DEMOCRACY, FREEDOM OF EXPRESSION AND DIGITAL SPHERE
ANALYSIS OF TRENDS AND TOPOLOGIES IN TWITTER
THE CASE OF #REDAMLOVE
published on February 28, 2019
(This is a free translation, y you want to collaborate to make it better please share in comments)
Composition of Signa-Lab from the photograph of: Faridh Mendoza / CC BY 2.0.
“Pleased to meet you
Hope you guess my name,
Ah, what’s puzzling you
Is the nature of my game “
The Rolling Stones
The centrality of the internet and the reconfiguration of the public
Internet, its different platforms, engines and especially the so-called social networks such as Twitter, Facebook or Instagram, have now become spaces for the construction of narratives and social and political imaginaries in constant dispute. In recent years, it has become common news to use intervention strategies of these tools to alter debates in the public sphere. From controversial issues that seek to silence the opposition or critical voices, to electoral processes as was made visible in the case of Cambridge Analytica , the company that intervened in the Brexit campaign and in the election in which Donald Trump was elected, or the confessions of the Colombian hacker, Andrés Sepúlveda, about the digital manipulation of the presidential campaign in Mexico in 2012, until the most recent case of the Brazilian elections, in which in addition to the mentioned networks, whatsapp was used to orchestrate a campaign of hate and fear.
Despite the advance of technopolitics, that is, the appropriation and critical use of networks for citizen participation, the visibility of key issues, the denunciation and especially the organization of discontent, the growth and expansion of bad practices on the Internet is remarkable . The use of automated accounts to accelerate the production of trending topics or trends, the creation and use of accounts known as “trolls” to direct attacks against specific accounts or “uncomfortable” issues for certain power groups, among other strategies, have proliferated in Mexico and in other countries. The real is constructed from images and narratives, from the mobilization of emotions, and networks are those interfaces in which it is possible, paraphrasing a text by Amador Fernández-Savater “construct a political climate” nowadays. That is, these platforms can be production and control tools of a certain hostile environment that prevents conversation.
During the administration of Enrique Peña Nieto (2012–2018), there were constant cases of the use of these strategies, in this period the term “peñabot” was coined, which even has an entry in Wikipedia . As of 2016, once Signa Lab was inaugurated, the laboratory focused part of its academic efforts on generating the tools, methodologies and in-depth multilayered analyzes of the relevant events and discussions for democracy, freedom of expression and justice not only in Mexico, but in different parts of the world, where the centrality of the digital is key to the plurality and strength of the public sphere. Since then, Signa_Lab has carried out different studies and analyzes around three key dimensions: the use of socio-digital networks as tools for the self-organization of society, as in the case of the feminist movement known as # 8M because of its network label; the growth of the polarization that finds in networks a favorable space for social tension, as in the case of the so-called migrant caravans ; and it has followed with special interest what we call organic and artificial narratives in the construction of the representation of reality, the latter, the artificial and atypical ones were recurrent throughout Peña Nieto’s presidency and at very different junctures .
Context.The disturbing polarization
In recent months, the political discussion around central issues, such as electoral processes, exodus of migrants and gender violence, in Mexico and internationally, has generated high levels of polarization, which has prevented the construction of conditions for dialogue.
In Mexico, within 100 days of reaching the government, Twitter followers of President Andrés Manuel López Obrador, have launched various attacks and ridicule campaigns to journalists, media and users with a critical stance towards the actions of the leader. Examples of this have been trends such as #PrensaFifi or #Chayoteros, which openly call on users who want to defend, attacking, to criticize the president.
Both hashtags have been used in different conjunctures, however during January and February 2019 they became several times trends that took up much of the online discussion.
The following graphs (a graph is the result of an algorithmic analysis of a set of data indicating relationships, in this case, of Twitter) show the relationships between users who used the terms #Chayoteros and #ProfileFifi, obtained through the download script of Twitter, THOTH, developed by the laboratory. There are 16,131 tweets from which they are detached, for #Chayoteros: 9,428 nodes or accounts, and 21,225 edges or interactions, grouped into 156 communities (colors); and in the case of # PressFifi: 3,220 accounts and 6,184 relationships, grouped in 177 communities.
