VICTIM VOICE

Team Members

Emmy Eklundh, Janna Joceli Omena, Liva Brice, Ariadna Matamoros, Francisco Freitas, Gustavo López, Andrea Fiorentini, Valentine Weydert, Katia Meggiorin, Katja Viarshynina, Stella Rieck, Nora Wohlfeil, Christoph Rauch, Lonneke van der Velden

Introduction

[ Back to top]

Questions

To what extent can Facebook pages indicate acts of racial and xenophobic violence?

General Methodology

In order to investigate the relevance of using Facebook to anticipate concrete acts of violence, the research has been divided into language matter. As the Winter School team is composed of fourteen members speaking different languages, eventually the entire research can cover four of the main languages spoken in Europe: English, German, Spanish, and Italian. More specifically, this research seeks to investigate the use of pro vs anti-violence words through the analysis of pro-extremist Facebook pages on four European countries: United Kingdom, Spain, Germany and Italy. For the most part, the methodology applied in this project is the same for all countries while the findings are expected to be different according to the country of interest; which will allow a language based cross-country comparison.

Approaches

In order to know the concrete ability of Facebook pages content analysis to foresee acts of violence, two types of content are analyzed and two approaches are explored. Each sub-team, working on a specific country in a specific language, is analyzing both/either pro and/or anti-violence Facebook pages.

- The first approach consists in retrieving the principal events that occurred in each country of interest over the past year (from January 2013 to January 2014). According to the language used for the research, information on events can be found in reports, online newspapers and any other online news aggregators. Thereafter, each group researches the relevant Facebook pages, that could contain any pro or anti-violence words related to those specific events. The purpose of this approach is to investigate posts, comments and likes on those pages before and after the actual event of interest occurred.

- The second approach consists in exploring the most popular pro and anti-violence Facebook pages in each country of interest in two categories: Pro-violence content and Non-Governmental Organizations, expected to support the ‘Victims voice’. (Mainly Italy focused on researching the NGO content as well as the extremist perspective, while Spain, UK and Germany focused on the right-wing pages). The point is to identify peaks in the number of likes of a page and see if they relate to specific events. Through both of those methods the research seeks to highlight the eventual correlation between the activity of online communities and offline acts of violence.

Process

All language groups searched for pages of interest on Facebook and acts of violence on Google using the translated keywords related to ‘violence’. E.g. [immigrant] in Facebook search bar and [racist attack in UK] in Google search bar. Nevertheless, the implementation of the methods previously announced had to shift from one country of interest to another. Indeed, each group in charge of a specific country used the language spoken in this country. Yet, Germany, Italy, Spain and United Kingdom do not have the same problems, concerns, issues or interests for the same type of acts of violence. Although each group took into consideration the number of likes, comments and replies to comments to judge the popularity of a Facebook page; each group adopted a specific method to assess the relevance of the selected pages as well as related events. With the keywords another method was implemented using the principle of snowballing (looking at the pages Facebook advises users to like when visiting a related page).

In order for the research to remain coherent in terms of national focuses, as well as to ensure the doability, we had to concentrate the research on Facebook pages, excluding groups from the analysis. Moreover, as certain groups retained during the first step of selection gathered more than 5 000 members, the Netvizz tool could not have precisely retrieved the information. If a group exceeds 5000 users, Netvizz randomly retrieves members which diminishes the accuracy of the analysis.

Tools

In order to investigate the correlation between actual acts of violence and Facebook page activity both a quantitative and a qualitative analysis needed to be carried out. For the purpose of accessing information from extremists Facebook pages Facebook users have to ‘like’ the page, thus all researchers have created a research Facebook account. Additionally, all groups used Netvizz in order to retrieve the data from the selected Facebook pages. Netvizz allows us to collect data for every post of the pages until the very first one as well as the comments on a post and likes per comment. This provides enough data for a quantitative data analysis taking into account the user engagement on the Facebook pages. The ‘engagement’ of a post (e.g. a status, photo or link published on the Facebook page) consists of the total number of likes, the number of comments, the number of likes per comment, as well as the number of times the post has been shared. This number has been used by all groups to evaluate the most popular posts on a page and then to manually explore its content. After making the quantitative data analysis, it is relevant to use the DMI comment search tool to analyze the sentiments that can be associated with those comments and their evolution over time. Each group entered the ‘comments’ files retrieved by Netvizz into the DMI comment search tool. Subsequently, each group queried the keywords selected into the ‘query’ bar of the tool and selected one file/page comments to generate a graph. This graph shows peaks that correspond with the number of times the specific keywords has been employed in comments over the period of interest. Overall, this analysis of comments allows researchers to investigate the evolution of the sentiment of hate in comments published in the selected Facebook pages. Thereafter, each group can correlate peaks in comments to specifics acts of violence that occurred in the country of interest.

[ Back to top]

Germany

Participants: Stella, Nora, Christoph

Introduction

With the austerity crisis as well as a rise in right-wing populism in several European countries, the climate in Germany is equally becoming more hostile over the recent years (Ward). Issues include racism towards Black, Turkish, Jewish, Muslim, and especially Sinti and Roma communities (European Commission against Racism and Intolerance). The ENAR shadow reports, providing an NGO perspective on the issue from 2012, declare that islamophobia has risen so highly that it is socially accepted. People no longer fear to voice it in public. Furthermore, an article by the German newspaper Die ZEIT declares that assaults on German asylum seeking homes have doubled since 2012 (Eine furchtbare Bilanz). Hence, our assumption in this research is that racist violence in Germany is also being voiced openly on social media platforms and that it could thus be detected and analyzed. In the following, in order to give the research a tighter focus we concentrate on racist pages mainly agitating against immigrant communities.

Specific Methodology - Germany

Selection of the German Facebook pages of racist nature

In order to select the German Facebook pages of racist nature, we first created a research account (with matching racist interests). We then queried some keywords such as [Asylanten raus] (asylum seekers out), [Anti-Muslim], [Roma und Sinti raus] (Roma and Sinti out) amongst many others and checked if the displayed results would rear very active pages. There were not many obvious findings at first, and we had to look for other pages by snowballing. Starting with the most influential political party in Germany, the German right-wing NPD, we continued by searching for other pages and networks. We specifically looked at the NPD ‘likes’ that forwarded us to similar pages. Also, we were looking at public Facebook user profiles and checking their groups or ‘like’ network. Since there were a wide range of results, we picked those that reflect the stringest right-wing extremism. Additionally, only the pages that were most active, ie. showing most engagement were finally picked.

Categories for page selection:

• racist nature

• radical in content (for instance racist jokes, comments clearly orientated towards the right, images and slogans indicating an apparent plea for fighting against the supposedly increasing number of immigrants)

• active within the past year and months

• big number of likes, at least 1000

Final selection of pages:

• Keine weiteren Asylantenheime in Deutschland

• NPD

• Freies Deutschland

• Für Familie, Volk und Heimat - Multikulti und Islamisierung stoppen

  Page	Short Description	Likes
  ‘Keine weiteren Asylantenheime in Deutschland’	Active since October 18 2013. A page that is against the establishment of more homes for people seeking asylum.	12.933 likes
  ‘NPD - die soziale Heimatpartei’	Active since January 26 2010. The page of German political right-wing party, that is seeking to keep Germany to Germans.	56.274 likes
  ‘Freies Deutschland’	Active since November 3 2012. A page that is for a “free“ Germany, positioned on the far right.	2.214 likes
  ‘Für Familie, Volk und Heimat–Multikulti und Islamisierung stoppen’	Active since October 7 2013. A page that is against Islam and a multicultural Germany. But instead wants to support German family, nation and homeland.	7.376 like

We did not choose to include any Facebook groups as a lot of groups are closed and the membership needs to be confirmed at first, which would have exceeded our research time frame. Furthermore, most obvious groups in the relevant right-wing milieu have only few members.

