Mapping War Atrocities across Platforms

Team Members

Facilitators: Stephanie de Smale, Alex Gekker

Members: Emanuela Blaiotta, Quynh Tu Hoang, Mick Jongeling, Janna Joceli Omena, Ginevra Terenghi, WenZhe Zu

Summary of Key Findings

In mapping the spheres and publics that remember Srebrenica, we can discern between dominant and marginal framings, or program and antiprogram. The dominant framing (program) present around the Srebrenica commemoration focuses on commemorating the Bosniak victims that died in the war. The marginal framings represent the Serb/Serbian perspective, found on YouTube (YT), Instagram (Insta), Facebook (FB), and Twitter (TW), and the Dutch perspective, found only on YT. It is important to note that the detection of marginal framing of Srebrenica required extra research efforts; meaning if one search for Srebrenica on social media platforms, one will certainly get the Bosniak victims perspective.

Across Insta, FB, and TW, we found that the Srebrenica commemoration is used by different publics to draw attention to their own victimhood. EN/BA language both focus on Bosniak victimhood, where SR language draws attention to the Serb/ian casualties that happened in the post-Yugoslav war, mainly between 1991-1995 and in 1999 during the Belgrade bombing.

There are differences in the visual framing of each platform. Where Instagram is more affect-based and focuses on solidarity and mourning, Twitter is used as a platform for calls to action, whether its in protest, peace marches or in general calling attention to local politics.

YouTube works differently than Twitter and Instagram. Where the latter is more user-led through the hashtag affordance, the YouTube viewer is more dependent on the recommendation algorithm what to watch next. The goal here is to more passively consume what is recommended to you. Here we found each country had their own specific spheres and agendas for Srebrenica, with a couple of minor bridging nodes connecting the two.

See our final presentation here: Mapping War Atrocities across Platforms

Link to Wiki: https://wiki.digitalmethods.net/Dmi/SummerSchool2018MappingWarAtrocities

1. Introduction

One of the blackest pages in the post-Yugoslav wars is the Srebrenica massacre. In July 1995 over 8000 Muslim men and boys were separated from their families by Serbian troops under the command of Ratko Mladić. Srebrenica was declared a “safe zone,” under the protection of the United Nations. Responsible for keeping safeguarding Srebrenica were the Dutchbats, who failed to protect its refugees during the massacre on July 11.

As illustrated in a study of Dutch, Serbian, Bosnian, Croatian, and Serbo-Croatian Wikipedia pages, different framings of Srebrenica circulate on the internet (Rogers & Sendijarevic 2012). One interesting finding in analysing the Wikipedia edits, is that these pages are edited especially around the Srebrenica memorial day on July 11. During this day, the genocide is remembered, and human remains identified that year are buried during the ceremony.

The Srebrenica memorial ceremony takes place on July 11 and is heavily mediatized on platforms such as YouTube, offering mourners the chance to witness the event from a distance. Around this day, other more political events, such as the Mars Mira (Annual Peace March), held July 7 - 10 July, are also organised to draw attention to the prosecution of war criminals and outstanding arrests.

On the one hand, digital commemoration offers long-distance mourning of the Bosnian diaspora spread out globally in Europe, USA, Canada, and Australia (Halilovic 2013). On the other hand, in parallel with everyday remembrance is the political side of commemorating Srebrenica, such as establishing legal accountability and a local memorial for the victims of the massacre. These efforts led to a series of resolutions commemorating Srebrenica, for instance in 2009 by the European parliament (Mehler 2017). The memorial day is a highly mediatised political event, visited by politicians across the globe.

The ‘recommended videos’ affordance of YouTube informs what the viewer of a video on Srebrenica may watch next. This affordance is particularly relevant when binge-watching. The question is whether or not these different framings of Srebrenica materialize in the YouTube sphere and whether or not they create “echo chambers.” And therefore, how the recommendation algorithm may consolidate, diffuse, or diffract different narratives of war atrocities.

2. Initial Data Sets

Instagram

A first dataset was used for an overall visualization consisting of the images scraped from #srebrenicamassacre #srebrenicagenocide #Сребреница (Srebrenica in Cyrillic alphabet).

A second dataset was used instead to create a co-hashtag and images visualization from the hashtags #Srebrenica and #Сребреница. We obtained #srebrenicamassacre #srebrenicagenocide images in Latin and #Сребреница images in Cyrillic. These images were analysed in Image Sorter and compared to each other and the images found on Twitter and Facebook.