Graph 1. Relationships User to User of HT #Chayoteros . Prepared by Signa_Lab.
Communities: 156
Nodes: 9,428
Edges: 21,225
* Move the cursor over the image to see detail.
Graph 2. User to User Relations of HT # Pressfifi . Prepared by Signa_Lab
Communities: 77
Nodes: 3,220
Edges: 6,184
* Move the cursor over the image to see detail.
The constant disqualifications have been escalating not only towards the journalists of the most conventional press in Mexico, such as Loret de Mola or López Dóriga, but especially to the journalists who cover “las mañaneras”, morning conferences in which López Obrador uses often adjectives such as “press fifí”, which accuses slander and “flown.” For example, Ivonne Melgar, reporter for Excelsior and Image Television, author of the column Rearview and writer of the magazine Woman is More, received a wave of attacks that are reflected in the first graph and that motivated her to write the following tweet.
In addition to the visualizations, a cloud of frequencies was made of the words used in the tweets that used both hashtags (#PrensaFifi and #Chayoteros), which yields an interesting result in terms of constructing a narrative of the press as an enemy.
Image 1. Word cloud of HT # Pressfifi . Prepared by Signa_Lab
Image 2. Cloud of HT #Chayoteros words. Prepared by Signa_Lab
The semantic relationship between the most used words in both hashtags shows a persistent expression of “repudiation” towards the press and spreads the idea that the president is being unfairly attacked.
Throughout several weeks we have followed qualitative and quantitative the issues, discussions and hashtags that are linked to the management of President López Obrador. For the purposes of this study, it is interesting to start with the smear campaign against the newspaper Reforma, launched from the account of the user @Fafhoo, which is identified as the founder of the # RedAMLOVE.
Through the hashtag #ReformaTodoLoDeforma a call was made to the followers of this network. Said hashtag was generated in the context of the discussion about some notes published by Reforma, about a department in Houston not included in the 3 of 3 of the Secretary of Communications and Transportation, José Jiménez Espriú , and days after it was made public, as well through the Reformation, that in the patrimonial declaration of the Secretary of the Interior, Olga Sánchez Cordero , to say the middle, did not appear a department of his property, also in Houston.
The @LoQueSigue portal published a brief report questioning HT #ReformaTodoLoDeforma. The response of this network was immediate, in addition to attacks and disqualifications to the account, HT #NoSoyBotSoyReal was reactivated, which had appeared in the electoral context of 2018 and that this time, operated as a form of response of these legions to the criticism leveled against them not only at Reforma, but also at different communicators and at the climate of constant harassment against journalists, which has been growing with the impact that the press conferences “tomorrows” have on the construction of the public agenda “ of the president.
1 Transparency exercise done by public officials that contains: 1) declaration of assets, 2) declaration of interests and 3) proof of payment of taxes.
Average behavior and atypical behavior online
During the 2018 presidential campaign in Mexico, Signa_Lab, in conjunction with OpenLabs and Enjambre digital, published a report to explain the difficulty in these latitudes to categorically separate profiles of Twitter as “bot” and as “no bot” from numerical indicators only, without taking into account qualitative analysis parameters, to the extent that today there are profiles with behaviors that integrate both automated and non-automated features.
Since then, we have continued the timely monitoring of the online political discussion in Mexico, and fine-tuning the methodology to analyze the behavior of Twitter accounts, to build useful categories in the identification of profiles that follow average patterns of content dissemination and profiles with patterns atypical content dissemination.
For the case that concerns us here, 98,806 tweets were obtained that used the hashtags #ReformaTodoLoDeforma, #NoSoyBotSoyReal, and #RedAMLOVE.
