Think about how technology can help fight media propaganda?

Photographer: Arkady Shaikhet.

Introduction

This article can be divided into two logical parts. In the first one, I consider the phenomenon of media propaganda: what it is, why is it dangerous, and to what terrible things it has already led in history. I will consider in detail the mechanisms of work of modern state propaganda in the media and try to understand why it is so effective and what characteristic features it has.



The second part is a reflection on how, based on the characteristic features discussed in the first part, one can try to automate the recognition of propaganda in the media. Since propaganda is the impact of text (besides the picture, of course), and I do not have the necessary knowledge in the field of natural language processing (Natural Language Processing), then my calculations are exactly what thinking aloud. I will describe in a more or less structured way the possible functions of a program that should look for signs of propaganda in media materials, but without technical specifics.



Therefore, I will say right away: do not expect to read in the text about specific algorithms and their methods of application. On the contrary, I'll be honest: I hope to find among you those who understand both NLP and browser plugin development and get feedback from you in the comments. If enough people get infected with my ideas, then maybe we can work together on an interesting open source solution!



Disclaimer



Since there should be no politics on Habré, I will not go into details in the text of how the idea of ​​creating this project came about. If someone wants to know about this - especially, probably those who want to join the project - I will tell you in the comments or in a personal. Let's just imagine a certain State that political scientists would describe as "authoritarian", that is, one in which the personal freedoms of citizens are suppressed, election fraud is made, the opposition is persecuted, and so on, and imagine that everything that will be discussed further, refers to this abstract State (although, given the cultural context, of course, examples will be primarily from our post-Soviet realities). Why I focus on state propaganda, I will explain later.

What is propaganda and why is it dangerous?

Photographer: Elliott Erwitt

One of the eternal pillars of authoritarian states is powerful state propaganda. We hear about this phenomenon quite often - but what is it like?



The American propaganda author Edward Bernays in 1928 defined propaganda as "a consistent, unremitting effort to create or shape events with the aim of influencing public attitudes towards an initiative, idea or group." In other words, this is an attempt to manipulate public opinion. It is interesting that in international law there is still no single definition of the concept of "propaganda" due to the fact that it is very difficult to draw a clear line between manifestations of freedom of speech and systematic propaganda, given that many authoritarian regimes with varying degrees of success can for a long time pretend to be "defective" democracies. However, organizations such as the OSCE distinguish between 2 types of propaganda:

1. , , , , , . – , , . « ».
2. , , .

This is quite an important point, let's fix it: there is propaganda for war, using hate speech, and other political propaganda . We will return to this division, as well as to a more precise definition of hate speech later. In the meantime, let's try to answer the question why propaganda is a danger to society at all?



History knows several tragic examples of how propaganda managed to awaken the feeling of hatred in people, which led to wars and genocide. The example of the Nazi Third Reich is so famous that I see no reason to dwell on it separately. Another example is the genocide in Rwanda, during which the Hutu people killed and raped hundreds of thousands of Tutsi people for three months; a year before these events, the radio station "Free Radio and Television of a Thousand Hills" began to urge listeners to exterminate cockroaches (the contemptuous name of the Tutsi) and indulged in racist humor, and with the beginning of the genocide began to voice the addresses of specific people whom, in their opinion, it was necessary to kill ... The private radio station Thousand Hills received funding from another, state-owned radio station,and its employees were exclusively Hutu. Over the years, Harvard economist David Yanagizawa-Drott conducteda study in the course of which he was able to prove the influence of radio propaganda on the scale of the tragedy by comparing the signal level at each point of the country with the number of people convicted of genocide at these points.



Another example of the influence of propaganda on the course of history is the Yugoslav wars. Numerous studies of the role of the media in the conflict in the former Yugoslavia have shown that media in the service of the regime have contributed to the incitement of war and hatred. Mark Thompson in his book " Forging War » (Forging the War) wrote that " verbal violence led to the physical ." Italian journalist Paolo Rumiz also wrote in his book " for the massacre Masks » (Masks for a Massacre), that "in 1988 the war was in headlines and articles."



These are all extreme examples, when propaganda almost directly called for the massacre and destruction of another people or social group. But it is obvious that propaganda is not always so straightforward, and it does not always pursue precisely such goals. Why, then, is it dangerous and harmful?



Journalism has a goal in front of itself - to understand and inform citizens about the results of their investigations into the actions of the state or others. Propaganda wants not only to inform, but to inspire some ideas. It manipulates citizens so that they do not behave consciously, guided by their own interests, but act in the interests of the manipulator (for example, the state), that is, they vote for the “correct” candidate or decision, or ignore obvious problems and do not ask unnecessary questions to the authorities. Under the conditions of freedom of speech, all this does not seem like some terrible things, but in an authoritarian state, besides propaganda, the authorities have other levers, such as pressure on independent media, the affiliation of media moguls with the government and the possibility of massive election fraud.as a result, the propaganda effect increases many times over. In addition, propaganda often achieves its goals by looking for illusory enemies, external or internal, and appealing not to reason and logic, but to feelings (see the term “post-truth ”), and because of these two factors, state propaganda sometimes actually teeters on the brink of“ hate speech ”.



In addition, as we will see later, modern state propaganda is damn effective, possessing a large arsenal of means and channels. A special danger is that, being in an endless stream of information, even high-quality media can sometimes pick up false information deliberately released into the media space by manipulators. Some examples of recent years: misinformation about the EU plans to abolish visas for Ukrainian men was so often mentioned in the Ukrainian media that the country's authorities at some point were forced to make an official refutation, and shortly before that, the German media picked up false information about the atrocities on east of Ukraine.

? ?

Christopher Paul and Miriam Matthews, experts from the NGO RAND Corporation, answered these questions very well. With Russian state propaganda in mind, they claim in their report that "the new Russian propaganda is entertaining, confusing and shocking the audience." And they explained its effectiveness with the following qualities: " It is operational, continuous, illogical and uses the principle of multiple repetition ." They analyzed each of these characteristics:



1. Operational

2. Continuous

3. Not logical

4. Uses the principle of repetition.



Let's go through each of these points and consider the mechanisms described by the authors of the report for the influence of propaganda on recipients of information.



Operative



Advocates are very quick at “making” material because they don't need time to check facts, gather opinions from different political camps, and so on. (although sometimes it takes time to determine the attitude towards certain events). The fact is that a person is more likely to believe the first message received on the topic, and in the future, when competing messages appear, they will tend to favor the point of view of the message received first. Moreover, the influence of the first message on the way of thinking may persist even if a refutation is received in the future. We are like an algorithm that once having learned something, can only unlearn it with great difficulty.



Continuous



Propaganda churns messages like a true infofabctory, leaving no gaps in the message flow. The recipients get a continuous picture of everything that happens around. This continuity gains additional effectiveness thanks to the multichannel nature of modern propaganda. After all, everyone has heard stories from a couple of propaganda TV channels and sites, but in fact, sites and telegram channels promoting the "party line", according to the authors of the report, dozens. And such multichannelism helps propaganda a lot in achieving its goals due to the fact that people tend to believe that information received from various sources is most likely based on different points of view, and therefore deserves attention. Since these channels together have a huge coverage, recipients find both in their environment,and on the Internet there are many people who approve of this information, and thus the reliability of this information in the eyes of the recipient increases. In addition, there is another interesting effect: when the consumer's interest in information is low (in other words, when the consumer does not have the time and desire to understand the information), the credibility of the message depends more on the number of arguments in its defense than on their quality.



Propaganda exploits another bug of our brain: it receives its dose of dopamine every time we click on the play button of a video on Youtube, and it is very easy to "add" to the consumption of video content and watching social media feeds. A person finds himself in a state of flow, from which it is very difficult to get out. It's just that someone gets used to watching interviews, series or short video vines, and someone - posts on political topics and talk shows.



Illogical



Propaganda knows how to change shoes on the go, but this does not in any way affect those who already associate themselves with propaganda channels of information. Here it is also the fact that people, as a rule, are too lazy to understand the information, and the fact that people simply forget that the TV presenter just a few weeks ago said something exactly the opposite, and that if the recipient once begins to trust a certain source, then he ceases to be critical of the information he disseminates.



Uses the principle of multiple repetition



As you already understood, propaganda exploits several interesting bugs in our thinking. One of them is the "illusory truth effect." People are more likely to perceive a statement to be true if they have heard it before than if it was a completely new message. The illusory truth effect stems from the fact that repetitive statements are more readily accepted and accepted as true. An interesting detail from the authors of the report: " Even in the case of ridiculous stories and urban legends, those who have heard them many times tend to believe it ."

How can you catch a first impression?

Photographer: Rene Maltete



So, we looked at the theory of propaganda, we understood why it is dangerous and why it is effective. Just imagine thousands, if not millions of your fellow citizens, to whom day after day they explain who to love and choose, and whom to hate and persecute - it is obvious that such a situation is unhealthy for the state, because citizens must make their own choices based on their own convictions. formed on truthful facts. I got into such a situation. And, as a programmer, I wanted to do something with my own hands.



I was particularly hooked on the fact that propaganda tries to create a first impression, and it achieves this through its multichannel nature - at least online. I immediately imagined a person who, after a hard day's work, goes online, opens news in his browser and opens a portal to the stream of consciousness of dexterous manipulators. And then I realized that the easiest way to intercept the first impression, or rather to prevent propagandists from creating it, is to create an extension for a web browser that can analyze materials in live time. I opened a web page - I received analytics with warnings: here, there and here - signs characteristic of propaganda!



Several other theories from sociology and the field of media studies speak in favor of developing such a plugin.

1. , (“agenda-setting theory”), – , (, ). , , , . , , , . , – , . , , . (, check-list; ).
2. (Eli Pariser) « » („filter bubble”), , , ( ) . - , , . , , , . , , , « » („confirmation bias“): , . « » . , - , , ( «-», “echo chamber”). , , , , , . ( ).

. – . , - – - , .



, , . . NLP (Natural Language Processing), . - - - , : , - . , ,

1. ;
2. .

, ( ), , , .

, , , , . – « » « » ( «, »). , , - .

Well ... that's how it should work somehow.



Obviously, there are different media formats: text broadcasts, video reports, analytical telegram channels and talk shows. And each format has its own characteristics, which should probably be taken into account. Or not? Either way, I suggest starting with a simple format: articles (i.e., texts) on news sites. The person opens the site and starts reading the texts on this site. Put yourself in his place: what should he pay attention to? How exactly can he understand what is in front of him - an attempt at manipulation?



I have developed a methodology that does not claim to be complete, but which should cover an essential part of the formal features of propaganda. It is impossible to formalize everything, and we will discuss this point later. But many of the signs can be found as a human being, and it is quite possible that algorithms can cope with this task.



I have divided all the rules for signs of propaganda in news texts into several logical categories.

Material feeding quality

This category contains several rules that news journalism must adhere to.

Was the context given?

Meaning : At the end of a news text, you can often find a brief description of the events that preceded the event. For example, if the news contains excerpts from Trump's last interview, then at the end it is usually explained that the presidential race is currently taking place in the United States, and according to sociological polls, so far the leadership is with someone, and that person thinks like this.



Criticism : Not all events have a context - some of them just happen. For example, the explosion in Beirut.



Automation :

• These can be bullet words such as "recall that."
• You can try to automate the extraction of the main meaning from the articles of the publication, and keep these main theses for some time. When new articles appear, the last paragraphs are also reduced to the main theses, and in the database of past publications of this edition, the search for the most similar theses is carried out. If none were found, then either the event is completely new, or the context was not given.
• Since not all events can have a context, the absence of it at the end of the article is not a strict criterion, and if there is an assumed absence of context, a warning should be displayed, not an "alarm".

Availability of sources

Meaning :

• As a general rule: if there is news, then it should be said where the news came from - after all, news always has some source.
• If the article contains a link to some source of information (for example, a politician or other media), then a hyperlink must be given.

Criticism :

• What to do with anonymous sources (without which journalism cannot exist)? What to do with investigative journalism (which is different in format from news)?
• Not all sources may have their own sites. Or, theoretically, a situation is possible in which information was broadcast on TV, but not on the Internet. Both factors can make hyperlinking difficult.

Automation :

• The algorithm should be able to determine the presence of information sources in the text. Either by keywords: "reported", "said", or using NLP techniques.
• Checking for hyperlinks is easy;
• , , ?
• , (, ) .

Meaning : if the article is translated in whole or in parts, then was the translation made correctly?



Criticism : It is almost impossible to automatically assess the correctness of a translation. Since manipulation can be hidden in the choice of individual words, we need an algorithm that is very sensitive to the smallest shades of concepts, which, as far as I understand, do not yet exist. This means that it will be unrealistic to find this kind of manipulation in translation. But what can really be discovered is whether the original article actually contained a thesis similar to the one in the translation (or was it completely invented?).



Automation : translating either the original text into the target language, or vice versa, through a call to some API, and comparing the two texts for semantic proximity (I hope I called it correctly?).

Content analysis

This category contains actions that can be applied in addition to those listed in the first paragraph.

If there is a conflict: have all parties been heard?

Meaning : the simplest example: if the president criticizes the oppositionist, it would be good if the news outlet asked the oppositionist's opinion on this matter.



Criticism: such a rule does not always make sense. If the president as a whole says about the protesters on the streets that they are puppets of the West, then it is not entirely clear who to ask for an opinion on this matter, and whether an answer to this remark will add information value to the material. Therefore, one should try to formalize this rule more rigidly: for example, if a particular person or organization accuses another particular person or organization of something specific, then you really need to ask the opinion of the other side. But the reaction of the other side may not follow immediately and is published some time later, which again causes the complexity of the implementation of this rule.



Automation :

• Just as we are going to automatically identify the source of information, we need to learn how to identify the parties to the conflict (if it is possible algorithmically).
• Just a thought: does it make sense to store information about socially significant persons and organizations in a social graph? In order to understand who to expect a response from?

Comparison: how did other publications cover the same news feed?

Meaning : compare the submission of events by this publication with the submission of other publications.



Criticism : to be honest, I don't see any weaknesses. If you find it, tell me.



Automation :

• Select the main theses or concepts from the text and search for these theses or concepts in other socially significant publications (we will talk separately about how to define them).
• A simpler solution is to offer keywords based on the analysis of the text, for which the reader can immediately start looking for additional information, for example, in Google News.

Analysis for manipulation

If the previous categories offered some quality standards for news materials, then in this category we will talk directly about intentional manipulation by means of expressive means.

Newsmaker bias

Meaning : manipulative actions can be committed not only by the media themselves, but also by those who are quoted by the media. Therefore, it is quite good to know what the subject previously said about the phenomena he is speaking about now.



Criticism : any person has a certain set of attitudes, values ​​and worldviews, that is, in fact, any person is biased. Will the recipient be helped by the knowledge that Vyacheslav Volodin had previously attacked the West? Well, perhaps, using the example of other people, it will help to think about your attitudes and how flexible they are. What do you think?



Automation : It is possible to maintain relationships between actors in a social graph. And under the edges of the graphs, keep the past statements of subjects about each other.

Marker words

Meaning : There is a whole bunch of words that can betray manipulations or gaps in the argument. For example, words of expression such as "obviously that" or "we all know that."



Criticism : it should be borne in mind that it is possible that in some contexts such expressions are appropriate (“it is obvious that violence is evil”). Therefore, this criterion may not be hard, but soft (warning, not alarm). Also, the process of creating a database of similar words and expressions can be quite costly in terms of effort (probably, you need to read a lot of texts).



Automation : search by words (and their declensions) and expressions (and semantically very close to them). The search can be carried out both by the statements of newsmakers and by texts in general.

Analysis of word shades

Meaning : I called this function this way because this name most closely conveys its meaning. Is there a difference between saying “protesters took to the streets” or “protesters have filled the streets”? Very often manipulations consist precisely in the use of certain vocabulary, which can probably be called "evaluative". That is, there is not just a statement of the fact, but a certain assessment of this fact is immediately given.



Criticism : This should be a difficult task for the algorithm, but I came up with another solution (which, however, is also not an easy one - see below).



Automation: Psycholinguists have developed a model called the Linguistic Category Model, which has a tool such as a “scale of concreteness / abstractness”. To understand how it works, let's look at the following statements:

• – .

I, of course, exaggerate, and forgive me, linguists, if I cited formulations that were not entirely correct from the point of view of this theory, but the general meaning is something like this: there are 4 levels of concreteness / abstractness, from the most concrete formulation (factual) to the most abstract (evaluative ).



In order to apply this model in the analysis of texts, one could create a markup of concepts. Since it is probably unrealistic to create markup for all linguistic situations, and it is not even necessary, we can concentrate precisely on political discourse, creating markup for situations that are often encountered in the political and media field. That is, it will be something like a synonym base, in which each word has its own "sparring" from other levels of the scale of concreteness / abstractness.

State propaganda narratives

Meaning : I have already said that I am primarily interested in state propaganda, and although all the previous points can be used to search for any propaganda, I dedicate this point to the state propaganda. The fact is that the set of plots that the state propaganda uses in its materials is quite limited and is rarely replenished with new plots. How many years have we heard about "puppets and special services of the West", or "Ukrainian nationalists"? That is, new mini-motives, of course, appear (“Navalny's diabetic coma”), but at some higher level of abstraction, as a rule, these are all the same old motives that were a year and five years ago.



Criticism : not found. If you find it, tell me.



Automation :

• , . , . , .
• .

Reference :



A few years after the October Revolution of 1917, the children's poet Kornei Chukovsky wrote: “ Soviet people remember with the greatest pride how the revolution has miraculously renewed our language. She cleansed him of such disgusting words as a Jew, a Little Russian, a foreigner, common people, peasant, etc. From the active words they immediately became archival . " Unfortunately, he was wrong, because not all of these words were eradicated, and many were replaced by even more offensive ones.



The words that the children's writer listed refer to the lexicon of the so-called "hate speech" (synonymous expression - "rhetoric of hatred"). There is no generally accepted definition of “hate speech” in international law, but basically this term is interpreted in the context of hatred of people who have the so-called “protected characteristics” - these are personal characteristics for which you cannot be discriminated against: race, skin color, gender, language, religion, political or other opinion, national or social origin, property, age, birth, indigenous origin or identity with an indigenous people, disability, migrant or refugee status, sexual orientation, gender identity, or intersex variation.



In general, this is a fairly broad category of features. Enmity, on the other hand, is “an intense and irrational feeling of condemnation and disgust towards an individual or group chosen as an object because of the possession of certain — real or attributed — protected characteristics (recognized by international law). “Hostility” is more than just bias, and it is necessarily discriminatory. "Feud" is a manifestation of an emotional state or opinion and is thus different from any overt action. " And this is an important point - the use of "hate speech" does not necessarily lead directly to any specific consequences. That is, when you hear “let's slaughter the hobbits,” you understand that the call for genocide sounds. And how to relate to the assessment “our rotten, corrupt,stinking opposition "(both formulations are invented, any coincidences are accidental)? There is no call to action, if someone voluntarily goes to do something with the opposition, then it will be problematic to find a connection between his actions and the assessment given by someone from public figures, and nevertheless, this assessment sounds at least bias, but judging emotionally, it may well be enmity.



In general, it is difficult to find clear interpretations and boundaries of these terms. For example, deeply offensive language does not automatically refer to hate speech - if it does not sound in relation to some object precisely because of its protected characteristics. In the context of "hate speech", the terms "in-group" and "out-group" are often heard - that is, when using "hate speech" it means a kind of confrontation between two groups, the presence of a division into "friends" and "aliens". But there are also clearer classifications of "hate speech". According to one of them, formulated by the researcher Verkhovsky (Verkhovsky A. M .: "My language ... The problem of ethnic and religious intolerance in the Russian media"), there are three large categories of hate speech:

• harsh (for example, direct incitement to violence)
• medium (for example, accusing the negative influence of a social group on society, the state, or justifying historical cases of discrimination and violence)
• and mild (for example, asserting the moral flaws of an ethnic group or mentioning ethnic group names in a derogatory context).

Therefore, the ethnophilisms so beloved by everyone (offensive nicknames of representatives of other peoples) are also hate speech.



Meaning : the media text should not contain hate speech. Not journalistic material, not a news story, not just a column with the author's opinion.



Criticism: Even representatives of NGOs, whose report I quoted above, admit that there are contexts when hate speech should not be punished. For example, when the intent could be either more frivolous (for example, an ill-conceived or frivolous comment on social media) or more complex (satire or provoking discussion on a difficult issue, for example through art) than an act of enmity. That is, it can be a rather difficult task to confidently determine the hate speech, especially since the criteria for different researchers differ (some attribute the denial of historical events to hate speech, while others do not).



Automation: Creation of a base of expressions related to hate speech. You can create a database of any offensive words and expressions at all, and although this may complicate the task of collecting data, it will simplify the search in the text for expressions that manipulate the recipient's opinion about a person or group of people. Then the function will be called not "search for hate speech", but "search for expressive expressions" or something like that. It is not very clear, however, what to do with Dostoevsky's The Idiot in this case.

Database of language stamps of state propaganda

Meaning : in the texts of state propaganda, you can often find common expressions that are characteristic of this particular genre. For example, "a rally was held in support of the current political course of the country."