Twitter

10.070 Tweets #Srebrenica written in Latin alphabet

166 Tweets #Srebrenica written in Cyrillic alphabet (Сребреница)

We obtained #Srebrenica images in Latin and #Сребреница #НаДанашњиДан images in Cyrillic. These images were analysed in Image Sorter and compared to each other and the images found on Twitter and Facebook.

Facebook

Query in Latin/Cyrillic

• “site:facebook.com Srebrenica”, in BA, NL, EN

  • 0 NLpages

  • 8 BA pages

  • 9 EN pages

• “site:facebook.com Сребреница”, in SR

  • 0 pages

We did a deep search for Serbian pages using a variety of other search terms In Cyrillic (genocide, bosnian war). In order to detect more Srebrenica-related pages, we relied on Netvizz module page like-network (depth 1) and analysed the network of the pages in BA, EN, and SR in Gephi. After analysing the clusters we took at least one page that represented each cluster and extracted all the images of each page by using Netvizz module Page Timeline Images. This resulted in two datasets: i) the dominant framing on Facebook constituted by four clusters: Women’s counter narrative (42 images), Bosniak Diaspora (241 images), Muslim Identity (10.669), and Bosnian Hero (1.107 images); ii) the marginal framing was also formed by four clusters: Proclaiming the truth about Serbia (946 images), Serbian Heroes (568 images), Serbian Nationalism (1144 images), and Solidarity (7649 images). These images were analysed in Image Sorter and compared to the images of other platforms. See below the page like networks both from dominant and marginal framing that led the data extraction process on Facebook pages:

Srebrenica dominant framing on Facebook [Page Like Network]

Srebrenica marginal framing on Facebook [Page Like Network]

YouTube

Four datasets of recommendation networks of Srebrenica on YouTube across four languages and regions using “srebrenica” as query (BS; BA; RS; SR; EN, US; NL, NL).

This resulted in four datasets per country, which we combined into one dataset in Gephi to see where the clusters overlap in their recommendations.

3. Research Questions

For this report, we set out to tease out how these platforms visually afford different types of commemoration. How do different groups emerge around the issue and claim a space to draw attention to their own issues, as well as how these dominant and marginal framings (program/antiprogram) collide and dissipate across platforms. The questions we answered were:

1. How do platforms afford different types of remembering?

2. How do they emerge and claim a space within the platforms?

3. What is the role of recommendation algorithms in structuring these experiences?

4. Methodology

The commemoration of Srebrenica is as much about politics as it is about remembrance. Therefore, our methodology relied on operationalising social movement theory on social media. Our methodology builds on Rogers (2018) critical analytics approach, where we used dominant voice, concern, positioning, and alignment as lenses to map the issue across platforms.

Dominant voice

Dominant voices looks at the most impactful voices within the issue network by analysing the ‘specific actors that give voice to the issue with greatest strength’ (ibid. 2018, 455). To operationalise this, we looked at dominant nodes in our network analysis across platforms (Co-hashtag-, page-like-, and recommendation network analysis) and by looking at the central clusters and nodes in the networks and classifying them.

Concern

Concern refers to the ways in which individuals or persons are present or absent within the space. Translating this to our research, we focused on different perspectives to the issue (Srebrenica) are present or absent. How are perspectives across countries on Srebrenica present or absent in these different spaces? By searching for the presence of the NL, EN, BA, and RS framings across platforms we found present and absent perspectives varying across platforms.

Positioning: positioning refers to the choice of words chosen in relation to an issue. For our research, we chose to map visual framings and textual relations on the different platforms. So what are the related words used in relation to Srebrenica? How is Srebrenica framed visually and textually? We used co-hashtag analysis (Insta, TW) recommendation network analysis (YT), page-like network analysis (FB) and image clustering via Image Sorter (Insta, TW, FB)

Alignment

Alignment focuses on group formation around the positionings. i.e. the groups that form around the different perspectives of the event. We analysed this by looking at cluster formations in Gephi (co-hashtag analysis, recommendation network, like network) that emerge that emerges around the commemoration of Srebrenica group emerges

Tools:

Facebook

• Google Search

• Netvizz (page-like network, timeline images)

• Image Sorter

• Gephi (Force Atlas2, modularity class 1.0)