The graph below shows the relations between users and hashtags (U2HT) 2 of the three mentioned tags. In this graph we identified 22,971 nodes, 63,814 edges and 66 communities. The evident blurring between the colors blue, green and pink (which represent the largest communities in this conversation), is a sample of the articulation between users who used the three hashtags. In other words, it is possible to affirm that these labels and these sets of users are part of the same conversation .
Graph 3. User Relationships to Hashtag # RedAMLOVE , #NoSoyBotSoyReal and #ReformaTodoLoDeforma . Prepared by Signa_Lab.
* Move the cursor over the image to see detail.
2 These types of relationships show the traction that a set of labels or specific users have among users. They allow identifying the centrality of hashtags or users, that is, the importance of condensing these as words, phrases, slogans or actors, based on the size of the user clouds around each relevant tag or account.
The subnets of the #RedAMLOve
From this data set , several filtering processes were carried out, with which the accounts with the greatest interaction in the conversation were searched. In addition, the accounts of political figures such as that of López Obrador, Felipe Calderón and Vicente Fox, which due to their enormous capacity to generate discussion, tend to make invisible the actors that operate the strategies related to this report. This process yielded a total of 176 particularly active Twitter accounts between January 31 and February 7, 2019, period in which the aforementioned hashtags were deployed and which constitute a representative sample of the # RedAMLOve.
From these data, user-to-user 3 and user-to-hashtag relationships were analyzed through Gephi, an open source program that allows working with large volumes of data that register relationships.
In this way, we obtained the degree of entry ( sum of the number of mentions and retweets that a specific profile receives ) and the degree of exit ( number of retweets and mentions that one account does to others ).
A high degree of exit indicates that an account retweeted in large quantities and constantly to one or several users; a low degree of exit, indicates that an account did not retweet or mention so often to other accounts.
Graph 4. Comparison of the degree of entry and degree of exit of #RedAMLOve users . Prepared by Signa_Lab. ‘
A high degree of entry indicates that an account was much mentioned or retweeted; a low degree of entry indicates that the account was not mentioned or retweeted.
Graph 5. Comparison of degree of entry and degree of exit of users of #ReformaTodoLoDeforma . Prepared by Signa_Lab
Once these two measurements were obtained, we separated those accounts that had the highest degree of entry (the most mentioned and retweeted) and the highest degree of exit (those that most retweeted and mentioned others). Then, to give more support to the methodology, we use the tools Atrapabot , Botometer and Proton Pack 4 , which use variations of the same script to throw a number that represents the probability that an account is “bot” or not. Next, we averaged the results of the three tools to have a quantitative indicator for each account. In turn, with Botometer we collect the following information:
- Hours of greatest activity of each account
- Average weekly tweets for each account
- Percentage of retweets made by each account
As part of the methodology and to provide comparative elements that facilitate the understanding of what a typical or organic behavior of an account means, and an atypical one, 10 highly active profiles were selected on Twitter, regardless of their tendency or ideology, with In order to compare the intense patterns of tweeting of these, with the patterns of the 176 accounts retrieved from the #RedAMLOVE.
From the detailed analysis of these profiles, we established maximum limits to consider an account as atypical, when its activity peaks exceed 30 tweets per hour .
Another element considered for the validation of the category that we propose to call “atypical content diffusion”, was to analyze profiles of media and national and international journalists, characterized by constantly publishing content. The tracking of these accounts shows that they tend to reach 200 publications in three days or more, while, in some cases, the most active accounts of the 176 selected for our analysis, reached that peak in just two days. That is, making more than 200 publications in two days is another indication of atypical pattern of content dissemination .
National and international media accountsUserTweets per week% RetweetsBBC World16003ExpansionMx790oneHuffPost7109nytimes63016the Economist6108washingtonpost5803reform5604ElFinanciero_Mx550oneAristeguiOnline4300RegeneracionMx210oneAVERAGE6674.6MEDIAN5953
Table 2. National and international media accounts on Twitter. Prepared by Signa_Lab
In a next step, and in order to give the study the greatest possible strength through a multilayer analysis, we enter another series of relevant data from each of the 176 accounts:
- Date of creation of the account.
