The ease of spreading fake news on digital channels such as Facebook and WhatsApp, and the difficulty of removing such content from the network facilitates the action of individuals with malicious intent to defame their opponents, manipulate the market for personal gain, etc.

The volume of fake news is big and the news comes up all the time. It is virtually impossible to manually scan to remove inappropriate content from the Internet. Thus, Artificial Intelligence (AI), with its ability to understand human language and  analyze various data per second, seems to be the ideal solution.

How has the spread of fake news started?

For the past year, we have been living in a very polarized political landscape, based on mutual aggression on social networks. Unfortunately, most of the time, these discussions are based entirely on fake news. Instead of respecting differing opinions, many people have preferred to launch fake campaigns against the honor of individuals.

For example, in recent weeks, the Federal Supreme Court (STF) has organized a task force to identify slander against the institution and the ministers. Judges favoring a lighter criminal policy are being attacked and accused of corruption without any factual basis. Everything is done using fake news.

In this context, the Federal Police collected several computers suspected of containing fake news message triggering software. That is effectively machine learning has been used for their propagation. Now it seems that he too will be the antidote himself to this serious problem.

Scoring web pages

Scoring web pages was Google’s solution to reducing the spread of fake news.

The company’s machine learning software can understand the content of the pages. Then, according to the language used, such as the use of sensational words, robots identify potentially suspicious content.

As a result, search page ranking drops dramatically, reducing the chance of fake news reaching a larger audience.

Weigh the facts

For fighting false news, one of the most important techniques is the balancing of facts. To this end, Artificial Intelligence has increasingly advanced in determining the semantic meaning of texts and photos on a web page.

Thus, based on the meaning of the text, the choice of images and the geographical location, machine learning will make a comparison of pages with what has been narrated on reliable sites, such as those of the mainstream media. If the content is too divergent, the page will be considered fake news.

Predict reputation

As a web address is constantly linked to fake news, AI will notice this pattern and classify this site as “unreliable”. From then on, all publications in the address will be considered false news sources.

Contrary to what happens in the first item, which analyzes news by news, the whole site is put under suspicion. This can do preventive work and prevent social networks and Google ranking from giving it more visibility.

Discover sensational words

In news, the headline is used to get the public’s attention. This explains why fake news sites often use extremely hyped headlines.

In this way, they capture a growing audience that, organically and without the need for firing tools, helps spread the news faster and more widely.

Machine learning is an essential tool to identify and combat this strategy. Over time, the robot will realize what the hyped keywords are and how they are used to decrease fake news visibility.

Predict User Behavior

Generally speaking, fake news spreaders follow a very similar pattern, especially if they are using message triggering software. Rather than understanding content, Artificial Intelligence will also be able to identify the pattern of fake news propagators.

For example, generally, in fake news campaigns, the number of referrals in WhatsApp is much higher than normal for a user. Similarly, there is excessive use of caps lock. All of these factors can help arouse suspicion of fake news.

Identify the language used

In addition to the sensational words, there is another very interesting strategy for identifying fake news: understanding the linguistic pattern of news.

In this sense, a study by USP may be essential to help Artificial Intelligence. Researchers realized that there are certain classes of words or grammatical errors that are much more frequent on fake news sites than on traditional news sites.

The surveys point out that:

  • The length of the sentences is usually much larger in fake news. In traditional virtual newspapers, phrases average 15 words. In fake news, this average rises to 21;
  • There are many more misspellings on fake news sites. The number is scary. While misspellings are present in only 3% of traditional newspaper page content, in fake news spreaders this is 36%!
  • Because they are run by ordinary non-journalist people, fake news sites also have other issues such as lack of lexical wealth, inadequate punctuation, and so on. All of this can be easily identified by AI.

Therefore, if until a while ago artificial intelligence was the main ally of fake news spreaders in Brazil, it seems that now it will be the great enemy. So many people have had their reputation damaged by false news that more forceful action needs to be taken. Otherwise, instead of a source of information, the Internet will become a place for misleading content. Therefore, actions are required to maintain network integrity.