Selection of incidents:

Seeing that the existing German official statistics are lacking depths and for 2013 none have been published yet, further resources of incidents needed to be acquired manually. We queried for different news sites, relevant headlines and titles that are affiliated with right-wing violence and reported confrontations both locally and nationally. Furthermore, we looked for archives and different news sources. In order to have a focus while researching and collecting the acts of violence, we chose to include those that were xenophobic, specifically anti-immigrants and incidents in regard to asylum seeker homes (considering that these have increased). After having set up a chronological list with incidents, we were then further categorizing them as ‘insult’, ‘minor injuries’, ‘actual bodily harm’, ‘murder’ and subsequently rated these incidents in regard to their intensity of violence on a scale from 1-8, 1 being the lowest for insult in order to show how violent these incidents are. Information on events was retrieved from: Mut gegen rechte Gewalt, Opferperspektive, Netz gegen Nazis, Süddeutsche, Endstation Rechts, Main Post, Mupinfo, Soziale Netze gegen Nazis, Neues Deutschland, Infoportal, Oireszene.

Problems in retrieving data:

Problems that we encountered when amassing incidents to compare the engagement of the separate Facebook pages were:

• Statistical data is not available for the year 2013. Thus, news websites and incidents needed to be searched manually.

• No central point of data collection.

• German police forces do not accurately register acts of violence as being motivated by racism. ( Radke et. al.)

Problems when searching for extremist Facebook pages:

• German right-wing extremism is not obviously displayed on Facebook sites. The German right wing sites seemed to have less obvious terminology. Some pages we could only find after researching like connections in some depth.

• Some of the Facebook pages are being repeatedly shut down, presumably because of their radical and racist content. Hence there are some pages that have not a long history (less data) and their launch date goes only back quite recently.

• In order to find really active sites, one would have to spend more time researching the extremist scene and investigate their language and habits, in order to detect possible specific key and code words used between right-wing members.

  [ Back to top]

Results and Discussion

Quantitative data analysis of general Facebook activity data:

Below the figures 1, 2 and 3 show the graphs of the three Facebook pages “Keine weiteren Asylantenheime für Deutschland“, “NPD – die soziale Heimatpartei“ and “Für Familie, Volk und Heimat“. All three graphs show the engagement (likes, posts and comments) for the relevant page and the graph of the incidents that happened during that time (June-January, October-January). We only chose those three as they have the highest number of likes and highest traffic on the Facebook pages out of the selected four.

Figure 1 shows the engagement of the Facebook page “NPD - die soziale Heimatpartei“ in correlation to the number of incidents from October 2013 to January 2014.

The graph of the political right-wing party “NPD“ Facebook page shows a major engagement during the end of August and in September 2013. During this time, there was the election campaign running in Germany, which explains the constant peak at this time. Although there was a heated debate on “Roma” in Duisburg and an ongoing discussion about a new asylum seeker home in Berlin-Hellersdorf, the content on the Facebook page of the NPD does not correlate with these incidents. Despite the fact that the NPD is a radical political party and thus the page content is always aggressive, the page itself is not picking up on specific incidents and is not as brutal as the other pages. Perhaps this is due to the fact that it is an official political party with representative media outreach. Interestingly though, the biggest peak in the graph relates to a photo titled “Asylflut stoppen“ (stop the asylum flood). The date of the post (October 28) does not correlate with a specific incident, but with the general trend in 2013 in Germany, that asylum seeker homes were attacked more often.

Figure 2 shows the engagement of the Facebook page “Keine weiteren Asylantenheime für Deutachland“ in correlation to the number of incidents from October 2013 to January 2014.

When looking at the graph it becomes obvious that there are much more peaks within the engagement on the Facebook page as there are incidents during that time. The page “Keine weiteren Asylantenheime für Deutschland“ has only been active since October 2013 and thus might still be used a lot. Moreover the debate on asylum seeker homes in Germany has been heated in the year 2013 and as this page is clearly about asylum seeker homes this can be considered to be an explanation for the constant engagement this site has. Although the incidents in October and January are attacks on asylum seeker homes and do correlate with peaks, it cannot be concluded that acts of violence really produce an increase in Facebook activity on this page.

Figure 3 shows the engagement of the Facebook page “Für Familie, Volk und Heimat – Multikulti und Islamisierung stoppen“ in correlation to the number of incidents from October 2013 to January 2014.

Similarly to the Facebook page “Keine weiteren Asylantenheime für Deutschland“, “Für Familie, Volk und Heimat“ shows a constant high engagement including two major peaks. These two peaks however are not incident related when looking at the activity on the page in detail. Furthermore, the first big peak seems to simply correlate with the launch of the Facebook site which was in the first week of October. Again it cannot be concluded that there is a distinct indication for an increase in activity that is related to specific acts of violence. Moreover, when reading the comments there is no particular mention of the incidents happening in October and January or a trend for future violent acts.

In summary, when looking at the actual comments on each page for the peak times of engagement, the message content rarely corresponded to actual incidents of violence on the streets. This proves the solely quantitative method, used in this section as a poor indicator of acts of violence.

DMI Comment Search tool:

As previously mentioned we have set our focus to violent incidents that were linked to asylum seeker homes, since the numbers of those have actually doubled compared to 2012. (Rechtsextreme Angriffe auf Flüchtlingsheime nahezu verdoppelt - Die Zeit). Apart from the overall quantitative graphs, representing the engagement of users on certain pages, we wanted to add a somewhat more qualitative aspect to the research by querying certain keywords in the collected comments of all German pages that were used (NPD, Freies Deutschland, Keine Asylantenheime, Familie und Heimat). The objective was to see if peaks in the use of sensitive keywords correlate with the peaks in the overall Facebook activity and subsequently with the incidents. Therefore we used the following specific keywords in the DMI Comment Search tool : [Roma], [Asylanten], [Heim], [Syrer] and [Hellersdorf]. [Roma] are the ‘Romani’ people, [Asylanten] means ‘asylum seekers’, [Syrer] stands for ‘Syrians’, [Heim] means home and is used in the context of ‘asylum seekers’ and [Hellersdorf] is a district of Berlin. In Hellersdorf, a new asylum seekers home was opened in August 2013 which led to lots of debates within the racist and radical circles, but was also covered broadly by the media. Furthermore, when looking at the results, it has to be taken into account that two of the pages (NPD, Freies Deutschland) have been active before 2013, whereas the other two pages have been active only since October 2013. As the data was generated from June 2013, respectively October to January 2014, the specific keywords were queried in the tool and represented by week and not by day as that would be too unclear to read.

Figure 4 Query for [Roma] - total of comments in peak time: 99

The graph in figure 4 shows the frequency of the word [Roma] within all comments on all four Facebook pages. The obvious peak with 99 comments is during the second week of October (7.10.-13.10.). From what we know there has not been a particular attack on Roma during this time. When looking at the graph there are two smaller peaks, one shortly before and shortly after the abrupt peak. The first smaller one starts during the end of August and is peaking in the middle of September. During this time there indeed has been a heated “Roma” debate and a demonstration in Duisburg at the End of August. The comments seem to suggest that this debate could have triggered a more frequent use of the term within discussions on the Facebook pages, but when actually looking into the comment section there was no indication that the frequent use refers back to the debate. Nonetheless, Roma may also be used just as an insult to foreigners, even if they are not Roma. The pages often lack this degree of exactitude.

Figure 5 Query for [Asylanten] - total of comments in peak time: 212

Figure 5 shows the peak of the queried word [Asylanten] which adds up to 212 and decreases again. However as mentioned before, two of the Facebook pages just joined Facebook in October 2013, one of which is called “Keine weiteren Aylantenheime in Deutschland” (No further asylum seeker homes in Germany). As the launch matches with the calendar week 42, dating to October 2013, when the frequency of the word Asylanten increases, this might be seen as an explanation for the increase considering it is specifically dealing with Asylanten. When looking into the comments, there are no mentions of the October-incidents on asylum seeker homes specifically. The debate on asylums is rather more general, and about how much they are disliked by the Facebook community of the relevant pages.

Figure 6 Query for [Heim] - total of comments in peak: 58

Clearly, when querying the word [Heim] as in home for asylum seekers, the curve resembles the one from the previous graph very much. There seems to be a peak in the use, when the two new pages become active towards the end of the year- both specifically agitate against immigrant homes. Furthermore, the quite frequent use of the term around August seems to point toward the nation-wide debate around this particular asylum seeker home being occupied in the Berlin district Hellersdorf. Evidently, in this case, word frequency and news correlate due to a national discussion covered by the media.

Figure 7 Query for [Hellersdorf] - total of comments in peak: 8

To solidify this assumption of figure 6, we queried the word [Hellersdorf], which is the actual name of this home of asylum seekers. Even though the term was used less frequently, it is specifically used around the time the asylum seeker home became an important debate in German news.

Figure 8 Query for [Syrer] - total of comments in peak time: 15

Lastly, the query for the word [Syrer] is included partially in order to see how informed the Facebook debate is when considering the immigrant groups migrating to the asylum homes. The asylum home in Hellersdorf shown above is especially open to refugees from Afghanistan and Syria. Clearly, the peak in the use of Syrians correlates with the peak in the word Hellersdorf. The number of mentions are relatively small however, and one can thus conclude that the right-wing discussion of the asylum home is of a racist nature in general rather than specifically targeting Syrians.

A conclusive note on the keyword query should add that, while the frequency in key terms does highlight the overall significance of certain terms in right-wing-circles, it does not account for the detailed and nuanced use of the terminology. On the pages we saw, words describing certain immigrant groups were often used interchangeably for all others.

Conclusion Germany

Apart from the problems due to the lack in data collection by any centralized entity in Germany, we also ran into difficulties when discovering smaller, but potentially very active right-wing Facebook groups, as they often used unusual names (“Ringelpietz mit Anfassen”). Clearly, a quantitative method alone does not prove any correlation between peaks in incidents and peaks in Facebook activity (even when these peaks do indeed correlate). In order to qualify the content further, we used a comment search analysis of keywords, which did show which terms were used most during certain periods in 2013. However, while this provides a slightly more nuanced view when combined with the engagement and incident data, it still can not serve as conclusive evidence to detect or foresee acts of xenophobic violence. Often words are used in several unexpected contexts.

[ Back to top]

Italy

Participants: Andrea, Valentine, Katia, Katja

Selection of Facebook Pages and Events

The sub-group started to select as many relevant pages as possible.We queried relevant keywords in Facebook search and made a selection according to the medium: pages with significant activity in each result page were selected. We used keywords like [immigrants in Italy], [fascist], [anti-immigrant]. We ended up with a large list of groups and communities. Eventually, we selected only four pages by picking the one with more likes and activity in its specific category: extreme-right, anti-immigrant, supporting immigrants and left-wing organization. Therefore, the most popular and/or active Facebook pages were identified as following:

I Giovani Fascisti Italiani, Fuori tutti gli immigrati dall'Italia, Sosteniamo gli immigrati di Rosarno, Osservatorio Sulla repressione.

The next step in the research was to identify key events. In order to verify whether monitoring violence on Facebook was possible at all, we decided to pick events that were already prominently featured on traditional media. For this reason, one could expect these selected events to generate noticeable reaction on social media and especially Facebook. Two kinds of events have been retained. The first event was related to immigrants issues, namely when a retired Italian has been attacked by an immigrant on 11th of May 2013. The second event was related to the issues between the right and left wing parties, namely when a neo-fascist gang attacked a left wing singer on the 5th of June 2013. These events caused many reactions in media. Then, it seems interesting to investigate whether they also caused peaks in activity on our pre-selected Facebook pages: pro-immigrant and anti-immigrant, left and right wing organizations.

[ Back to top]

Findings

PAGE ANALYSIS

---

For the page analysis, Netvizz was used in order to retrieve information on posts, likes and comments. We made graphs on Excel out of the data retrieved by Netvizz, taking into consideration a period of approximatively two weeks before and two weeks after the incidents. We created graphs that visualize the engagement of four pages from the 1st of May 2013 to the 20 of June 2013. In this way, we were able to amply cover the period during which our pre-selected events took place: the 11th of May 2013, when an Italian has been attacked by an immigrant, and the 5th of June 2013, when a neo-fascist gang attacked a left wing singer.

• I Giovani Fascisti Italiani

With 68 261 likes and 1700 posts published in 2013, I Giovani Fascisti Italiani is the most active extreme right wing political Facebook page we could find. Over the two months we took into consideration, five peaks of engagement have been observed: on May the 12, 22 and 25 as well as on June the 3rd and the 19th.

We can actually see an increase in the engagement of the Facebook page that corresponds to the date when the Italian has been attacked by an immigrant. Unfortunately, going back to the Facebook page we can notify that the posts of the 12 of May do not relate to the event but refer to the desire for fascists to return to the Musolini’s “Rupubblica Sociale de Italiana”. In the same logic, peaks observed on the 22 and 25 of May 2013 as well as on the 19th of June do not refer to concrete acts of violence.

The peak observed between the 3rd and 6th of June 2013 is actually due to an important number of posts published during this period but those actual posts do not individually gather so much popularity. Nevertheless, while the 7th does not appear to be a key day according to the graph, we can observe that only one post gather the 6 760 ‘points’ of engagement. This post supports Vladimir Putin’s homophobic propaganda. It has been liked by 1249 members and shared by 980 persons which, approximately, is ten times more than for any other post.

Overall, none of the acts of violence selected are related to the peaks observed over May and June 2013 on this page.

Figure 9: Engagement graph of the Facebook page I Giovani Fascisti Itiani, generated on January the 17th, 2014 out from the Netvizz data retrieved on 'posts by page and users'

• Fuori tutti gli immigrati dall'Italia

Fuori tutti gli immigrati dall’Italia is a highly xenophobic page dedicated to information regarding immigrants in Italy. With 5911 total page likes and 7157 posts in 2013 alone, this is the biggest and most active anti-immigration page we could find in Italian. It focuses on news related to immigrants, in particular violent incidents and politics.

