, although the “democratization” of information balances its veracity. If someone edits a content with wrong information, it is easy for someone to correct it in no time.

Does the same thing happen with information related to jurisprudence? Taken to an absurdity, one could see the currency of their destiny fall into prison because a law student has published fabricated content. This, obviously, is not the case because the information cited is always accessory and does not form the structure of the sentence.

This study, to everyone’s peace of mind, has highlighted that citations to Wikipedia were higher in low-ranking courts; that is, never in resources to the Supreme Court or the like. And we can get an idea of ​​the cause: low-profile lawsuits are processed faster and it is easier to look for padding information. Still, the study’s authors have called their findings “worrying.” And we agree.

Categorized in: