When analyzing the press, it is important to distinguish between plausibility and truthfulness. While the first concept is associated with reasoning and assertions that meet the potential conditions to be something concrete, but that are not necessarily real facts; the second refers only to defined, supported and verifiable facts.
Although there are methods to detect false news, beyond general culture and common sense, within the margins of plausibility propaganda messages can infiltrate that, hand in hand with real events, can divert the message towards the pursued political purpose. Through an artificial intelligence system, the analysis of these phenomena is proposed.
AI Detection of Covert Political Propaganda in Newspaper Articles
Persuasion, whose main element is the emotional factor, is one of the most used resources in press articles to convince a social group of an idea.
A team of professionals from the department of AI Amazon Alexa AI, the Department of Mathematics of the University of Padova in Italy, the Laboratory of Computer Science and Artificial Intelligence of MIT, Cambridge and the Institute of Research in Computing of HBKU, Qatar; carried out a study that proposes a new methodology for the detection of propaganda, through an interpretable model created specifically for the task.
Through this instrument, the researchers quantitatively deduced the relationship between the semantic or syntactic characteristics and the propaganda techniques, through elements such as the feeling that can be identified behind a sentence, the characteristics of its structure and the similarity of the theme of a news item with its headline.
“Today’s online users are exposed to misleading and propaganda news articles and media posts on a daily basis. To counteract this, a series of approaches have been designed with the aim of achieving a healthier and safer consumption of news and online media ”, points out the presentation of this tool in the report that reports data from the study associated with its development.
With a “colder” interpretive capacity, these automatic systems have the power to help humans detect propaganda manifestations in this content.
As this analysis dangerously borders on subjectivity, this AI system, along with detecting a covert propaganda message, also presents in its reports which techniques were used on a specific basis, as a way of offering better accessibility interpretability.
In particular, this system is in charge of defining descriptive characteristics in a qualitative way, analyzing their suitability with the central theme, to detect potential deception techniques.
This model is presented as a complementary tool, which can be combined with other analysis tools, oriented to large-scale work.