Polarization Rank: A Study on European News Consumption on Facebook

Polarization Rank: A Study on European News Consumption on Facebook

Ana Lucía Schmidt Dept. of Environmental Sciences, Informatics and Statistics
Ca’ Foscari University of Venice, Venice, Italy
analucia.schmidt@unive.it
Fabiana Zollo Dept. of Environmental Sciences, Informatics and Statistics &
Center for the Humanities and Social Change
Ca’ Foscari University of Venice, Venice, Italy
Antonio Scala ISC-CNR, Rome, Italy Walter Quattrociocchi Dept. of Environmental Sciences, Informatics and Statistics
Ca’ Foscari University of Venice, Venice, Italy
Abstract

The advent of WWW changed the way we can produce and access information. Recent studies showed that users tend to select information that is consistent with their system of beliefs, forming polarized groups of like-minded people around shared narratives where dissenting information is ignored. In this environment, users cooperate to frame and reinforce their shared narrative making any attempt at debunking inefficient. Such a configuration occurs even in the consumption of news online, and considering that 63% of users access news directly form social media, one hypothesis is that more polarization allows for further spreading of misinformation. Along this path, we focus on the polarization of users around news outlets on Facebook in different European countries (Italy, France, Spain and Germany). First, we compare the pages’ posting behavior and the users’ interacting patterns across countries and observe different posting, liking and commenting rates. Second, we explore the tendency of users to interact with different pages (i.e., selective exposure) and the emergence of polarized communities generated around specific pages. Then, we introduce a new metric – i.e., polarization rank – to measure polarization of communities for each country. We find that Italy is the most polarized country, followed by France, Germany and lastly Spain. Finally, we present a variation of the Bounded Confidence Model to simulate the emergence of these communities by considering the users’ engagement and trust on the news. Our findings suggest that trust in information broadcaster plays a pivotal role against polarization of users online.

keywords:
Facebook, News Consumption, Misinformation, Polarization, Social Media

1 Introduction

The advent of social media changed the way we get informed and shape our opinion. In 2016, post-truth was selected by the Oxford Dictionaries as the word of the year. The definition reads “relating to or denoting circumstances in which objective facts are less influential in shaping public opinion than appeals to emotion and personal belief”€™, that is, we select information and interpretations adhering to our system of beliefs (confirmation bias).

This phenomenon is not new, our cognitive abilities have always been limited, and social media and the consequent disintermediated access to an unprecedented amount of information solely exacerbated the process. Recent studies on massive datasets (376 million users) schmidt2017anatomy () showed that major disintermediated access to information is creating segregation of users into communities where they share a specific worldview and ignore dissenting information. Confirmation bias dominates news consumption and informational cascades foster the emergence of polarized groups around shared narratives quattrociocchi2016echo (); del2016spreading (); del2017mapping (); quattrociocchi2017inside (); zollo2017debunking ().

Important results (that served to inform the Global risk report of the World Economic Forum in 2016 and 2017) pointed out the pivotal role of confirmation bias –i.e., the attitude of acquiring information coherently with the individual system of belief– in viral processes as well as in the collective framing of narratives. In particular, one of these works zollo2017debunking (), showing the inefficacy of debunking, convinced the Washington Post to close its weekly column dedicated to debunking false rumors dewey_2015 ().

The process of acceptance of a claim (whether documented or not) may be altered by normative social influence or by the coherence with the individual system of beliefs as well-documented in the literature on cognitive and social psychology of communication nowak1990private (); moscovici1982coming (). At the extreme of the spectrum, conspiracy theorists tend to explain significant social or political aspects as plots conceived by powerful individuals or organizations, and with the so-called urban legends€ they share an important characteristic: the object of the narratives are inevitably threatening the established social order or well-being and are always an indicator of what communities and social groups deeply fear franks2013conspiracy (). These phenomena are evidently of great interest and can be considered as a sort of “thermometer” of social mood. Since these kinds of arguments can sometimes involve the rejection of science, alternative explanations are invoked to replace the scientific evidence. For instance, people who reject the link between HIV and AIDS generally believe that AIDS was created by the U.S. Government to control the African American population.

In this paper we focus on the interplay between users and news outlet on Facebook by comparing four European countries: France, Germany, Italy and Spain. First, we compare the pages’ posting behavior and the users’ interacting patterns across countries and observe different posting, liking and commenting rates. Second, we explore the tendency of users to interact with a variety of pages (i.e., selective exposure) and the polarized communities of pages that emerge from the users’ consumption habits. Then, we introduce a new method to calculate the percentage of polarized users when more than two communities are involved and thus rank the four countries accordingly. We find that Italy is the most polarized country, followed by France, Germany and lastly Spain. Finally, we present a variation of the Bounded Confidence Model (deffuant2000mixing, ) to simulate the emergence of these communities by considering the users’ engagement and trust on the news.