Over May and June 2013, the analysis of page engagement highlighted 3 peaks: one occurring on 11-12 of May, the second on the 20th of May and the last one on the 30th of May. After manually checking the post published during these peaks, we found that the first one is indeed related to our preselected event: several posts featured pictures of the incident and used it to support their claim that Italy would do better without immigration. On the contrary, the posts published during the second peak showed no relation with our preselected event. Nonetheless, the posts were related to violence: they covered an aggression at the expenses of an old man carried out by immigrants. Finally, the 30th of May peak was not related to our preselected event. Even though it is not directly related to violence, it is indirectly connected to it. The posts cover a scandal regarding the process of hiring lifeguards for Jesolo beach: apparently, they were systematically rejecting applications made by female lifeguards because Islamic customers were not respecting their authority. The page claims that these arguments did sometimes result in aggressions and fights. In conclusion, peaks in engagement on this page seem to be an indicator of violent incidents.

Figure 10: Engagement graph of the Facebook page Fuori tutti gli immigrati dall'Italia, generated on January the 17th, 2014 out from the Natvizz data retrieved on 'posts by page and users'

• Sosteniamo gli immigrati di Rosarno

This Facebook page publishes content supporting immigrants living in Italy. With 15 569 likes, this page is considered as being popular even though only one person (on the 6th of January 2014) is talking about it and that the only 101 post have been published in 2013, the last post dating from the 20 of October 2013. Indeed, during the period of interest the page was more active than today. Nevertheless, neither the attack of the Italian guy by an immigrant, nor the aggression of the left-wing singer by neo-fascists do appear in this page. However, the content published remains relevant with regard to the title of the Facebook page, probably used as a news aggregator rather than a real active online community. This statement is actually confirmed by the observation of both peaks of engagement of the page that occur on the 6th and 18 of June 2013. On the 6th of June, an article has been published by the page about an African immigrant living in Italy who started to object to his ‘boss’ authority, considered as exploiting lower class of immigrant in the ‘ghetto’. This post has been liked 26 time and has been shared by 110 persons. This post actually appears to gather more reactions than any other in this page. The second peak, on the 18th of June 2013, relates to a generic post in favor of immigration and asking for peace. It is a picture where one can see two hands touching each other; one white and one black. This picture, saying “racism is not an opinion but a crime” has been liked 22 times and shared by 69 persons.

Overall, we can say that this page has been liked by many Facebook users probably because they thought it was for ‘a good cause’ but never really participated to its activity.

Figure 11: Engagement graph of the Facebook page Sosteniamo gli immigrati di Rosarno, generated on January the 17th, 2014 out from the Natvizz data retrieved on 'posts by page and users'

• Osservatorio Sulla repressione

Osservatorio sulla Repressione is a left-wing organization dedicated to the promotion of research and the production of reports regarding every kind of abuse. In particular, it focuses on activism, immigration and problems with the penitentiary system. Its Facebook page counts, at the moment, 20334 likes.

The analysis of the engagement on the page during May and June 2013 highlighted a peak in activity the 11th of May, which is exactly the date on which our preselected event took place. Unfortunately, a deeper analysis revealed that the peak was not related with the event nor with violence, hence suggesting that activity on this page is not a good indicator of violent incidents.

Figure 12: Engagement graph of the Facebook page Osservatorio Sulla repressione, generated on January the 17th, 2014 out from the Natvizz data retrieved on 'posts by page and users'

COMMENTS ANALYSIS

---

For the comments analysis we focus on four preselected Facebook pages by going through the comments manually on the data retrieved by the Netvizz tool. We used Excel files in order to look for the most used aggressive keywords. Therefore, we were checking the frequency of the keywords. Further, we were entering the probable keywords into the comment search DMI tool checking for peaks or interesting trends in every group separately. Referring mainly to two events, as it was mentioned above we tried to relate the appearing peaks to these events.

Thereby, we identified main keywords in relation to the events and the peaks. It seemed of interest to use the same list of keywords for the four Facebook pages selected. Indeed, we can expect to see different result for the same keyword as the angle of approach of the same topic must differ from the anti-immigrant page to the pro-immigrant page. Eventually, the keywords retained are: [vendetta] (revenge), [aggressione] (attack), [immigrat] (immigrant), [clandestino] (clandestine) and [attacco] (aggression).

• The page against immigrants Fuori tutti gli immigrati dall'Italia

The figure 13 presents an example of a peaks caused by the keyword [clandestino] which means “clandestine” in English. The word itself refers to a specific term as might be used to describe immigrants. Seemingly, peaks do not corresponded with our preselected events. Nevertheless, it is of interest to remind that this word has a pejorative connotation. However, others keywords more related to violence and action like [attaco], [vendetta] and [aggressione] are used only couple of times on really specific period. Their do not describe the sentiment of the users, the tone of the conversations and posts published on this page. On the other hand, the keyword [immigrat] appears to be central in the content published in this page which seems evident considering the entitle of the Facebook page.

Figure 13: Use of the keyword [clandestino] (clandestine) on the comments of the Facebook page Fuori tutti gli immigrati dall'Italia between the 1st of January 2013 to the 31 of December 2013. Data retrieved by the comment search DMI tool on the 17th of January 2014.

• Extreme right wing group I Giovani Fascisti Italiani.

On the figure below, we can clearly see that keyword [immigrat] is quite often employed and in a regular way. This can be justified as this page is revealed to be the most active one that we found (e.i. ‘page analysis’). But on the other hand, immigrants are not obviously supposed to be the central subject of conversation as this page appears to be fascist and not specifically anti-immigrant. However, this shows that [immigrant] remain a central subject of interest for the fascists Italian. One can expect posts/comments about immigrants to be anti-immigrants. But at the same time, the keywords related to violence like [vendetta], [aggressione] and [attaco] as well as [clandestino] are neither frequently nor quantitatively employed.

Figure 14: Use of the keyword [immigrat] (immigrant) on the comments of the Facebook page I Giovani Fascisti Italiani between the 1st of January 2013 to the 31 of December 2013. Data retrieved by the comment search DMI tool on the 17th of January 2014.

• Pro immigrants group _SOSTENIAMO GLI IMMIGRATI DI ROSARNO_

This page contained very few comments to identify the keywords. The group administrators and members use very polite and official language, no insults or offensive words used. Particularly with this Facebook page we could not identify any activity regarding our selected keywords. For all the preselected keywords [vendetta] (revenge), [aggressione] (attack), [immigrat] (immigrant), [clandestino] (clandestine) and [attacco] (aggression) the graph showed no activity, as shown in the Figure 15 with the [vendetta] keyword example. Those findings can be justified by the fact that this specific page is not active anymore (e.i. ‘page analysis’). Moreover, as highlighted in the above analysis, users to not interact but rather use this page as a new aggregator, sharing articles rather than comment on them.

Figure 15: Use of the keyword [vendetta] (revenge) on the comments of the Facebook page SOSTENIAMO GLI IMMIGRATI DI ROSARNO between the 1st of January 2013 to the 31 of December 2013. Data retrieved by the comment search DMI tool on the 17th of January 2014.

• The left wing group _OSSERVATORIO SULLA REPRESSIONE_

This page,expected to support immigrants cause in Italy, does not appear to really engage with the selected keywords. The peaks do not result from/in any pre-selected events related to act of violence. For example, the keywords [clandestino] and [attacco] do not result in any activity at all. The keyword [aggressione] caused three peaks, but those did not present much activity (only 1-2 comments contained the keyword) and did not relate to any of our preselected events. Despite that, there were peaks in the number of comments relating to different events, for instance, the keyword [immigrat] caused a peak right after the sea tragedy in Lampedusa, on the 4th of October 2013. This seems to be really low regarding to the cause supposed to be defended by this page.