- Number of followers
- Description of the account
This information expands the qualitative description of the atypical profile of content generation, since if they are less than a year old, they have many followers and follow a few accounts, have responded little or much to other users and have or not a Description in your profile, will help to understand if they were created expressly to orchestrate the attack strategy, if historically they have behaved as they did at this juncture, if they only responded at this time, or if they voluntarily joined the same .
Image 3. Cloud of words most used in the description of active accounts in #RedAMLOVE #ReformaTodoLoDeforma and #NoSoyBotSoyReal . Prepared by Signa_Lab
The semantic relationship between the words used in the descriptions of the accounts articulated with the # RedAMLOVE, shows a strong positive charge and the appeal to emotions in the self-description with which these figures are presented on Twitter.
The result of this systematization has been a database (with which we continue working in the laboratory) with more than 20 observations of each account, which has allowed, until now, the development of four qualitative categories to identify the profile of an account in a conjuncture of conflict or online attack, its characteristics in the topology of the network and its role in a given discussion.
It is important to remember that there are patterns of publication of certain human accounts that resemble the pattern of publication of automated accounts used for propaganda purposes or infiltration of public, a situation that only confirms the risk of asserting in a clear manner that an account is or not a “bot”, since the automated behavior not only responds to the programming of an account as a content repeater, but can also respond to the will of a user to reaffirm a posture by consecutively replicating the same message. Therefore, it is necessary to develop more complex categories than “bot”.
3 This type of graph shows the modules or communities of users (by colors) interacting within a network. It allows to identify tensions between actors, groups of dispute by the sense of a topic or conjuncture, and volume of users around a discussion.
4 Script developed by Signa_Lab to detect automated accounts and which, unlike the previous two, has the capacity to analyze up to ten thousand accounts in a single run.
Categorization of accounts
Master of ceremony (MC)
We take this denomination as a metaphor for the so-called masters of the hip hop culture, an “em si” (pronunciation in the music scene), who creates the letters, recites the lyrics and especially masters the metrics. In this analysis, the MC figure has a high degree of input, a low degree of output (with little content generates a lot of movement); MCs are the generators of content that circulates more in a tendency, they mark and define the narrative. That is, they are the ones who make the “call”, they articulate the “chorus” and they encourage the fans (as in the case of #ReformaTodoLoDeforma ).
Master of Ceremony (MC)UserAverage Tweets per week% RetweetsAleFerruzcaL63048ELPoderdelaInfo53083SinlineaMx500oneFafhoo39066AntiTelevisaMx25089lovrega2108rubbistein20047AlvarezBDaniell1300Luouis0097410LuisLeonardoF136455AVERAGE297.840.7MEDIAN23047.5
Table 3. Users identified as MCs. Prepared by Signa_Lab
Chorus (Bots or semibots)
They are accounts with a high degree of output and very low or sporadic entry level, have an automated behavior, “like” and massively retweet the contents produced by the MC’s.
BotsUserAverage Tweets per week% Retweetsssamtaoying8500100OrlandoSannz640083FaridGonzalez2570067camaney195200100melflor93100100Veronic712629102300100_AceDCopular_220074ALMARGG12220079MaasAguila220097azteca_vikingo1900100AVERAGE397090MEDIAN270098.5
Table 4. Users identified as bots. Prepared by Signa_Lab
Troll
These are accounts that are used to attack other users, usually with reference to a specific topic or trend;they fluctuate between the production of own content (high degree of entry) and the replication of contents of the MC’s (high degree of exit), according to the narrative in dispute.
TrollsUserAverage Tweets per week% Retweetsvalme1453061kokoGracia1135041Mx_SinCorruptos30074act_just2903677Faustino7725038workshop20061004mags_6138838abymmora7169mimatsal4658Stupormundi73717AVERAGE206.243.6MEDIAN17539.5
Table 5. Users identified as trolls. Prepared by Signa_Lab
Fans
They are real accounts, people, followers or supporters who do not participate in the inorganic or atypical logic in a trend, but who rely on the contents promoted by the MCs and therefore, tend to retweet (high degree of output), adding to a tendency without dimensioning the strategy with which they are joining.
