Political Bias at Wikipedia and in AI

The Washington Post has a story today taking note of renewed interest by folks on the right in seeing more balance in articles at Wikipedia. 

Larry Sanger, ousted from the online encyclopedia by co-founder Jimmy Wales a year after he helped launch it in 2001, argues that Wikipedia has strayed from its commitment to neutrality. He says the site is mismanaged and has a penchant for defaming people. And he sees in its coverage of hot-button issues such as crime, religion and climate change a leftward slant that he believes is contrary to its founding principles…

Sanger’s long-standing criticism is finding fresh traction with leading conservatives as they push in the second Trump administration to reclaim cultural institutions and information sources that they argue have been captured by the left. Billionaire Elon Musk, Sen. Ted Cruz (R-Texas) and pundit Tucker Carlson are among those calling to reform or supplant a website that has become one of the country’s last bastions of shared truth — and a bedrock of factual information for the AI systems that increasingly answer the world’s search queries.

Earlier this month, Cruz sent an open letter to Wikipedia’s parent nonprofit, the Wikimedia Foundation, demanding answers about what he called its “ideological bias.” In August, Republicans on the House Oversight Committee launched an investigation into possible manipulation of the platform by what they called “hostile nation-state actors to expose Western audiences to pro-Kremlin and anti-Western messaging.” And Musk has said he is set this week to launch his own, AI-powered alternative to Wikipedia, called Grokipedia.

The efforts amount to a high-stakes bid to shape the world’s most influential reference source toward what conservatives consider a more balanced approach.

Wikipedia is quite biased when it comes to anything with any political valence and that’s especially true of hot topics. For instance, last year I wrote about an investigation by the site Pirate Wires which found a group of Wiki editors who were working in coordination to rewrite articles about Hamas and Israel.

Led by around 40 mostly veteran editors, the campaign has worked to delegitimize Israel, present radical Islamist groups in a favorable light, and position fringe academic views on the Israel-Palestine conflict as mainstream…

One of the most prominent members of the pro-Palestine group is the user Iskandar323, a prolific editor whose nuanced approach to historical and even esoteric articles is representative of the larger effort. In the article on “Jews,” for example, he removed the “Land of Israel” from a key sentence on the origin of Jewish people. He changed the article’s short description (a condensed summary that appears on Wikipedia’s mobile version and on site search results) from “Ethnoreligious group and nation from the Levant” to “Ethnoreligious group and cultural community.” Though subtle, the implication is significant: unlike nations, “cultural communities” don’t require, or warrant, their own states.

Iskandar also worked to sanitize articles on Hamas, in one case removing mention of Hamas’ 1988 charter, which calls for the killing of Jews and the destruction of Israel, from the article “Hamas.” (The edit remains intact today.) He removed mention of Hamas’ 1988 charter in at least three other articles…

All together, the top 20 editors of this group made over 850,000 edits to more than 10,500 articles, the majority of them in the Palestine-Israel topic area, or topically connected historical articles.

Wikipedia’s problem is not just that its editors can be anonymous and are not subject matter experts. The broader problem is that its reliance on “reliable sources” means it only parrots what is reported by left-leaning news sites like Mother Jones, CNN, the Washington Post, the NY Times and refuses to consider anything published by right-leaning sources like the NY Post or the Daily Mail.

Underlying the site’s biases, Sanger argues, is a list maintained by Wikipedia editors that categorizes commonly cited information sources according to their reliability.

Sources deemed “generally reliable” include mainstream media organizations as well as some left-leaning outlets, such as the Nation and Mother Jones, as Sanger has pointed out. (While it’s true that both are listed as generally reliable, the entries for both point out their liberal bias and urge editors to attribute their claims rather than simply stating them as fact.)

On the other end of the spectrum are what Wikipedia calls “deprecated sources,” which editors are discouraged from citing in support of factual claims. Those include not only state-controlled media outlets and notorious propaganda sites, but also some popular right-leaning news outlets, such as the New York Post and the Daily Mail.

As Sanger put it on Tucker Carlson’s show, that means that if Carlson is widely criticized in liberal outlets but defended only in conservative ones, “then you won’t be defended in the [Wikipedia] article about you, and they will call the article about you ‘neutral.’”

That’s exactly how it works. And the results are entirely predictable. Political articles read like left-wing group-think and contrary views get no traction. This problem is magnified by the fact that Wikipedia is now used as training data for a bunch of AI LLMs. So the bias is now being passed up to another layer of tools.

A new study released on Thursday by a conservative think-tank is giving scholarly credibility to long-held conservative suspicions of bias among Wikipedia editors on entries related to current events. 

Wikipedia entries for conservative political figures and organizations do in fact contain more negative attitudes than entries for their liberal counterparts, according to a new Manhattan Institute report released Thursday. This bias could have profound implications for the training of large-language artificial intelligence models, according to the study’s author, David Rozado, a computer scientist who previously researched the apparent left-wing bias of artificial intelligence chatbot ChatGPT and other large-language models.

“In general, we find that Wikipedia articles tend to associate right-of-center public figures with somewhat more negative sentiment than left-of-center public figures; this trend can be seen in mentions of U.S. presidents, Supreme Court justices, congressmembers, state governors, leaders of Western countries, and prominent U.S.-based journalists and media organizations,” Rozado’s report states.

As I argued yesterday, AI doesn’t really know anything except what it has absorbed from sources. If all of its training data on partisan issues is biased left because it relies on Wikipedia, you’re going to get biased output.

As mentioned above, Elon Musk is promising to rollout a rival called Grokipedia. That was supposed to be out this week but hasn’t appeared yet.

Editor’s Note: The Schumer Shutdown is here. Rather than put the American people first, Chuck Schumer and the radical Democrats forced a government shutdown for healthcare for illegals. They own this.

Help us continue to report the truth about the Schumer Shutdown. Use promo code POTUS47 to get 74% off your VIP membership.