Figure 16: Use of the keyword [immigrat] (immigrant) on the comments of the Facebook page OSSERVATORIO SULLA REPRESSIONE between the 1st of January 2013 to the 31 of December 2013. Data retrieved by the comment search DMI tool on the 17th of January 2014.

Overall, this analysis on comments does not reveal correlation between the comment of the posts of selected Facebook page and actual act of violence. However, it would be interesting to further investigate the comments of more Facebook pages. Perhaps the most popular pages are not the most likely to foresee violent incidences. In general, this specific analysis could be more insightful if more and more precise keywords could have been used.

[ Back to top]

Discussion

As a result of the analysis on the level of engagement on Facebook pages, we can say that it is difficult to make any distinction between posts relating concrete act of violence to posts relating other content immigrant related. For this reason, it seems like a evidence to focus on Facebook pages highly related to the topic of interest: pro-anti immigrants. During this research, we were able to identify only one Facebook page where all identified peaks related to that seems to concrete violent incidents: Fuori tutti gli immigrati dall’Italia. The characteristic of this page is that it is highly focused on crimes committed by immigrants, hence violent incidents are very likely to be featured and to provoke engagement in its followers. We can then conclude that pages highly focused on reporting crimes and violence may be a good indicator of the number of violent incidents taking place in Italy. Nonetheless, we cannot determine at the moment if sufficiently active pages dedicated to other specific kinds of crimes exist or not.

From the comments analysis part, it seems like the right wing group and anti-immigrant group present more activity and share hate expressions within their own community. The left wing groups and pro-immigrant community are not very active on Facebook. These groups use more polite, official and formal language, which makes it difficult to identify keywords to be tracked. Therefore, for further analysis it makes more sense to monitor the right wing and anti-immigrant groups. In regards to the right wing and the anti-immigrant communities we monitored, we saw that the selected keywords were successful in identifying major aggression events in the news. However, they show two major drawbacks: they also identify events not related to aggressive episodes (e.g. anniversary of Mussolini death for right wing community), they seems not to work well as predictors of the aggressions but rather as comments to the already happened violent episode. Consequently, the peaks provoked by keywords do not specifically relate to the preselected events. When a keyword provoked a peak we checked comments and their content but no meaningful result was found.

Overall, although the method of Facebook pages selection could be improved by investigating in further content of the pages and using others additional keywords, it seems difficult to say whether we can foresee acts of violence through the analysis of the Facebook pages and posts’ comments.

[ Back to top]

Spain

Participants: Ariadna, Gustavo, Francisco

Introduction

In Spain, four extreme right-wing political parties announced last December their will to run together for the European elections in May 2014 under the umbrella of the platform ‘España en Marxa’. To officially present their candidacy they will need the support of other extreme right parties, such as Alternativa Española (Madrid), España 2000 (Valencia), Movimiento Social Republicano (Madrid) and Plataforma per Catalunya (Catalonia), most of them currently attaining legislative seats in different local governments ( Info Raxen ). According to human rights organizations, the legally permitted extreme right parties in Spain contribute enormously in the creation of a fertile climate for xenophobic hate crimes ( Pro Igual ). These parties display anti-immigrant propaganda and intolerance speech which can be reflected in their Facebook pages.

[ Back to top]

Specific Methodology

Selection of the Spanish Facebook pages of racist nature

The identification of the Facebook pages was carried through a snowballing search method. Our starting point was the Facebook page of the xenophobic political party ‘Plataforma per Catalunya’, with 67 representatives in local governments in Catalonia, a northern region of Spain. This political party has an openly hate speech against foreigners, and this fact is reflected on its Facebook page. From that point, we looked for Facebook pages that were linked, via ‘like’, in order to analyse what kind of organizations were connected with ‘Plataforma per Catalunya’. We obtained some findings, such as two Facebook pages entitled ‘ Unitat contra el comunisme [unity against communism]’’ (407 likes) and ‘ El marxismo cultural debe perecer [marxist must die]’ (520 likes) that, despite their title, were platforms solely to spread propaganda against immigrants. Both were created recently, in 2013. Even though these findings, we decided to only focus on the extreme right Spanish political parties, that contribute to create an anti-immigrant feeling among society. The 8 right-wing Spanish political parties that have a clear hate speech are:

• Democracia Nacional (Madrid): 3.434 Likes
• La Falange (Madrid): 2.817 Likes
• Alianza Nacional (Madrid): 996 Likes
• Alternativa Española (Madrid): 1.396 Likes
• España 2000 (Valencia): 7.950 Likes
• Movimiento Social Republicano (Madrid): 1.852 Likes
• Plataforma per Catalunya (Catalunya): 1.978 Likes [ Back to top] [ Back to top]

  Selection of incidents:

Simultaneously we looked for recent hate crime events that had took place in the same period to analyse if we could make some connections regarding the spikes of activity in the right-wing Facebook pages and the dates when these crimes, aggressions or minor incidents occurred. In Spain, on contrary to other European countries like the UK, there is no official data about hate crimes. There is one NGO, Movimiento contra la Intolerancia (Movement against the intolerance), which recollects manually these data from the nineties to nowadays. The problem, though, is that the last report is from 2012. We faced this downside searching for hate crimes that appeared in the Spanish news during the last year, and especially in the last three months. Thus, we identified six hate incidents over that period occurred in different parts of Spain. The selection of the events does not attempt to be complete, since the process of searching was done manually and through a snowballing search process. We also used the Facebook pages of anti-hate crimes organizations to look for incidents occurred against foreigners. The main incidents that we found are: three attacks to mosques in Valencia (9 Jan.), attack to a couple of foreigners (9 Jan.), attacks towards immigrants that were demonstrating in Andalusia (6 Jan.), the detention of a young Cameroon without apparent reason (8 Jan.), an assault to the Complutense University in Madrid (17 Dec.), the police insulted with a racist remark a moroccan when they asked him to identify himself without apparent reason (29 Nov.). The information on events was retrieved from: Movimiento contra la Intolerancia, Plataforma ciutadana contra la Islamofobia, Levante, el mercantil valenciano; El Paiís, El Mundo, Público.es, El País. We graded these incidents regarding their intensity on a scale from 1 to 8 to further generate a graph:

2= insult

4= minor injury / building attack

6= actual bodily harm / police aggression

8= murder

[ Back to top]

Quantitative and qualitative analysis

For the quantitative analysis we used Netvizz to scrape the data from the Facebook pages of the 8 main right-wing political parties to see its activity over the last year (from 15 Jan. 2013 to 14 Jan. 2014). Then we selected three facebook pages out of our 8 examples to generate graphs and compare any possible correlation between the engagement activity and the hate events.

In reference to the qualitative approach, we investigated if the language used on the Facebook pages of the eight right-wing political parties could be conductive of violence. In this regard we used the Comments search DMI tool, in which we queried keywords related to “hot topics” or simply that implied a violent connotation. The tool gave us the number of comments in which the queries appeared, not the number of times each keyword is mentioned. The results were presented by number of comments per week. We queried “inmigrante” (immigrant) to see if this word had been most used over the past months. Immigrant, although it does not have a bad connotation by itself, in the Spanish right-wing Facebook pages is always used in a context where the foreign collective is criticized. We also queried “moro” (moroccan), a derogatory word to describe people from Morocco. Then we queried one violent verb: “matar” (kill).