Twitter:

• TCAT (Srebrenica in Latin and Cyrillic, scraped for two weeks from 02-07-2018 until 15-07-2018)

• Gephi (co-hashtag network)

• Image Sorter

Instagram:

• Python script Instaloader to scrape dataset and hashtags (https://github.com/instaloader/instaloader),

• Get Them All/ Downloader for Instagram Chrome Extension to scrape all images,

• Image Sorter,

• Gephi (co-hashtag network)

• Open Refine (to modify dataset to replace nodes with images)

  • Upload images in tekstedit program (used BBEdit) to change code

  • Export Gephi to change to the link to the images

YouTube

• DMI YouTube Scraper

• Gephi (recommendation network, Force Atlas2, modularity class 1.0)

• Illustrator

5. Findings

Findings on Twitter

The Twitter Data Set on which we worked on was collected via the digital methods tool Twitter Capture and Analysis Toolset (TCAT), until the 10th of July, the day before the commemoration day.

We, initially, extracted the images from the output of the TCAT (a .csv file).

We were expecting to find more active Dutch war veterans on Twitter. We did find a couple of references to the Dutchbats with a negative undertone.

A second part of the Twitter analysis was focused on a co-hashtag analysis of both the hashtags #srebrenica and #Сребреница, that were again scraped through the TCAT and then visualized on Gephi.

From what is possible to see, there are six main clusters, connected together with a main bigger node “Srebrenica”.

1. The main cluster (26, 27%) mainly consists of hashtags connected with "genocide", "massacre", "victims" in both in English and Turkish. It then goes on toward the words connected with "bosnia" (bih, war in bih, potocari) and the geographical Balkans realm. Another realm is connected with "remembrance"(never forget, don't forget Srebrenica, remember 8372, Srebrenica memorial week). A further important thing to point out is the presence of the Turkish term "Ölüm Yürüyüşü", the march of death, referring to those escaping from genocide and "mars mira", march of peace, that indicates the annual march conducted in memoriam of the Srebrenica deaths.

2. The second cluster (2,65%) is mainly connected with the central node via the hashtags "serbia", "sarajevo", that lead toward the war crimes realm. In addition, there’s a connection with a bicycle marathon that was lead in Germany in memoriam of Srebrenica.

3. 1,99% - Presents a number of UK cities that are connected with Srebrenica via the specific twitter accounts dedicated (Remembering Srebrenica UK) - Twitter presents a community in the genocide from the UK, probably due to a large community of Bosnian present in the UK.

4. 1,99% - It is mainly related to the realm of "courts" and "judgment"

5. 0,66% - Connected with Swedish 2018 elections, again probably because of the large number of bosnian that went there.

6. 0.66% - A small cluster of italian hashtags, connected with “11 july” and “genocide”

Thus, we can state that the marginal voice occupies a lower space in Twitter with respect to other platforms, but, instead, what is possible to find is a large support for the event coming from the Turkish community.

Findings on Facebook

We did not find any specific Dutch pages or Serbian pages on Facebook. Using the same method as with Twitter and Instagram images, we analysed images from 7 pages for dominant framing and 11 pages for marginal framing. For the dominant framing, we identified clusters that depicts the women’s counter narrative, for instance women mourning at the memorial sites, as well as Bosnians organizing memorial events all over the world, especially in the US. An emerging cluster is the military and Bosnian flag as representing nationalism or national pride.

The extracted images from the marginal voices show a more extensive result than from Twitter and Instagram. While the same image calling for attention to Serb victims presents in Facebook dataset, two framings identified are images of attacks on Serbian territory for example NATO bombing Yugoslavia in 1999 on the name of humanitarian intervention, and Serb militarians. The marginal voices on Facebook seem to emphasize nationalism and historical evidence.

Findings on Instagram

The first step was to create a general overview of the Instagram photos for two specific hashtag #srebrenicamassacre and #srebrenicagenocide. Unfortunately, it was not possible to find any hashtags with the same meaning in Serbian (cyrillic), thus, we decided to use the general #Сребреница (Srebrenica). It is important, in addition, to underline that the number of images scraped was largely different between the first two hashtags (three thousand) and the third one (two-hundred and twenty).

Images were scraped with the Google Chrome Extension “Get Them All” and Firefox extension DownThemAll, then, fed into the ImageSorter.