2 Materials and Methods

2.1 Ethics Statement

The data collection process was carried out using the Facebook Graph API fb_graph_api (), which is publicly available. The pages from which we downloaded data are public Facebook entities and can be accessed by anyone. Users’ content contributing to such pages is public unless users’ privacy settings specify otherwise, and in that case their activity is not available to us.

2.2 Data Collection

We generated a list of top news sources, in their official language, of France, Germany, Italy and Spain. The list for each country was compiled considering the Reuters Digital News Reports newman2015reuters ()newman2016reuters ()newman2017reuters (). We then obtained the official Facebook page of each news outlet and proceeded to download all the posts made from 1st January 2015 to 31st December 2016, as well as all likes and comments that have been made on those posts. The exact breakdown of the data can be seen in Tab. 1, while the complete set of downloaded pages is reported in Tab. 8 in the Supporting Information.

France Germany Italy Spain
Pages
Posts
Likes
Likers
Comments
Commenters
Users
Population
Table 1: Dataset Breakdown. Population according to the Reuters Digital News Report (2017) newman2017reuters (). Likers is the number of people that gave at least one like. Commenters is the number of people that gave at least one comment. Users is the number of people that gave at least a like or comment.

2.3 Preliminaries and Definitions

In this section we provide a brief description of the main concepts and tools used in the analysis.

2.3.1 Projection of Bipartite Graphs

A bipartite graph is a triple where and are two disjoint sets of vertices, and is the set of edges, i.e. edges that exist only between vertices of sets and . The bipartite graph is described by the rectangular matrix where , if an edge exits between and , and otherwise.

We consider bipartite networks in which the two disjointed set of nodes are users and Facebook pages. That is where is the set of Facebook pages of country and is the set of users active on pages belonging to . Edges represent interactions among users and pages, that is, either likes or comments.

As an example, a like given to a post on page constitutes a link between the user and the page so . We can then build the co-occurrence matrices and that quantify, respectively, the number of common neighbors between two vertices of or .

Only two graphs per country will be relevant for the analyses, and . These are the result from the projection of two bipartite graphs: one considering the users’ liking activity () and another considering the comments ().

2.3.2 Community Detection Algorithms

Community detection algorithms serve to identify groups of nodes in a network. In this work we apply three different community detection algorithms.

  1. FastGreedy (FG). It takes an agglomerative bottom-up approach: initially each vertex belongs to a separate community and, at each iteration, the communities are merged in a way that yields the largest increase in the current value of modularity clauset2004finding (). The algorithm stops when it is no longer possible to further increase the modularity. Due to its speed and its lack of parameters in need of tuning, this algorithm will be the main reference to compare against the partitions resulting from the application of other community detection algorithms.

  2. Multilevel (ML). It uses a multi-level optimization procedure for the modularity score blondel2008fast (). It takes a bottom-up approach where each vertex initially belongs to a separate community and in each step, unlike FastGreedy, vertices are reassigned in order to achieve the highest modularity.

  3. Spinglass (SG). It interprets the problem of community detection as one of finding the ground state of an infinite range spin-glass. In this algorithm, the community structure of the network would be the spin configuration that minimizes the energy of the spin glass, with the spin states being the community indices newman2004finding ()reichardt2006statistical ().

To compare the various community partitions and the similarity between different clustering methods, we use the Rand index rand1971objective (), where a comparison between two partitions yields a value between 0 and 1, such that 0 indicates that there is no agreement on any vertex between the two partitions, whereas 1 indicates that the partitions are exactly the same.

3 Results and Discussion

3.1 Attention Patterns

As a first step we characterize how different countries consume news on Facebook. We focus particularly on the allowed users’ actions through the entire period of the data collection: likes, shares and comments. Naturally, each action has a prescribed meaning. A like represents a positive feedback to a post; a share expresses the user’s desire to increase the visibility of a given piece of information; and a comment is the way in which online collective debates take form. Therefore, comments may contain negative or positive feedback with respect to a post.

In Fig. 1 we show the distribution of the number of likes, comments and shares received by the posts belonging to each country. As seen from the plots, all the distributions are heavy-tailed, that is, they are best fitted by power laws (as shown in Tab. 2) and possess similar scaling parameters with some notable differences when looking at the number of comments and likes (Tab. 3).

Action Poisson Log-Normal Exponential PowerLaw
FR comment
DE comment
IT comment
ES comment
FR like
DE like
IT like
ES like
FR share
DE share
IT share
ES share
Table 2: Maximum-Likelihood fit of the actions received by the posts of each country. FR: France, DE: Germany, IT: Italy, ES: Spain.
Figure 1: Complementary Cumulative Distribution Function of the comments, likes and shares received by the posts of each country.
Comments Likes Shares
FR
DE
IT
ES
Table 3: Powerlaw fit of the actions received by the posts of each country.