[ Back to top]

Results and Discussion

Quantitative data analysis of general Facebook activity data:

Below the figures 17 and 18 show the graphs of two Facebook pages “Plataforma per Catalunya” and “España 2000”. All two graphs show the engagement (likes, posts and comments) for the relevant page and the graph of the incidents that happened during that time (from 15 Oct. 2013 to 14 Jan. 2014). We chose those 2 as both have the highest activity, and in the specific case of the page “España 2000” because some of its sympathizers are related with some of the hate incidents occurred over the last three months.

Figure 17 shows the engagement of the Facebook page “Plataforma per Catalunya“ in correlation to the number of incidents from 15 Oct. 2013 to 14 Jan. 2014

The graph of the political right-wing party “Plataforma per Catalunya” Facebook page shows a 3 major engagement spikes during the last week of October and the second week of December. The first peak in October corresponds to a post of a photo of an immigrant threatening the police of throwing his son from the balcony if he was arrested. This pictured received a lot of comments and users dehumanized the immigrant for this action. In December, the first spike corresponds to a video posted the 11th of that month in which immigrants say racist to a group of Catalan elderly people. And the second peak is a post of a photoof a family from Morocco to which the city council of Rubí (Barcelona) gave a protected house. All the engagement peaks are related to events related to foreigners that awake hate feelings and violent comments among the users. Nevertheless, these peaks do not correspond in time with the incidents occurred during that period.

Figure 18 shows the engagement of the Facebook page “España 2000“ in correlation to the number of incidents from 15 Oct. 2013 to 14 Jan. 2014

The graphic shows that this Facebook page does not have a lot of engagement despite two major peaks. The first one in October corresponds to a photo posted the 21th against ETA, the Basque terrorist group. The second one in November corresponds to a photoposted on the 14th in which judges, attorneys and politicians are being accused to be too indulgents with immigrants committing crimes. In this case one of the peaks coincides with content related to foreigners.

In this case, though, we wanted to analyse whereas the three attacks on mosques in Valencia in January had some relation with any activity change on the Facebook page of the extreme right political party “España 2000” in that period. According to the press, in these attacks there was a painting on the mosque wall where one could read “E 2000”, as if the perpetrators were sympathizers of this political party. Although we did not find explicit comments that made us think they were a prior indicator before the attacks, it is true that during Christmas of 2013 there were a lot of comments where people complained about to the bad situation that Spanish people are living in comparison to the situation of the immigrants. The comments found followed a similar argument: “Hay q dar a los españoles que los inmigrantes nos han hecho mucho daño q se vayan a su puto país [Spanish people have to receive help, the immigrants did a lot of harm to us, let’s send them to their fucking country]” (Posted on the 22 Dec.). Nevertheless, looking back over time, in November, we identify comments against mosques: “Menos hablar por aqui y mas actuar joder . A tirar cerdos en las mezquita por la noche en serio eso sirve[Less talk and more action. We should throw picks against mosques during the night. Seriously, this is useful] (13 Nov.); “Poco a poco están haciendo una ocupación silenciosa por toda España [Little by Little they [muslims] are silently occupying all Spain]”. (13 Nov.). This does not express causality but at least indicates the hate speech against immigrants on the “España 2000” Facebook page.

DMI Comment Search tool:

Apart from the quantitative graphs, we add a somewhat more qualitative approach to the research by querying certain keywords in the collected comments of all Spanish pages that we used (Plataforma per Catalunya, La Falange, Movimiento Social Republicano, Alianza Nacional, Alternativa Española, Democracia Nacional, España 2000, Nodo Patriótico). The objective was to see if peaks in the use of sensitive keywords correspond to the peaks in the overall Facebook activity and subsequently with the incidents.

Figure 19 Query for [inmigrante (immigrant)] - total of comments in peak time: 267_

The graph in figure 19 shows the frequency of the word [Inmigrante] within all comments on all eight Facebook pages. “Inmigrante (immigrant)”, although per se it does not have a bad connotation, on the Facebook pages of the extreme right political parties is used in a context where the foreign collective is criticized, especially the muslims. We observe a sharp peak with 267 comments during the first and second week of January. This period corresponds with the publication of the European Social Survey, which revealed that 50% of the Spanish people would restrict the entrance of non-european immigrants to the country. In January also appeared on the news that more than 7.000 immigrants obtained the Spanish nationality in 2013, and different crimescommitted by moroccan people were posted in these pages. Further, January was a month with a lot of news regarding the bad consequences of the economic crisis on the Spanish society. The combination of news alerting that Spanish families are experiencing difficult times with news blaming the immigrants to be either criminals or favored by governments regarding social benefits could have lead to more comments against foreigners in the extreme right political parties Facebook pages .

*Figure 20 Query for [moro (Moroccan)] - total of comments in peak time: 423

Figure 20 corresponds to the frequency of the word [moro (moroccan)] within all comments on all eight Facebook pages. The word “moro” appeared in 423 comments in January compared to the 9 comments that contained this word in the last week of December. During the past year, “moro” was never present in more than 70 comments in a particular week. The sharp rose of the use of this word could be due to the same reasons exposed above in Figure 19

Figure 21 Query for [matar (To kill)] - total of comments in peak time: 423

Finally, regarding the verb “matar (to kill)”, the curve resembles the one from the previous graph but less abrupt. We observe two peaks: one again in January (the keyword is present in 22 comments) and another one in November (the keyword is present in 18 comments). Through a manual analysis on the content of the comments we found out that this verb is often used in relation to immigrants. Some of the sentences identified are: “A matar putos moros [Let’s kill fucking Muslims]” (Feb. 2013); “Islam puta mierda, todos los musulmanes son racistas con los occidentales. Yo los mataría a todos pero tenemos que salir a la calle a la caza del inmigrante [Fucking Islam, all muslims are racist with the westerns. I would kill them all but for that we should go to the streets to hunt the immigrants]” (Feb. 2013). These comments are clearly explicit against immigrants and might conduct to violence.

To sum up, by querying the keywords in the DMI Comment Search tool we observed that January had a big peak regarding the use of sensitive keywords. This increase of the interaction of users corresponds with three incidents occurred during January: attacks to mosques in Valencia (9 Jan.), attacks to immigrants that were demonstrating in Andalusia (6 Jan.) and the detention of a young Cameroon citizen in Madrid without apparent reason (8 Jan.). Nevertheless it could not be proved a straightforward correlation between the incidents and the use of certain keywords among anti-immigrants organizations. On contrary, word frequency correlates more to information appeared on the news regarding immigrants and the consequent discussion around these thematics.

[ Back to top]

Discussion

Overall, a quantitative method approach it is not enough to deduce causality between the peaks on engagement on Facebook (number of likes, posts and comments) with the peaks in hate incidents. Thus, the lack of official data about hate crimes in Spain complicates the task of the researchers. Nevertheless, we found that the sympathisers of the 8 extreme right political react with strong language to Facebook content related to immigrants. This hate speech against foreigners, especially muslims, could be considered conductive to a violent climax taking place offline, albeit it is difficult to use social media to foresee future acts of violence. Thus, for a better understanding of the dynamics of the extreme right political parties each one should be studied separately through a more in depth qualitative analysis beyond the one we used with the comment search tool, in which often keywords were used in unexpected contexts. The tools helped us to identify some anomalies, but then we had to studied them manually, looking for the content of the comments on an specific day to understand the peak in the activity on Facebook. As in the German case, it was also difficult to find potentially active anti-immigrant facebook groups, as they often use names with no relation to foreigners, such as “unitat contra el comunisme [unity against communism]. Further research could be conducted to build an extended list of all these smaller pages and try to identify their connections and its possible radicalization.

