Research report on judging reliability and political bias of net news using AI

The worldwide spread of the Internet brought about the overwhelming amount and speed communication. At the same time, the credibility of information became strongly questioned, and especially in social media, fake news aimed at biased political activities and self-interests became more common. Meanwhile, Massachusetts Institute of Technology (MIT) and the University of Qatar have revealed that they are developing " AI to see if it is fake news by scrutinizing information sources ".

Predicting Factuality of Reporting and Bias of News Media Sources
(PDF file) https://arxiv.org/pdf/1810.01765.pdf

Detecting fake news at its source | MIT News
http://news.mit.edu/2018/mit-csail-machine-learning-system-detects-fake-news-from-source-1004

In social media, fake news, discrimination / defamation, and biased political advertisement are widely deployed. Facebook has already launched more than 10,000 operators to work on strengthening the safety of SNS, and it is reported that it will increase to 20,000 by the end of 2018. Also, YouTube · Twitter is handling measures with a large number of people as well.

Researchers at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) and at the Qatar Computing Laboratory (QCRI) at Qatar University not only individual arguments but also "focusing on the news source itself" are the best for fake news detection , We are developing a system that uses machine learning to judge "source accuracy" and "political neutrality".

by freie-presse.net

The research team registered data acquired from " Media Bias / Fact Check (MBFC)" which manually checks bias of 2000 or more media sources in the machine learning data set, and AI registers news like MBFC We programed to check the reliability and political bias of the site.

AI can evaluate the reliability of news in three stages of high, medium and low, and at the same time it can classify political bias as three types of left wing, right wing and neutral. Whenever new news were actually reported, I checked it with AI, and it seems that I could judge the reliability of news 65% and the political bias 70% with accuracy.

Through scrutiny of the news by AI, it turned out that fake news tended to use more subjective and emotional words. Also, in terms of political bias, it is reported that the tendency of words such as "harm / anxiety" and "fair / mutual benefit" tend to be more frequent, especially from those deviated to the left wing.

According to ACRI researcher Preslaf Nakov, this AI system not only detects words such as "extreme" "conspiracy theory" but also whether the URL text to Wikipedia etc. exists in the page source I was checking it. It seems that new sources of URLs using complex subdirectories and special characters were considered to be unreliable.

At the same time as development of AI, the research team created an open source data set that evaluated the reliability and bias of more than 1000 news sources. In addition, the research team says that "whether AI learned in English can be applied to other languages" and "whether to be able to explore region-specific prejudices" will also be studied.

by Cody Williams

According to Mr. Nakov, AI is still in the process of development and it is most effective to use it in combination with conventional fake news check to the last rather than using AI alone even if the accuracy improves. "There are quite interesting things to think about ways to tell the news to different people, these tools will help people think about the problem and help us to explore perspectives that have never been before," Mr. Nakov is commenting.

Dr. Lamy Barry , chief administrator of the paper at CSAIL said, "If you have about 150 articles, you can detect whether the news source is trustworthy, not just the website that publishes the articles, By examining the news source, we can prevent new fake news from being born before articles are spread. "