We continue our analysis by examining how users from each country interact with the pages. In Fig. 2, we show the distribution of the number of likes and comments given by the users according to each country. Once again, all the distributions are heavy-tailed, as seen in Tab. 4, with some notable differences in their scaling parameters when considering the commenting activity of the users of the different countries (Tab. 5).

Figure 2: Complementary Cumulative Distribution Function of the users’ likes and comments of each country.
Action Poisson Log-Normal Exponential PowerLaw
FR comment
DE comment
IT comment
ES comment
FR like
DE like
IT like
ES like
Table 4: Maximum-Likelihood fit of the users’ different actions by country. FR: France, DE: Germany, IT: Italy, ES: Spain.
Comments Likes
FR
DE
IT
ES
Table 5: Power law fit of users’ attention patterns.

3.2 Selective Exposure

The overall number of likes given by each user is a good proxy for their level of engagement with the Facebook news pages. The lifetime of a user, meaning the period of time where the user started and stopped interacting with our set of pages, can be approximated by the time difference between the time-stamp of their latest and earliest liked post. These measures could provide important insights about news consumption patterns, specifically, the variety of news sources consumed over time.

We say that a user has consumed a page in a given time window, if the user has at least one positive interaction with that page in that period, that is, the user liked a post made by that page. We do not consider comments as a valid interaction for regular consumption because they have very diverse meanings and, dissimilar from the likes, they do not unambiguously represent positive feedback. Thus, we can measure the collection of pages consumed in a weekly, monthly and quarterly basis while taking into account the activity (total number of likes) and lifetime time difference of their first and last liked post) of the users of each country.

Fig. 3 shows the number of news sources a user interacts with considering their lifetime and for increasing levels of engagement for each country. For a comparative analysis, we standardized between 0 and 1 the number of pages present in each country, as well as the lifetime and engagement over the entire user set. The results were calculated considering the quarterly (right), monthly (middle) and weekly (left) rates.

Figure 3: Selective Exposure. Maximum number of unique news sources that users with increasing levels of standardized lifetime (top) or standardized activity (bottom) interact with weekly, monthly and quarterly for each country. The user’s lifetime corresponds to the normalized time difference between the time-stamp of their latest and earliest liked post. The user’s activity corresponds to the number of likes given in their lifetime.

Note that, for all countries, users usually interact with a small number of news outlets and that higher levels of activity and longer lifetime correspond to a smaller variety of news sources being consumed. We can also observe clear differences between the countries. When considering the users’ lifetime, France has clearly a more varied news consumption diet than the rest; and when considering the users’ activity users in Germany consume consistently the less diverse set of news sources. We can conclude that there is a natural tendency of the users to confine their activity to a limited set of pages, news consumption on Facebook is indeed dominated by selective exposure schmidt2017anatomy () and users from different countries display different rates for the decreasing variety of news outlets they consume.

3.3 Emerging Communities

User tendency to interact with few news sources might elicit page clusters. To test this hypothesis, we first characterize the emergent community structure of pages according to the users’ activity for each country with . We project the users’ page likes to derive the weighted graph (and ) in which nodes are pages and two pages are connected if a user likes (or comments on) both of them. The weight of a link on a projected graph is determined by the number of users the two pages have in common.

- Country Type ML SG
France Likes
Germany Likes
Italy Likes
Spain Likes
France Comments
Germany Comments
Italy Comments
Spain Comments
Table 6: Algorithm comparison. Comparison between the FastGreedy (FG) communities against the MultiLevel (ML) and SpinGlass (SG) communities for both likes and comments projections for every country.

We then apply the FastGreedy community detection algorithm to see if there are well-defined communities for each case. To validate the community partitioning, we then compare the membership of other community detection algorithms using the Rand method rand1971objective () and find high level of similarity for all four countries (see Tab. 6).

We also compared the communities of and against each other using different community detection algorithms and find, overall, low levels of similarity (see Tab. 7). This indicates that, for all four countries, the set pages users generally approve of (like), differ from the set of pages where they debate (comment).

Comparing - Country FG ML SG
- France
- Germany
- Italy
- Spain
Table 7: Likes and comments projections comparison. Comparison of the communities detected in and of each country with FastGreedy (FG), MultiLevel (ML) and SpinGlass (SG).

3.4 User Polarization

By examining the activity of users across the various clusters and measuring how they span across news outlets, we find that most users remain confined within specific groups of pages. To understand the relationship between page groupings and user behavior, we measure the polarization of users with respect to the communities found for each country where .

For a user with likes with such that each belongs to the community (, where equals the number of communities). The probability that the user belongs to the -th community will then be . We can define the localization order parameter as:

(1)

Thus, in the case in which the user only has likes in one community, . If a user, on the other hand, interacts equally with all the communities () then ; hence, counts the communities. Since we are considering many users, each with their likes and their frequency , we can plot the probability distribution and the complementary cumulative distribution function of along the user set of each country . This would allow for a fair comparison of the polarization of the users between countries.