[ Back to top]

United Kingdom

Participants: Emmy, Janna, Liva, Lonneke

This part outlines the specificities of the section dedicated to the UK. We will first describe the quantitative analysis conducted with the help of the Netvizz application, and subsequently analysed in Excel. After that we will concentrate on the qualitative findings, analysed in Semistrength.

Selection of Facebook accounts:
The first step in this analysis was to identify the main pages on Facebook that were to be analysed. In this process, we used the common keywords developed by the group, in order to narrow down the search for groups which would be relevant to our query. The words used were “(anti-)immigrants”, “(anti-)illegal”, “(anti-)muslim”, and “UK”. From the search results in Facebook, we chose the groups that had more than 500 members, and groups that had been active in the last year. In the search, we also came across a fair amount of groups which were obviously not anti-immigrant, such as anti-fascist groups, and we naturally disregarded these.

The list of the pages as follows:
Infields of Britain
Boycott Halal
Immigration Issues in the UK
UK Independence Party (UKIP)
British National Party
Stop ALL immigration into the UK
English Defence League
Britain First

Selection of events:
As noted in the reports provided to us by the HRW, we could see that hate crime is a tangible problem in the UK. However, much of the statistics provided no overview of the actual events, something which is crucial for our type of analysis. As such, we were forced to consult other sources, which also included other non-governmental organisations. One of these are Tell Mama, an organisation which monitors hate crime, and they had assembled a highly relevant map, which illustrated the “Anti-Muslim Incidents Post Woolwich. In particular, this map looks at mosques & Islamic institutions attacked post Woolwich”. Naturally, we found this extremely helpful, as this gave us a baseline for the acts of violence.

https://maps.google.co.uk/maps/ms?msa=0&msid=209838309277789756669.0004ddefdf328805d9d8d&hl=en&ie=UTF8&t=m&ll=52.683043,-2.614746&spn=6.396147,14.0625&z=6&source=embed. On the basis of this map we composed a spreadsheet containing all the titles and location of incidents from the map, and corrected the data (when available).

Our plan was thus to run these two different things together, as to investigate if there is any correlation between the two.

Data collection:

The data collection has been done by using the Netvizz application. All of the “actors” that we are using are pages, as the groups in these sort of communities are quite often closed. When having liked the page, we downloaded the posts by page and users, and then focussed on the stat file. We decided to only cover the time period for which we had any data on the acts of violence, which was April 30 - October 17 (2013).

Data analysis
When having opened the stat file in Excel, we decided to run the column engagement against the time column, in order to be able to identify certain peaks. These graphs, or , more specifically, their peaks, were in turn compared with the dates taken from the Tell Mama website.

The second part concerns, not only the frequencies of activity, but the frequencies of specific words: We queried the frequency of specific key words in the 'comment analysis' tool. In this way we profiled different Facebook pages.

Findings

The Facebook page "Infidels Britan" is more heathened up than "Stop Immigration", on the basis of the use of the term "burn". Presentation: https://prezi.com/tvp6iacb780t/victim-voice-uk/

Discussion

From the work we have performed this week arose several questions, problems, as well as possibilities for further research. First of all, we did have some constraints in retrieving as well as analysing the data. Some of the data files for the Facebook accounts were very large, and thus both hard to download, as well as to analyse. In our last day, we understood that in order to plot the right type of frequency graph, we needed to clean the data sets, which was quite time consuming. As such, we only have complete results for 4 Facebook pages. Naturally, this project could be complemented with a larger number of Facebook accounts/pages. In addition, the pages that were larger tended to be the bigger organisations, such as parties, and it would have been really interesting to match their frequencies with the offline events.

Now, let us turn to some of the caveats. One important issue is, of course, the selection of events, as well as the selection of groups. The selection of events is, naturally, biased, since it is assembled by a group with a clear, vested, interest in these matters. Nor is the list complete, since it only encompasses some hate crime events during this time period. However, as seems to be a problem for all people conducting this sort of analysis, there is always an issue with access to complete data, and the number of hate crimes reported might not reflect the real number. The selection of groups, on the other hand, is also subject to a few constraints. Whilst we have selected the most prominent anti-immigrant groups in the UK, some of the pages followed were previously unknown to us, and followed since they matched the keywords put into Facebook. This search could probably be improved by including more actors, as well as performing some kind of network analysis on the centrality of specific actors.

The other thing we must consider is the validity of the indicators. Here, we believe that it is important to point out that we by no means mean to infer causality based on our findings. The only thing that we can point to is a possible correlative relationship between the frequency of general activity of these pages, and the violent incidents taking place online. In some of our results, we can see a somewhat increased activity during June time. In truth, this is no surprising result. As pointed out by Tell Mama, the weeks following the Woolwich incident saw an increase in hate violence directed towards Muslims. While this all sounds quite intuitive, we would like to point to the possibilities of showing this empirically, with the methods described above. In other words, the project does not aim at proving that social media is a cause of violence, but that social media platforms, such as Facebook, could serve as indicators in outlet of hate opinions.

An aspect of further research that could be to incorporated into the analysis is the tool which was developed during the Winter School. Since we also have descriptions of the hate crimes, we could potentially search for keywords relating to the incidents, to see if there would be any connection between the frequency of certain words prior to the incident itself. Another aspect could be developing the use of Google maps. We found that the use of Google maps as a tool of visualising locale was quite convenient, and might be developed further.

[ Back to top]

General Conclusion

The aim of this research project was to explore the possibility of measuring the magnitude of violent incidents through the analysis of Facebook activity. Specifically, we focused on four European countries: UK, Germany, Italy and Spain. The study was carried out as an attempt to fulfill HRW’s (Human Rights Watch) needs: at the moment, Europe lacks a reliable and consistent mechanism to monitor violent incidents. Therefore, an automated online-based system would be the perfect substitute to allow the organization to estimate the current situation in specific countries and take action where it is needed the most.

Firstly, each group made a selection of Facebook pages that were considered relevant to the phenomenon. In this regard, each nation adopted a unique criterion. Secondly, the problem was approached with two main quantitative analysis methods: the analysis of the level of engagement and an automated text analysis of user-generated comments on selected Facebook pages. Finally, a qualitative approach was paired with the previous ones: potentially meaningful results obtained with automated analysis were manually cross-checked to determine if the results indeed correlated with violent incidents or not. In most cases, the results did not show any correlation with our data regarding incidents, but one specific Italian Facebook page engagement seems to be an indicator of violent incidents in its specific niche. Overall, the results are not particularly encouraging regarding the possibility of monitoring violent incidents through the study of Facebook. Especially when considering that the Italian page engagement spikes were caused by events that were already prominently featured on traditional media and that the victims were not members of minorities. This may suggest that, even with a deeper analysis, only news produced by influential but potentially biased sources have the ability to produce a significant activity on Facebook. Thus, the minorities that would need support by HRW or other NGOs the most seem to be more likely to be invisible to Facebook monitoring. Another important aspect seems to be finding pages that are highly focused on the phenomenon that one wants to observe, since more generic pages, such as the ones associated to political parties, consistently failed in producing good results. On a side note, it seems that Facebook analysis may be much more appropriate to study hate speech. This may particularly true for Spain and Italy, while Germany groups seem more prone to the use of slang that is hard to capture with automated processes.