The 176 categorized accounts can be appreciated in the following interactive visualization :
Interactive categorization of the 176 accounts analyzed. Prepared by Signa_Lab
The detailed analysis of network posting patterns is now shown, which strengthens the hypothesis that we are facing a non-organic architecture and narrative:
Display of schedules of greater activity by type of atypical content diffusion account
As can be seen in this analysis, it is possible to assume, with quantitative and qualitative criteria, methodological strategies for critical monitoring and generation of depth layers, that the #RedAMLOVE is a sophisticated operation of producing political narratives through, at least, three dimensions:
- Automated content replication
- Attacks and confrontation.
- Production of agenda and framing.
The network and its geolocation
From georeferencing 5 the information of the accounts was obtained a distribution of each tweet issued in the downloads made on the #RedAMLOVE, (64,619 tweets) and #ReformaTodoLoDeforma (20,103 tweets).As can be seen in the images that follow, the distribution of the accounts that make up the RedAMLOVE is practically similar to the network that tweeted about Reformatodolodeforma.
Image 4. Geo-referenced distribution of tweets issued with the HT # RedAMLOVE. Prepared by Signa_Lab.
Image 5. Geo-referenced distribution of tweets emitted with the HT #ReformaTodoLoDeforma. Prepared by Signa_Lab.
Most of the followers are located in CDMX but the amount of accounts present in the US is also significant, mainly on the west coast. However, within the use of the same hashtag support messages coexist as well as some critical or ironic messages with the network. So the number of total tweets is not representative of the number of followers or the size of the network, but they account for the expansion of the discussion and debate, which is still much broader than other debates or controversies about the Mexican territory.Here is the interactive map where the tweets issued in the discussion on #reformatodolodeforma appear.When approaching the clusters, each tweet is broken down and the text of the tweet can be read:
Interactive map of Tweets issued in the discussion on #reformatodolodeforma. Go to full screen version.
5 The georeferencing of the accounts is done from the assignment of latitude and longitude coordinates to each one based on the information that the user provides as text in the description of his account and / or if his geolocation has been activated. It is a semi-automated process managed by a Google API. So the location is only indicative and not definitive.
Analysis
In a logic of war that is confused with politics, in which you have to choose sides or stick to the friend-enemy distinction proposed by Carl Schmitt (1999) 6 , the strategy exerted by # RedMLOVE seems to be extremely effective. Well, as Delgado (2011) comments on Schmitt’s criterion, “the purpose of war is not to annul the enemy, but to disarm him, to domesticate him, so that he surrenders to the opponent in the relationship.”
One of the effects of the coordinated and massive attacks on users who disagree or disseminate information deemed anti-amlo by this network, is that it contributes to generating a negative image of the followers of AMLO. Well, several messages are highly offensive or include threats. In the case of the trend #ReformaTodoLoDeforma, it is difficult to know if there is an intentional use of the “network” pro-AMLO to delegitimize the image of the followers of the 4T 7 , or if, on the contrary, adherence to the network and its messages, agglutinates an adequate representation of these followers. In any case, we consider that this does not favor at all to the image of the government and specifically to that of the president, and that, on the contrary, it subscribes to the unnecessary climate of polarization.
Another effect is to produce a coercive fence in relation to the information or critical opinions towards AMLO, so that they find it increasingly difficult to produce, in the first place, without waiting for the reprisals of the network. This may result in critical voices, necessary in a democracy, retreating due to the virulence of attacks (as in previous sexenios in which an army of bots and trolls was used to silence criticism and dissent).
Finally, whether intentional or not, another effect is that a separation is achieved between the official channels of the president and the multitude of followers and initiatives related to the president, as is the case of the network in question. Not being very clear the extent to which the president supports or condemns the actions of the other party, or actions that can be attributed as their own. This opens a double game in which, for convenience, one can demarcate and at the same time use those same channels as a sample of validation of proposals. Within these separations, one should also ask whether the emergence of these networks, coordinated in appearance by a single node, effectively translate the feelings and thoughts of the followers of AMLO.