Figure 4: Probability Density Function of for each country. The dotted vertical line indicates the median value.
Figure 5: Complementary Cumulative Distribution Function of for each country.

For each country, Fig. 5 shows the Complementary Cumulative Distribution Function of the localization , and Fig. 4 shows the Probability Density Function. Both figures consider only users with at least 10 likes.

As we can see in Fig. 4, the densities are well behaved, that is, present a single peak around 1. By looking at the CCDF of each country, we can rank the four countries from the one with least polarized users to the one with the most: Spain ), Germany (), France () and Italy ().

3.5 The Model.

In this section we provide a simple model of users’ preferential attachment to specific sources that considers the users’ trust on the media as a parameter and reproduces the observed community structure.

The entities of our model are pages and users . Each page is characterized by a set of opinions (an editorial line) modelled as a real number that ranges . We assume that the values are uniformly distributed. Each user has an initial opinion that is modelled as a real number , which ranges between and it is uniformly distributed. Each users also has a measure of trust in the media modelled by the real number , which ranges between . User’s trust will follow a truncated normal distribution.

We suppose and to be homogeneous such that the quantity is the distance between the opinion of user and the editorial line of page . We mimic confirmation bias by assuming that if user interacts with a page and the opinion distance is less than a given tolerance parameter , the preference of user will converge toward the editorial line of page according to the modified BCM deffuant2000mixing () equation:

(2)

To mimic user activity we give each user an activity coefficient that represents the number of pages a user can visit. Thus, the final opinion of a user will average the editorial lines of the pages the user likes. If is the set of pages that matches the preferences of user , then the average opinion will be:

To mimic the long tail distribution of our data we set the activity distribution to be power law distributed with exponent .

We use numerical simulation to study our model. A user randomly selects a subset of with which to interact. The user likes a page only when . When this occurs, the feedback mechanism reinforces the user’s page preference using the trust parameter to control the extent of the feedback. Thus the final opinion of a user will be the average of the editorial lines of the pages the user likes.

Figure 6: Analysis of the synthetic pages-to-pages graph . It shows the number of communities as a function of the mean user trust.

When a user’s opinion converges, we build in the bipartite graph where the set of edges are the couplings with which user likes page . Hence, represents users interacting with their favorite pages, and from we can build the projected graph that links the pages according their common users.

Figure 6 shows an analysis of as a function of the mean values used for the truncated normal distribution that models the trust , with different standard deviations and tolerance. Each point of the simulation is averaged over 100 iterations.

We can see that increasing the tolerance leads to a reduction of the number of communities, that is, agreement is reached faster and polarization takes place. Very low and very high values of user trust also display similar behavior. Absolute trust or no trust in the media leads to fast polarization, either the user will trust what they read fully and change their opinion accordingly, or they won’t.

The simulation displays an interesting behavior at where the number of communities formed by the users’ consumption habits seem to peak. This indicates that some skepticism might actually factor against polarization. Users’ who distrust the news they interact with, even when their opinions were similar, are more reluctant to further change their own beliefs. Perhaps a solution for the issue of false and misleading narratives could be found by fostering critical readers.

4 Discussion

In this paper we use quantitative analysis to understand and compare the news consumption patterns of four European countries: France, Germany, Italy and Spain. We show that while there are similarities in the consumption behaviours between the four countries, the posting and consumption behavior is not universal.

The results also show that all users, regardless of country, display selective exposure, that is, the more active a user is on Facebook the less variety of news sources they tend to consume. This behavior is seen in all four countries, with different rates of selective exposure for each case. News consumption on Facebook is dominated by selective exposure.

Additionally, we studied the cluster of news pages that emerge from the user’s activity, and found that users, regardless of their nationality, are polarized. We then measure the polarization of the users of each country, and ranked them accordingly, finding that Italy presents the most polarized users, followed by France, Germany and finally Spain. Further studies might gain insights into the reasons behind the slight variations in consumption habits.

Finally, we introduce a variation on the Bonded Confidence Model (deffuant2000mixing, ) that mimics the users’ behavior of selective exposure taking into account user trust. The simulation seems to indicate that users’ who have some distrust of the news they interact with, even when the narrative presented conforms to their beliefs, are more reluctant to further change their own beliefs. Thus, a tentative solution to mitigate user polarization might be found by fostering critical readers.

