A study by the Manhattan Institute, a conservative think tank, has uncovered evidence suggesting Wikipedia articles are biased against conservatives. This finding has significant implications, as AI chatbots increasingly rely on Wikipedia for information dissemination.
Researcher David Rozado examined the portrayal of 1,628 political figures on the English-language Wikipedia. His analysis of 3.8 million paragraphs revealed a pattern: right-leaning figures, including Donald Trump, were often presented negatively, while left-leaning figures like Barack Obama received more positive coverage.
Wikipedia claims to adhere to a “neutral point of view” (NPOV) policy, aiming for fair and unbiased presentations. However, Rozado's study shows this goal is not being met. The study found bias affecting not just American politicians but also Western media and leaders, though British politicians and US think tanks were portrayed neutrally.
The potential impact is significant, given Wikipedia's role in training large language models like OpenAI's ChatGPT. Rozado noted biases in Wikipedia content appearing in these AI models, suggesting a broader issue of misinformation bc of the potential to spread these biases across the web.
This isn't the first criticism of Wikipedia's neutrality. Larry Sanger, Wikipedia's co-founder, claimed in 2021 that the site had become a tool of left-leaning propaganda, marginalizing conservative viewpoints and sanitizing content about liberal figures. He stated that conservative voices were “kicked out” and that negative content about Joe Biden and other liberal figures was removed from the platform. Supporting these claims, a 2012 study found that the site used language preferred by Democrats on topics like taxes and the Iraq war, while sidelining terminology favored by Republicans.
Wikipedia's influence is vast, with the English version alone receiving over 84 billion views last year, primarily from the US, Britain, India, Canada, and Australia.
