The term Fake News refers to a series of news items and contents built for specific purposes, very often completely divergent from the intention to provide information. There is a taxonomy of fake news, with different variations, which enables the classification of news as propaganda, libel, conspiracy theory, unfounded story designed to hide the reality of the facts (hoax), sponsored content (clickbait), satire or hate speech and, finally, simple error.
Each of these types of news puts specific communication strategies in place and uses distinct dissemination tools. Social Networks ensure rapid propagation of news, including through the use of IT tools (in particular bots) which publish the latter massively. The greater the speed of distribution, the more complex it is for news agencies and newspapers to verify their veracity, before being called on to give evidence.
In electoral periods in particular, there is a multiplication of circumstances of suspicious news, which can also be created simply by forging images, audios and videos, inserting them in contexts different from the original ones, integrating satire with real news, reporting incorrect or difficult to verify numbers and so on.
Fact-checking as an antidote to Fake News?
Given such a variety of information, its speed of diffusion and the volume of subjects involved, there is an increasing use of practices and tools of fact-checking, which make use of Artificial Intelligence on Big Data technologies and make it possible to act on a large scale, processing unstructured data such as text and multimedia contents. Machine Learning enables training algorithms that learn the typical syntax with which fake articles are written, which is undoubtedly more effective than the use of specialized algorithms that enter into the merit of information content, which is difficult to verify.
Fact-checking is the responsibility of journalists above all, but recently also social networks like Facebook are taking precautions by integrating digital tools for the verification of photos and videos, with the aim of reducing the number of hoaxes and false information published, a phenomenon which has recently also plagued the electoral campaigns of different countries.
Worldwide, new forms of sensitizing public opinion are proliferating, which aim to be captivating and immediately understandable: memes simplify information as much as possible and are exchanged on social networks much more quickly and easily than other content. To understand its meaning, text and image must be interpreted simultaneously, increasing the complexity of fact-checking.
Machine learning and manipulated images
Machine learning can be used to identify three common types of image manipulation: different images combined (splicing), cloning of objects within an image, and removing portions of an image. When such changes are made, digital artifacts are left, such as inconsistencies in random variations in color and brightness created by image sensors (also known as image noise). By combining two different images, for example, or copying/pasting an object from one part of an image to another, this background noise does not match, like a stain on a wall covered with a slightly different color of paint.
Unfortunately, just as AI tools evolve that effect validations on digital counterfeits in post production, there are increasingly sophisticated techniques for generating images and videos that simulate reality: for example, Generative Adversarial Networks (GANs) are a class of Artificial Intelligence algorithms used in non-supervised machine learning, implemented by a system of two neural networks that challenge each other to create photographs that seem authentic to human observers.
The figure below shows a famous modified image of the launch of a missile released by the Iranian government in 2008.
Does Fake News condition people’s choices?
There is an increasing number of research studies coordinated by various American universities that provide evidence of how fake news distorted the outcome of an election due to the consumption of false news during the 2016 US presidential campaign. Specifically, a recent study conducted in 2018 by Princeton University, Dartmouth College and the University of Exeter shows that about one in four Americans visited a false news website, but this consumption was observed in particular among Trump’s supporters, who went to choose news that would confirm their opinions. This phenomenon is called “selective exposure to misinformation”. In particular, among Trump’s supporters, 40% read at least one article from a false news site about Trump, compared with supporters of Hillary Clinton.
Ironically, various polls report that, following the suggestion of conspiracy theories on social media, the majority of American citizens are convinced that fake news is really spread by mainstream TV and newspaper channels. 77% of respondents in a survey by Monmouth University in early 2018 claimed to believe that mainstream traditional TV and newspaper media report “false news”, marking a sharp increase in distrust of news organizations with respect to the previous year, when 63% expressed concern about the spread of misinformation. US President Donald Trump also endorsed this perception through some memorable tweets like the one below:
Not only in the United States do many political exponents rely on Twitter for most of their communications to their electorate, but for example, the Russian Embassy in London boasts a digital club that offers Twitter users “regular contests and prize draws” and even invitations to the ambassador’s residence in exchange for the possibility of using their account for automatic retweets of official posts.
Due to the disruptive effects it is generating in the current situation, the great interest – both technological and social – aroused by the phenomenon of fake news threatens to hold center stage for several years to come before being stopped and it can only be fought by sensitizing the population.