References

References

  • (1) A. L. Schmidt, F. Zollo, M. Del Vicario, A. Bessi, A. Scala, G. Caldarelli, H. E. Stanley, W. Quattrociocchi, Anatomy of news consumption on facebook, Proceedings of the National Academy of Sciences (2017) 201617052.
  • (2) W. Quattrociocchi, A. Scala, C. R. Sunstein, Echo chambers on facebook, Available at SSRN.
  • (3) M. Del Vicario, A. Bessi, F. Zollo, F. Petroni, A. Scala, G. Caldarelli, H. E. Stanley, W. Quattrociocchi, The spreading of misinformation online, Proceedings of the National Academy of Sciences 113 (3) (2016) 554–559.
  • (4) M. Del Vicario, F. Zollo, G. Caldarelli, A. Scala, W. Quattrociocchi, Mapping social dynamics on facebook: The brexit debate, Social Networks 50 (2017) 6–16.
  • (5) W. Quattrociocchi, Inside the echo chamber, Scientific American 316 (4) (2017) 60–63.
  • (6) F. Zollo, A. Bessi, M. Del Vicario, A. Scala, G. Caldarelli, L. Shekhtman, S. Havlin, W. Quattrociocchi, Debunking in a world of tribes, PloS one 12 (7) (2017) e0181821.
  • (7) C. Dewey, What was fake on the internet this week: Why this is the final column, The Washington Post.
    URL https://www.washingtonpost.com
  • (8) A. Nowak, J. Szamrej, B. Latané, From private attitude to public opinion: A dynamic theory of social impact., Psychological Review 97 (3) (1990) 362.
  • (9) S. Moscovici, The coming era of representations, in: Cognitive analysis of social behavior, Springer, 1982, pp. 115–150.
  • (10) B. Franks, A. Bangerter, M. Bauer, Conspiracy theories as quasi-religious mentality: an integrated account from cognitive science, social representations theory, and frame theory, Frontiers in psychology 4 (2013) 424.
  • (11) G. Deffuant, D. Neau, F. Amblard, G. Weisbuch, Mixing beliefs among interacting agents, Advances in Complex Systems 3 (01n04) (2000) 87–98.
  • (12) Facebook, Using the graph api, Website, last checked: 20.12.2017 (8 2013).
    URL https://developers.facebook.com/docs/graph-api/using-graph-api/
  • (13) N. Newman, D. A. Levy, R. K. Nielsen, Reuters institute digital news report 2015, Reuters Institute for the Study of Journalism.
  • (14) N. Newman, D. Levy, R. K. Nielsen, Reuters institute digital news report 2016, Reuters Institute for the Study of Journalism.
  • (15) N. Newman, R. Fletcher, A. Kalogeropoulos, D. A. Levy, R. K. Nielsen, Reuters institute digital news report 2017, Reuters Institute for the Study of Journalism.
  • (16) A. Clauset, M. E. Newman, C. Moore, Finding community structure in very large networks, Physical review E 70 (6) (2004) 066111.
  • (17) V. D. Blondel, J.-L. Guillaume, R. Lambiotte, E. Lefebvre, Fast unfolding of communities in large networks, Journal of statistical mechanics: theory and experiment 2008 (10) (2008) P10008.
  • (18) M. E. Newman, M. Girvan, Finding and evaluating community structure in networks, Physical review E 69 (2) (2004) 026113.
  • (19) J. Reichardt, S. Bornholdt, Statistical mechanics of community detection, Physical Review E 74 (1) (2006) 016110.
  • (20) W. M. Rand, Objective criteria for the evaluation of clustering methods, Journal of the American Statistical association 66 (336) (1971) 846–850.

Supporting Information

In this section we provide the list of all the downloaded pages. Table 8 contains the 225 news pages that form the dataset. Pages are identified by their name, website and Facebook ID, followed by the country code of their corresponding country. The countries are indicated with their ISO Alpha-2 international code (FR: France, DE: Germany, IT: Italy, ES: Spain).

