AI Program Demonstrates Capability to Verify Wikipedia Citations, Enhancing Reliability and Trustworthiness

You can’t trust everything on Wikipedia, which is why it’s essential to consult the original sources cited in the footnotes. However, even primary sources can sometimes mislead you. To address this, researchers have developed an AI tool called SIDE, aimed at enhancing the reliability of Wikipedia references by training algorithms to identify questionable citations.

SIDE performs two key functions: it verifies the accuracy of primary sources and suggests new references. Notably, the AI operates under the assumption that Wikipedia claims are true, meaning it can validate sources but cannot verify the claims themselves.

A study revealed that users preferred the AI’s suggested citations over the originals 70 percent of the time. About half of the time, SIDE identified sources already cited by Wikipedia as the top reference. Additionally, 21 percent of its recommendations were already accepted by human annotators.

While SIDE shows promise in assisting editors with Wikipedia claims, researchers acknowledge that other programs could potentially surpass its quality and speed. Moreover, SIDE is limited to references corresponding to web pages, while Wikipedia also includes books, scientific articles, and various media formats.

The very nature of Wikipedia allows any contributor to add references, which can introduce bias depending on the topics. Additionally, AI programs, particularly those reliant on training data, may reflect the biases of their developers, which could limit SIDE's effectiveness.

Despite these limitations, using AI to streamline fact-checking has significant potential applications in combating misinformation. Misinformation has become increasingly rampant, particularly in contexts such as the Israel-Hamas conflict and the upcoming US presidential elections. Leveraging AI tools like SIDE could be crucial in addressing this urgent issue, although further advancements are still required for optimal effectiveness.