Criticism : It is hard to find such expressions and make sure that they were used in the propaganda discourse. Furthermore, we cannot assume that these expressions are always used only in propaganda media. Perhaps this criterion is redundant, but if its expediency is nevertheless proved, then it should be a "soft" criterion for propaganda.



Automation: As in the previous paragraph, I do not know in what format these expressions should be saved. After all, we probably need declensions / conjugations if we want to search by marker expressions? If we follow the path of learning the algorithm, then do we need a text corpus? Sorry, as I said, I'm a layman in computational linguistics.

Sentiment analysis

Meaning : propaganda, as we found out, tries to influence emotions. It often evokes negative emotions not because of the presentation of a fait accompli (anger over the beating of a person), but because of interpretation, blame, and value judgments.



Criticism : I do not know how good modern sentiment analysis algorithms are, and what limitations they have (for example, the size of texts) to ensure good results.



Automation : analysis of sentiment by paragraphs, chapters or the entire text, and if the negative was found, then the question to the user: "Probably you feel X. Because of a fait accompli or someone's interpretation in relation to Y)?"

Analysis of media quality

This category contains several suggestions on how to assess the “quality” of media. Quality means the extent to which the media meets the criteria described above for the quality of presentation of the material and honesty to the recipients (that is, when it does not try to manipulate them).

Ignoring events

Meaning : quite often, propaganda resources simply ignore certain events, so that these events do not exist in their "world", and the recipients do not learn about certain events that the manipulator does not want them to know about. For example, the Russian state media have long ignored the presence of such an opposition figure as Alexei Navalny.



Criticism : this begs the question of how to define socially significant events. You also need to take into account the fact that the media may have some kind of specialization - for example, economics and finance. In this case, the business newspaper may not mention the event in political life. To take into account such nuances, it will be necessary to carry out an objective classification of the media on the basis of some transparent criteria, which, for a number of reasons, will not be such an easy task.



Automation: squeezing out the main meaning from each media piece  comparing such "tapes" from different media outlets and identifying socially significant events  checking whether the given media outlet wrote about this event. To be honest, I don't know how to bring this idea to life.

Analyze past media stories to find patterns

Meaning : it is difficult to characterize media outlets based on one or several materials. It is also problematic to give the media a label of “quality” once, because the editorial staff may change. Therefore, it is necessary to characterize the media by a certain number of the last N materials. Here are the aspects to analyze media materials in a certain time continuum:

• Bias towards information from government agencies; lack of materials where the opinion of the opposition or ordinary people is presented;
• Past use of propaganda narratives or hate speech in journalistic material
• Systematic violation of quality standards for the presentation of materials

Criticism : I did not find "weak points".



Automation : saving the features detected by the above functions in the time series format or some other. Perhaps the user can also save some notes himself.



Reason 1: You can create a broader list of media quality criteria, on the basis of which the plug-in users or some group of experts (although I advocate the openness of all lists, procedures and data) will evaluate the media.



2: , . : , 33 «» 170 « ». , : « ...». , , , , . , .

There is an objective difference between different media formats, which creates the preconditions for defining separate functions for some formats. For example, in talk shows or other video and audio formats, you can try to analyze people's voices and measure their loudness (after all, we remember that propaganda likes to whip up, press on emotions - and that is why they shout so much in political talk shows). Some formats are difficult to analyze automatically. For example, video news, because even a person will not always be able to recognize manipulation in a report (that's why I would deduce the golden rule - “do not trust the edited video - only the primary sources, given without comments”).



Otherwise, I'm not inclined to see the difference in formats. That is, of course, it is. For example, in telegram channels, value judgments are found through the word - this is often their "salt", while in the news materials of a serious editorial office, they will be inappropriate outside of direct quotes. But this difference is insignificant for the user, because our goal is to teach him to recognize manipulations, and when he learns, he will be able to understand the difference in formats without our plug-in and choose an acceptable presentation of information.

Conclusion

I would like to finish the material with a few important remarks.

1. The purpose of the plugin is to increase people's awareness of reading news, to teach them to identify manipulations through expressive and other means.
2. I want to do this through as formalized rules as possible. Of course, not everything can be formalized, and very often we recognize propaganda, as they say, “in a complex”.
3. The plugin is expandable. For example, you can create a whole system for evaluating the media and even individual journalists, bloggers and authors in general (for example, authors of posts on Telegram). That is, each plugin user will be able to rate the media or the author, and these ratings will be aggregated into a kind of rating for the media and authors. But that's all later.