The "fake news" phenomenon may have attracted Americans' imagination during the 2016 presidential campaign, and later investigated Russia's attempt to use Facebook's fake news and other plans to push the election to Donald Trump.
In fact, in the years before the 2016 general election, false or false news has become a temporary tool and a tool for disseminating propaganda and conspiracy theories. Websites including InfoWars and Brietbart have been disseminating fake news supporting their agenda.
However, since the election, it has become a political and social issue, and Facebook has become a typical representative of the site that failed.
Recently, the social media company acknowledged the mistake and tried to do it right with the subscriber. It is now tagging news articles posted to Facebook members through their news feeds. It is using AI to achieve this goal.
The company is using artificial intelligence to identify words or phrases that might mean that the article is actually fake. The data for this task is based on articles that are individually marked as false stories by Facebook members.
The technology currently uses four methods to discover false news. They include:
- Score page. The first to use this technology is Google. It uses facts to create scores for your website. Obviously, scoring websites is an ongoing action. However, as Google has been doing, the technology has grown significantly.
- Weigh the facts. This method uses a natural language processing engine to review the theme of the story. AI uses other models to discover whether other sites report the same facts.
- Predict reputation. The technology is based on AI using predictive analytics and machine learning to predict the reputation of a website by considering many features including domain names and Alexa Web ratings.
- Discover the sensational words. Fake news supporters use sensational headlines to attract the interest of potential audiences. The technology uses keyword analysis to find and tag fake news headlines.
Actually detecting these types of articles through AI is a daunting task. Of course, it involves the analysis of big data, but it also involves the accuracy of the data. Identifying it does involve determining the authenticity of the data. This can be done using the method of weighing the facts. What happens if fake news articles appear on hundreds of websites at the same time? In this case, using techniques that weigh the facts may cause the AI to determine that the story is legal. Perhaps using a combination of predictive reputation and weighing facts may help, but there may still be problems. For example, a reliable news source website that doesn't spend time verifying news stories can assume it is true.
Obviously, using AI to identify these articles requires more development. Many organizations are involved in improving the capabilities of AI. One such institution involved is West Virginia University.
Reed Media College has partnered with West Virginia University's Benjamin M. Statler School of Engineering and Mineral Resources to create a course that focuses on using AI to identify news articles.
Senior students taking computer science elective courses are also developing and implementing their own AI projects on the team.
Another group known as the fake news challenge is also looking for ways in which artificial intelligence can successfully fight fake news. It is a grassroots organization of more than 100 volunteers and 71 academia and industry teams to solve the problem of fake news. It is developing tools to help people check and identify fake news stories.
As organizations work to enhance artificial intelligence to discover these stories, there are a variety of tools to combat them. These include:
Spike, used to identify and predict breakthroughs and virus stories, and use big data to predict what will drive participation.
Hoaxy, a tool that helps users identify fake news sites.
Snoopey, this is a website that helps identify mobile news articles.
CrowdTangle, a tool that helps monitor social content.
Meedan, this is a tool to help validate news online breakouts.
Google Trends, monitor search.
La Decodes is from Le Monde, a database of fake news and real news sites.
Pheme, a tool for verifying the authenticity of user-generated and online content.