For further research, an exploration of Twitter’s potential is an interesting direction. Especially if carried out with the aid of sophisticated software such as DiscoverText, since distinguishing relevant social media activity from noise seems to be a major issue. In the case of Facebook, the process of selection of pages seems to be the key. A study with a huge sample of pages and niche-based approach (i.e. check for crimes by category) could provide the precision and the numbers needed to successfully monitor violent incidents on social media.

[ Back to top]

Bibliography

Alker, Hayward R., and Kinhide Mushakoji. "Three Times for Tomorrow.” Conferencia Internacional a Construcao do Tempo e os Futuros Possiveis. Rio de Janeiro. 12 May 1999. Speech.

“Angriff auf Asylbewerberheim in Berlin.” Soziale Netze gegen Nazis: Forum. Spin.de. 21 Aug. 2013. Web. Accessed: 17 Jan. 2014. < https://www.spin.de/forum/171398/-/17ea >

“Archiv der Kategorie 'Veranstaltungen der extremen Rechten'” Oire Szene - Infos zu Rechtsrock und Grauzone. Web. Accessed: 17 Jan. 2014

< http://oireszene.blogsport.de/category/veranstaltungen-der-extremen-rechten>

Card, Brittany, Justine MacKinnon, and Patrick Meier. "#Westgate Tweets: A Detailed Study in Information Forensics." iRevolution.net 18 Nov. 2013. iRevolution.net.

Chan, Jason, Anindya Ghose, and Robert Seamans. "The Internet and Hate Crime: Offline Spillovers from Online Access." Social Science Research Network ­ (2013). Social Science Research Network.

“Chronik der Gewalt.” Mut Gegen Rechte Gewalt. in: Projekt Magazin stern.de. Web. Accessed: 17 Jan. 2014. < https://www.mut-gegen-rechte-gewalt.de/service/chronik >

“Die netten Rechten von nebenan. - Asyl-Streit in Berlin Marzahn-Hellersdors bleibt aggressiv.” Süddeutsche.de. 07 Aug. 2013. Web. Accessed: 17 Jan. 2014.

< http://www.sueddeutsche.de/politik/aggressiver-asylstreit-in-berlin-die-netten-rechten-von-nebenan-1.1740306 >

“Endstation Rechts.” Endstation-Rechts.de. Web. Accessed: 17 Jan. 2014. < http://st.endstation-rechts.de>

European Commission against Racism and Intolerance. “Ecri Report on Germany.” ECRI 19 (2009).

Gerstenfeld, Phyllis B., Diana R. Grant, and Chau­Pu Chiang. "Hate Online: A Content Analysis of Extremist Internet Sites." Analyses of Social Issues and Public Policy 3.1 (2003): 29­44.

Grizelski, J. “Böllerattacke aufs Asylantenheim.” MUPINFO.de. 12 Oct. 2013. Web. Accessed: 17 Jan 2014. < http://www.mupinfo.de/?p=26792>

“Güstrow: Angriffe auf Flüchtlingsheim nur wenige Stunden nach Eintreffen erster Flüchtlinge.” Infoportal für antifaschistische Kultur und Politik aus Mecklenburg-Vorpommern. 20 Jun. 2013. Web. Accessed: 17 Jan. 2014

< http://ino.blogsport.de/2013/06/20/guestrow-angriffe-auf-fluechtlingsheim-nur-wenige-stunden-nach-eintreffen-erster-fluechtlinge>

Heiligenthal, F. “Brand im Asylbewerberheim: Zwei Menschen verletzt.” MAIN-POST. 18 October 2013. Web. Accessed: 17 Jan 2014.

Hieronymus, Andreas et al. “Racism and Related Discriminatory Practices in Germany.” ENAR Shadow Report.

Howard, Philip N. and Muzzamil M. Hussain. Democracy’s Fourth Wave? Digital Media and the Arab Spring. New York: The Oxford University Press, 2013.

Huang, Yen­Pei, Tiong Goh, and Chern Li Liew. "Hunting Suicide Notes in Web 2.0 – Preliminary Findings." Ninth IEEE International Symposium on Multimedia 2007 ­ (2007). IEEE Xplore.

Human Right’s Watch. 2014. Human Right’s Watch. 17 January 2014. < http://www.hrw.org/ >.

"Informe Raxen - Servicio de Noticias de Movimiento Contra La Intolerancia" Web. Acessed: 14 Jan. 2014.

< http://www.informeraxen.es >

“Massive rassistische Mobilisierung: Der Jahresrückblick 2013 aus Brandenburg.” Netz-Gegen-Nazis.de. in “Opferperspektive e.V. Brandenburg”. 24 Dec. 2013. Web. Accessed: 17 Jan. 2014. < http://www.netz-gegen-nazis.de/artikel/massive-rassistische-mobilisierung-der-jahresrueckblick-2013-aus-brandenburg-9166 >

Meier, M. “Diese Asozialen haben kein Recht, hier zu sein.” neues deutschland. 26 Aug. 2013. Web. Accessed: 17 Jan. 2013

< http://www.neues-deutschland.de/artikel/831250.diese-asozialen-haben-kein-recht-hier-zu-sein.html >

Meier, Patrick. "Early Warning Systems and the Prevention of Violent Conflict." Peacebuilding in the Information Age: Sifting Hype from Reality. ICT4Peace, 2011. 12­16.

"Pro Igual - Centro de Investigaciones en Derechos Humanos" Web. Acessed 15 Jan. 2014

< http://proigual.org >

Radke, J., Staud, T., Jansen, F., Kleffner, H. “Eine furchtbare Bilanz.” ZEIT Online. 30 May 2012. Web. Accessed: 16 January 2014.

“Rechtsextreme Angriffe auf Flüchtlingsheime nahezu verdoppelt.” ZEIT Online. 23 Dec. 2013. Web. Accessed: 17 Jan. 2014.

< http://www.zeit.de/gesellschaft/zeitgeschehen/2013-12/rechtsextremismus-angriffe-asylheime >

Rechte Gewalt in Brandenburg.” Opferperspektive. Web. Accessed: 17 Jan. 2014.

< http://www.opferperspektive.de/event/events_by_criteria/1?year=2013>

“Unbekannter zündet Asylbewerberheim an.” Süddeutsche.de. 08 Jan. 2014. Web. Accessed: 17 Jan. 2014.

< http://www.sueddeutsche.de/bayern/germering-unbekannter-zuendet-asylbewerberheim-an-1.1858485 >

Starbird, Kate, Grace Muzny, and Leysia Palen. "Learning from the Crowd: Collaborative Filtering Techniques for Identifying On­the­Ground Twitterers during Mass Disruptions." Proceedings of the 9th International ISCRAM Conference ­ (2012). http://epic.cs.colorado.edu/.

Ward, B. “Europe’s Own Human Right Crisis.” Human Right’s Watch. 2011. Web. Accessed: 17 Jan. 2014 < http://www.hrw.org/world-report-2012/europe-s-own-human-rights-crisis >