plus -1fill

Table 8: List of pages of each country in the dataset.
Name and Website Facebook ID Community
1 ARD - ard.de 48219766388 DE
2 Augsburger Allgemeine Zeitung -
augsburger-allgemeine.de 121104385783 DE
3 Badische Zeitung - badische-zeitung.de 177670301122 DE
4 Berliner Morgenpost - morgenpost.de 46239931235 DE
5 Berliner Zeitung - berliner-zeitung.de 137267732953826 DE
6 Bild - bild.de 25604775729 DE
7 B.Z. - bz-berlin.de 57187632436 DE
8 Das Erste - daserste.de 176772398231 DE
9 Der Spiegel - spiegel.de 38246844868 DE
10 Der Tagesspiegel - tagesspiegel.de 59381221492 DE
11 Der Westen - derwesten.de 243001859426137 DE
12 Die Tageszeitung - taz.de 171844246207985 DE
13 Die Welt - welt.de 97515118114 DE
14 Die Zeit - zeit.de 37816894428 DE
15 Express - express.de 172718036608 DE
16 Focus - focus.de 37124189409 DE
17 Frankfurter Allgemeine Zeitung - faz.net 346392590975 DE
18 Frankfurter Rundschau - fr.de 134100583282150 DE
19 Freie Presse - freiepresse.de 375109771472 DE
20 Freitag - freitag.de 313744767921 DE
21 GMX - gmx.net 187741777922914 DE
22 Hamburger Abendblatt - abendblatt.de 121580125458 DE
23 Hamburger Morgenpost - mopo.de 196072707519 DE
24 Handelsblatt - handelsblatt.com 104709558232 DE
25 Hannoversche Allgemeine Zeitung - haz.de 198530121257 DE
26 Huffington Post DE - huffingtonpost.de 366193510165011 DE
27 Junge Freiheit - jungefreiheit.de 13479664941 DE
28 Kölner Stadt-Anzeiger - ksta.de 141063022950 DE
29 Leipziger Volkszeitung - lvz.de 114360055263804 DE
30 Mitteldeutsche Zeitung - mz-web.de 141558262607 DE
31 n-tv online - n-tv.de 126049165307 DE
32 Ostsee-Zeitung - ostsee-zeitung.de 374927701107 DE
33 ProSieben Newstime - prosieben.de/tv/newstime 64694257920 DE
34 Rheinische Post - rp-online.de 50327854366 DE
35 RTL aktuell - rtluell.de 119845424729050 DE
36 SAT1 Nachrichten - sat1.de/news 171663852895480 DE
37 Schleswig-Holsteinischer Zeitungsverlag - shz.de 248528847673 DE
38 Stern - stern.de 78766664651 DE
39 Stuttgarter Nachrichten - stuttgarter-nachrichten.de 144537361776 DE
40 Stuttgarter Zeitung - stuttgarter-zeitung.de 129349103260 DE
41 Süddeutsche Zeitung - sueddeutsche.de 215982125159841 DE
42 tagesschau - tagesschau.de 193081554406 DE
43 t-online - t-online.de 24897707939 DE
44 WAZ - waz.de 117194401183 DE
45 WEB.DE - web.de 56488242934 DE
46 Wirtschafts Woche - wiwo.de 93810620818 DE
47 Yahoo News DE - de.nachrichten.yahoo.com 166721106679241 DE
48 ZDF - zdf.de 154149027994068 DE
49 ZDF heute - heute.de 112784955679 DE
50 20 MINUTOS - 20minutos.es 38352573027 ES
51 ABC - abc.es 7377874895 ES
52 Antena 3 - antena3.com 55353596297 ES
53 Cadena Ser - cadenaser.com 15658775846 ES
54 Canarias 7 - canarias7.es 85160277321 ES
55 Cinco Días - cincodias.elpais.com 36280712574 ES
56 COPE - cope.es 15829535820 ES
57 Cuatro news - cuatro.com/noticias 96876562265 ES
58 Diario de Cádiz - diariodecadiz.es 128335533904779 ES
59 Diario de Ibiza - diariodeibiza.es 255177630236 ES
60 Diario de Mallorca - diariodemallorca.es 155352736257 ES
61 Diario de Navarra - diariodenavarra.es 103384039711468 ES
62 El Comercio - elcomercio.es 64673887657 ES
63 El Confidencial - elconfidencial.com 63830851925 ES
64 El Confidencial Digital - elconfidencialdigital.com 202726949863885 ES
65 El Correo - elcorreo.com 280982578099 ES
66 El Correo Gallego - elcorreogallego.es 152802838075123 ES
67 El Día - eldia.es 165210860204301 ES
68 ElDiario.es - eldiario.es 417471918268686 ES
69 El Diario Montañés - eldiariomontanes.es 109434489075314 ES
70 El Diario Vasco - diariovasco.com 91085818678 ES
71 El Economista - eleconomista.es 56760767000 ES
72 El Español - elespanol.com 693292367452833 ES
73 El Mundo - elmundo.es 10407631866 ES
74 El Norte de Castilla - elnortedecastilla.es 98474974005 ES
75 El País - elpais.com 8585811569 ES
76 El Periódico - elperiodico.com 93177351543 ES
77 Expansión - expansion.com 93983931918 ES
78 Faro de Vigo - farodevigo.es 123746764304270 ES
79 Heraldo de Aragón - heraldo.es 130012437016272 ES
80 Hoy - hoy.es 85593393832 ES
81 Ideal - ideal.es 64258697112 ES
