Researchers have developed artificial intelligence that aims to improve the reliability of Wikipedia citations by training algorithms to identify questionable citations on the site.
The Wikipedia encyclopedia is considered one of the most famous websites, with the number of page views reaching five trillion annually, and it has also become one of the most important sources of knowledge.
Not everything on a Wikipedia page can be trusted, which is why it is important to consult the original sources cited in footnotes, even though these primary sources may sometimes mislead you.
An artificial intelligence program called SIDE checks primary sources for accuracy and suggests new ones. AI assumes that Wikipedia's claims are true, so it is able to verify the authenticity of the source, even if it cannot actually verify the claims made in the entries.
In one study, people preferred an AI-suggested display over the original display 70% of the time.
The researchers found that in about 50 percent of cases SIDE provided a source using Wikipedia as a reference, and that in 21 percent of cases, SIDE came out on top by providing recommendations that the study's interpreters found appropriate.
It turns out that artificial intelligence can help verify Wikipedia's claims, although researchers acknowledge that alternative software could be better than SIDE in terms of quality and speed.
SIDE has limited functionality, which means that the program only takes into account links to websites. Wikipedia cites books, scholarly articles, and information provided by other non-text media such as images and videos.
The researchers note that the use of Wikipedia itself may have been a limitation of the study because, beyond technical limitations, the basic principle of Wikipedia is that any author, anywhere, can assign a reference to a topic.
The researchers explain that people who enter quotes into the site may be biased depending on the nature of the topic in question. Training-based AI is susceptible to programmer bias, and the data used to train and evaluate SIDE models may be limited in this regard.
Regardless, the benefits of using AI to streamline the fact-checking process, or at least as a supporting tool, may find resonance elsewhere.
Wikipedia and social media platforms must contend with malicious actors and bot accounts that flood the Internet with misinformation.