Wikipedia Might Be Due A Spring-Cleaning From New AI System For Credibility

For many, Wikipedia is a go-to source for all things knowledge. It’s where people dive into history, science, pop culture, and more. However, as it’s user-driven, reliability can sometimes be a question mark.

In a bid to tackle this, there’s a new player on the scene: an AI system called SIDE. According to a study from the Nature Journal, it’s here to make sure those references back up their claims and offer better options when they fall short.

SIDE, powered by neural networks, was specifically engineered to analyze whether the references within Wikipedia articles substantiate the claims they are linked to. If not, it steps in to propose alternative, more suitable references.

To fine-tune its abilities, SIDE was trained on a selection of featured Wikipedia articles, equipping it with the skill to discern high-quality references from subpar ones. When a reference falls short, SIDE takes the initiative to scour the internet for better, more reliable sources, thus maintaining the credibility of the content.

During testing, it was observed that in nearly 50% of cases, SIDE’s top reference choice was already cited within the article. This redundancy may raise questions about the system’s efficacy. When presented to a group of Wikipedia users, 21% preferred the AI-suggested citations, while 10% favored the existing ones. Notably, 39% of users had no particular preference, indicating a degree of neutrality.

It’s important to acknowledge that any software program, especially an AI system that relies on training, has the potential to reflect the biases of its creators. The information used to train and assess the models used by SIDE may have some limitations in this regard.

However, using AI as a tool to streamline fact-checking or support could have significant applications. Wikipedia and social media companies face challenges from automated bots and malicious actors who spread false information on digital platforms, and tools like SIDE could help mitigate that.

But whether or not it will do a better job than manual fact-checking depends on future advancements and reviews of such tools.

[via Nature, cover photo 231143550 © MrFly | Dreamstime.com]