82 Información - diarioinformacion.com 410523955526 ES
83 La Gaceta de Salamanca - lagacetadesalamanca.es 319669591452311 ES
84 La Nueva España - lne.es 51837272861 ES
85 La Opinión de Málaga - laopiniondemalaga.es 80999977105 ES
86 La Opinión de Murcia - laopiniondemurcia.es 106647502704110 ES
87 La Opinión de Tenerife - laopinion.es 112238345503995 ES
88 La Provincia - laprovincia.es 124641092828 ES
89 La Razón - larazon.es 113080018770027 ES
90 La Sexta - lasexta.com 39172614918 ES
91 Las Provincias - lasprovincias.es 20810574989 ES
92 La Vanguardia - lavanguardia.com 156552584408339 ES
93 La Verdad - laverdad.es 120857625399 ES
94 La Voz de Asturias - lavozdeasturias.es 101351926940208 ES
95 La Voz De Galicia - lavozdegalicia.es 350393845757 ES
96 Levante-EMV - levante-emv.com 106329485190 ES
97 Libertad Digital - libertaddigital.com 141423087721 ES
98 MSN España - msn.com/es-es 35966491049 ES
99 Onda Cero - ondacero.es 99040469027 ES
100 Público - publico.es 75084861845 ES
101 QUE! - que.es 97090259641 ES
102 RTVE - rtve.es 133623265400 ES
103 Sur - diariosur.es 52107727250 ES
104 Telecinco - telecinco.es 50353113909 ES
105 Última Hora - ultimahora.es 114680095225282 ES
106 Yahoo News ES - es.noticias.yahoo.com 284428852938 ES
107 20 Minutes - 20minutes.fr 51555073310 FR
108 Agence France-Presse - afp.com/fr 114100038626559 FR
109 BFMTV - bfmtv.com 43896752783 FR
110 Canal+ - canalplus.fr 144056732332683 FR
111 Challenges - challenges.fr 79566127213 FR
112 Charente Libre - charentelibre.fr 144375072241306 FR
113 Charlie Hebdo - charliehebdo.fr 106626879360459 FR
114 CNES Matin - cnewsmatin.fr 181111805243991 FR
115 CNEWS - cnews.fr 76952916976 FR
116 Corse Matin - corsematin.com 107249929306302 FR
117 Courrier international - courrierinternational.com 142114104887 FR
118 Dernieres Nouvelles d’Alsace - dna.fr 19004867327 FR
119 FranceInfo - francetvinfo.fr 135112586936434 FR
120 France Soir - francesoir.fr 53638966652 FR
121 France Télévisions - francetelevisions.fr 179086202130933 FR
122 Huffington Post FR - huffingtonpost.fr 284129444969978 FR
123 La Croix - la-croix.com 108828257010 FR
124 La Dépêche du Midi - ladepeche.fr 271219815470 FR
125 L’Alsace - Le Pays - lalsace.fr 181480351879611 FR
126 La Montagne - lamontagne.fr 146949065315655 FR
127 La Nouvelle République du Centre Ouest -
lanouvellerepublique.fr 87693933163 FR
128 La Provence - laprovence.com 119213845538 FR
129 La République des Pyrennées -
larepubliquedespyrenees.fr 148446219817 FR
130 La République du Centre - larep.fr 211082695569481 FR
131 La Tribune - latribune.fr 18950434380 FR
132 La Voix du Nord - lavoixdunord.fr 76635774021 FR
133 Le Bien Public - bienpublic.com 106094599409 FR
134 Le Courrier Picard - courrier-picard.fr 58080584133 FR
135 Le Dauphiné Libéré - ledauphine.com 122601757780987 FR
136 Le Figaro - lefigaro.fr 61261101338 FR
137 Le Journal du Dimanche - lejdd.fr 246577183385 FR
138 Le Monde - lemonde.fr 14892757589 FR
139 Le Monde Diplomatique - monde-diplomatique.fr 34398236687 FR
140 Le Nouvel Observateur - tempsreel.nouvelobs.com 198508090036 FR
141 Le Parisien - leparisien.fr 36550584062 FR
142 Le Point - lepoint.fr 49173930702 FR
143 Le Populaire du Centre - lepopulaire.fr 240500052515 FR
144 Le Progrès - leprogres.fr 104985642868265 FR
145 Le Républicain Lorrain - republicain-lorrain.fr 142638581774 FR
146 Les Échos - lesechos.fr 123440511000645 FR
147 L’Est Républicain - estrepublicain.fr 190366851765 FR
148 Le Télégramme - letelegramme.fr 97539957978 FR
149 L’Express - lexpress.fr 9359316996 FR
150 L’Humanité - humanite.fr 254585183694 FR
151 Libération - liberation.fr 147126052393 FR
152 L’Indépendant - lindependant.fr 52697519148 FR
153 L’internaute - linternaute.com 156569814356922 FR
154 L’Opinion - lopinion.fr 445890365491209 FR
155 L’Union - lunion.fr 100163350071823 FR
156 Marianne - marianne.net 369717525444 FR
157 Mediapart - mediapart.fr 116070051527 FR
158 Metro France - lci.fr 411124728976705 FR
159 Midi Libre - midilibre.fr 183518182558 FR