The logics of acceleration in the dissemination of events in the digital arena have been useful in recent years to activate mobilizations and protests against various injustices on a global level. The speed with which public pressure on political actors such as governments and companies scale from on to offline is a central factor in understanding current social reconfigurations. The more people share, comment and react to a set of common ideas (tweets, images, memes, videos, etc.), the more likely that the topic permeates public opinion and is placed on the agenda. Under this same logic, attack campaigns such as those analyzed here have been able to build online censorship and harassment strategies through the dissemination of an idea through tweets and hashtags that act as “pollinator” agents, in charge of to spread a message in a capillary way. What is the message in this case? One dangerous for any democracy: keep those who criticize the president at bay.
According to these first analyzes, it would be desirable — to avoid the ambiguities in the use of pro-AMLO attack networks, that there be an official pronouncement from the presidency or from the office of social communication of the same, to clarify what is the position of the government against these online non-organic strategies.
Finally, the exploration of categories of accounts that participate in these political operations on Twitter is not limited to those shown in this report, but is likely to be extended. The category of “mutant” can be added in future analyzes to define those profiles that tend to change names in short periods of time and / or that delete their tweets after a certain period, this with the intention of camouflaging or hiding from the crawl, or maybe as a sign of the completion of a contract.
Coda The attacker observed
One of the attributes that we obtain about these accounts is the “ID”. It is a numerical sequence with which Twitter calls each account that is registered in the network. This number, unlike the username, can not be modified. Signa_Lab continues to monitor these trends to identify these profiles, regardless of whether they change their name.
6 In this regard, says Delgado, M (2011) that “The friend-enemy criterion implies the autonomy of the opposition and is conceived in relation to any other endowed with its own consistency. This shows the specific and controversial feature of the political. It is possible to love the enemy in the private sphere and in the public sphere develop the most intense political antagonism until the end of the war. “
7 Fourth Transformation. This is what Andrés Manuel López Obrador called his term in government, placing himself in a line of continuity with the historical “transformations” to which he intends to adhere to his mandate: the independence of Mexico (1810–1821), the reforms of Benito’s government Juárez (1858–1861) and the Mexican Revolution (1910).
References
- Delgado, M. (2011). The friend-enemy criterion in Carl Schmitt. The concept of the political as a ubiquitous and deterritorialized notion. Cuaderno de Materiales, 23, pp. 175–183. Retrieved on February 15, 2018 at: http://www.filosofia.net/materiales/pdf23/CDM11.pdf
- Schmitt, C. (1999). The political concept. Barcelona: Editorial Alliance.
Annexes
1. Categorization of accounts by atypical publication patterns
Creation of accounts by year. Prepared by Signa_Lab
2. Scope of the user’s network @Fafhoo
Graph 6. Output from User to User, combination of the @Fafhoo and @MxFaby accounts. Prepared by Signa_Lab.
Communities: 216
Nodes: 20,248
Edges: 52,253
* Move the cursor over the image to see detail.
Graph 7. User exit to Hashtag, combination of hashtags # RedAMLOVE , #NoSoyBotSoyReal and #ReformaTodoLoDeforma . Prepared by Signa_Lab.
Communities: 73
Nodes: 19,289
Edges: 50,797
* Move the cursor over the image to see detail.
3. Geolocation of Tweets from #RedAMLOVE
Geolocation of Tweets from #RedAMLOVE . Prepared by Signa_Lab.
4. Results of Proton Pack
Results of accounts processed by Proton Pack. As of .8 there are high chances that the account is a bot. Prepared by Signa_Lab.
Technological and Higher Studies Institute of the Western Peripheral South Manuel Gómez Morín # 8585 CP 45604 Tlaquepaque, Jalisco, Mexico Phone: (33) 3669 3434
