How could AI help detect infox spread on the web?
The introduction of AI in our daily lives would now make it possible to detect the infox that spread on the web. Indeed, it is becoming more and more difficult to distinguish between what is true and false,orapproximate, which could put readers at risk. The phenomenon can be seen mainly in social networks since billions of people have access to it.
Weak social supervision to detect infox
In a paper published on Arxiv.org,researchers affiliated with Microsoft and Arizona State University propose an infox detection approach that uses a technique called “weak social supervision.” Scientists say that through the implementation of infox detection training, this low social supervision would allow to study the interaction of readers with certain pages. It would be thanks to this database that we could verify the veracity of certain information..
Social networks: a danger as a source of news
A Pew Research Center study found that about 68% of U.S. adults would be kept up to date with the latest news through social media. The phenomenon is worrying because misinformation, especially on the Covid-19, is a danger to the community. Therefore, Facebook and Twitter are looking for automated detection solutions. On the other hand, the task remains difficult, because the infox are so numerous and use very different writing styles, even sophisticated.
The “interaction network” to detect unreliable news
Based on a study published in April,the co-authors of this work suggest that poor supervision could improve the accuracy of infox detection. A structure called TiFN,for “Tri-relationship for Fake News”, wasdesigned in this direction. It models social network users and their connections as an “interaction network” to detect infox.
TiFN divides users into several categories to understand their behavior on networks. In practice, media with strong political commitment would be considered dubious, as they could spread news from unreliable sources. The same would be true for accounts admitted as unbelievable. The various studies carried out to verify the effectiveness of TiFN have achieved a 90% accuracy rate.
Given the first test results, TiFN seems to be the solution to this incessant problem of infox. It could be the new bulwark against misinformation that does more good than harm. In any case, it is more prudent to see how the structure evolves over the long term, and complements the other measures already in place on the various platforms.