160 MSN France - msn.com/g00/fr-fr 136932803018290 FR
161 Nice-Matin - nicematin.com 388223307574 FR
162 Nord-Littoral - nordlittoral.fr 344969675415 FR
163 Ouest France - ouest-france.fr 270122530294 FR
164 Paris Match - parismatch.com 117714667328 FR
165 Paris Normandie - paris-normandie.fr 195238257180091 FR
166 Révolution Permanente - revolutionpermanente.fr 732277203520737 FR
167 Sud Oest - sudouest.fr 58305334711 FR
168 Télérama - telerama.fr 109520835773096 FR
169 TF1 news - tf1.fr/news 34610502574 FR
170 Var Matin - varmatin.com 365009223614 FR
171 Yahoo News FR - fr.news.yahoo.com 138207559575213 FR
172 Alto Adige - altoadige.gelocal.it 447795960541 IT
173 Ansa - ansa.it 158259371219 IT
174 Avvenire - avvenire.it 128533807252295 IT
175 Corriere Adriatico - corriereadriatico.it 431943793507773 IT
176 Corriere della Sera - corriere.it 284515247529 IT
177 Corriere del Mezzogiorno -
corrieredelmezzogiorno.corriere.it 84805991975 IT
178 Gazzetta di Modena - gazzettadimodena.gelocal.it 131613613524326 IT
179 Gazzetta di Reggio - gazzettadireggio.gelocal.it 102328739818445 IT
180 Giornale di Brescia - giornaledibrescia.it 352193836938 IT
181 Giornale di Sicilia - gds.it 211307618890745 IT
182 Huffington Post IT - huffingtonpost.it 276376685795308 IT
183 Il Blog di Beppe Grillo - beppegrillo.it 56369076544 IT
184 Il Centro - ilcentro.gelocal.it 261504285205 IT
185 Il Fatto Quotidiano - ilfattoquotidiano.it 132707500076838 IT
186 Il Foglio - ilfoglio.it 61703722992 IT
187 Il Gazzettino - ilgazzettino.it 154142713068 IT
188 Il Giornale - ilgiornale.it 323950777458 IT
189 Il Giornale di Vicenza - ilgiornaledivicenza.it 154836331469 IT
190 Il Manifesto - ilmanifesto.info 61480282984 IT
191 Il Mattino - ilmattino.it 210639995470 IT
192 Il Mattino di Padova - mattinopadova.gelocal.it 189556995002 IT
193 Il Messaggero - ilmessaggero.it 124918220854917 IT
194 Il Messaggero Veneto - messaggeroveneto.gelocal.it 195905383236 IT
195 Il Piccolo - ilpiccolo.gelocal.it 341809745380 IT
196 Il Resto del Carlino - ilrestodelcarlino.it 200174860861 IT
197 Il Secolo XIX - ilsecoloxix.it 36493277214 IT
198 Il Sole 24 Ore - ilsole24ore.com 38812693516 IT
199 Il Tirreno - iltirreno.gelocal.it 75980429042 IT
200 LA7 - la7.it 252449503661 IT
201 L’Adige - ladige.it 134572506600855 IT
202 La Gazzetta del Mezzogiorno -
lagazzettadelmezzogiorno.it 184749620911 IT
203 La Gazzetta di Mantova - gazzettadimantova.gelocal.it 62769612287 IT
204 La Gazzetta di Parma - gazzettadiparma.it 309928567597 IT
205 La Nazione - lanazione.it 87812020989 IT
206 La Nuova di Venezia e Mestre - nuovavenezia.gelocal.it 338049475695 IT
207 La Nuova Sardegna - lanuovasardegna.gelocal.it 226626114877 IT
208 La Provincia Pavese - laprovinciapavese.gelocal.it 57687391957 IT
209 L’Arena - larena.it 108431819182401 IT
210 La Repubblica - repubblica.it 179618821150 IT
211 La Stampa - lastampa.it 63873785957 IT
212 La Tribuna di Treviso - tribunatreviso.gelocal.it 243933437208 IT
213 L’Eco di Bergamo - ecodibergamo.it 197197145813 IT
214 L’Espresso - espresso.repubblica.it 259865949240 IT
215 Libero Quotidiano - liberoquotidiano.it 188776981163133 IT
216 L’Unione Sarda - unionesarda.it 231465552656 IT
217 L’Unità - unita.tv 292449724097 IT
218 MSN Italia - msn.com/it-it 232690009759 IT
219 Nuovo Quotidiano di Puglia - quotidianodipuglia.it 119992291359480 IT
220 RAI News - rainews.it 124992707516031 IT
221 Rai.TV - raiplay.it 88988179171 IT
222 Sky TG24 - tg24.sky.it 215275341879427 IT
223 TgCom24 - tgcom24.mediaset.it 40337124609 IT
224 Trentino - trentinocorrierealpi.gelocal.it 82383189226 IT
225 Yahoo News IT - it.notizie.yahoo.com 81262596234 IT
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
""
The feedback must be of minimum 40 characters and the title a minimum of 5 characters
   
Add comment
Cancel
Loading ...
208747
This is a comment super asjknd jkasnjk adsnkj
Upvote
Downvote
""
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters
Submit
Cancel

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test
Test description