For the hashtags #srebrenicagenocide and #srebrenicamassacre the following clusters were identified:

Identified Bodies Images (with mourning women), Flower of Hope (the specific symbol of the genocide), 8372 (the number of bosnian deaths), Museum images (strictly related to the images present in the museum dedicated to the event, Gallery 11/07/95), Historical Pictures, Butterfly, Cups of Coffee, Fatima, Memorial Site, Murales, Quotations, Text Images, Pop Art. These clusters are especially related to the realms of remembrance, commemoration and empathy. This is done both never to forget the events of Srebrenica, but also to appeal to people’s emotions and feelings of solidarity. This plainly justifies the content of the images, that shifts from iconic (of those days, of the memorial site, of the identified bodies) to more symbolic ones (the number of deaths, cup of coffee, butterflies).

Considering the shift toward Cyrillic with the third hashtag #Сребреница (Serbian), the appearance of two new clusters differing from the previous ones in terms of content can be justified. These two clusters are respectively defined as: “Serbs Victims”, and “Celebrations for the Russian Veto”. The Serbs, or better, Bosnian-Serbs, demand and request the recognition of their 3267 victims. It is interesting how, a number is used again to emphasize a loss, in this case of the perpetrators of the crime, that want to rehabilitate their image and position in history via what we defined as an “tactic of appropriation ”. An important thing to state here is the absence of any visual evidence of Serb victims and Sarajevo genocide, in plain contrast withthe Srebrenica genocide.

The second cluster, instead, is referred to the celebrations as a consequence to the Russians vetoing the Srebrenica genocide resolution of the UN in 2015. Serbians, in this case, convey their nationalized brotherhood, for finally being recognized. Another symbol is presented here: the national flag.

A second step focused more deeply on the co-hashtags and images visualization on Gephi scraped through the Python script Instaloader, using the two hashtags #srebrenica and #cребреница.

As it can be seen by the Gephi visualization, for the latin #srebrenica word, the results reflect the dominant position that was explained before. There is a big connection with the idea of genocide, massacre, empathy and solidarity, especially from the turkish-muslims.

The marginal voice connected with the term #сребреница can be, instead, more deeply analyzed. In fact, new clusters of words appear, thanks to the co-hashtag. There’s a cluster completely dedicated to Mladic seen as a hero, but als another one that depicts the Serbs heroes. It is necessary, furthermore, to highlight that the cluster of “Serbs as Victims” is still the most representative cluster.

Findings on YouTube

YouTube recommendation networks

Combining the language and region specific datasets in one recommendation network, and using the modularity class (1.0) statistic on Gephi visualised 13 clusters. Different modularities related to different contents are divided into different colour with Gephi. As well as there are also different languages in these videos.

• 22.09% Insight pieces about the 'betrayal' of Srebrenica, the cowardness of the UN and war footage.

• 18.26% Slavic news media about 'local' stories about the war. There is a video of the attack on the Serbian Prime Minister and other small panels on tv-shows.

• 16.19% Marches and 'local' videos about the present time.

• 10.76% Bosnian (or Muslim) news media about the bosnian victims and effects.

• 7.98% Dutch news media about the events of the war and the role of the Dutchbats.

• 7.61% Remembrance of the suffering of the Bosniaks / recollection of the decisions made by the leaders of the fighting parties.

• 4.4% 'Latin' news media about general war atrocities.

• 4% Cyrillic news media about the war atrocities of the 'Serbians', and how it affected the (surviving) victims.

• 2.89% High school / college life in Srebrenica

• 2.4% General war atrocities

• 1.62% Polish news media about the war atrocities (Personal / documentary)

• 1.07% Local news media commemoration

• 0.74%- Youtube Sketch Channel

There are many actors involved in the complex web of the events surrounding the genocide of Srebrenica genocide, however, we decided to narrow them to either Bosnian, Serbian or Dutch. Upon looking at the network, one of the first thing that stand out is the shape of it.

Certain ‘spikes’ are seen, most notably by the Dutch and Serbian news media. The distance from the center is due to the relevance to the main subject ’Srebrenica’. Although the Dutchbats played a major role in the staging of this genocide, it can be noted that the Dutch News Media Network is very distanced from the other clusters.

Closer analysis show a low presence of videos about the genocide in general, and a medium degree of videos explaining the role of the Dutchbats. The highest amount of videos are about the ‘trauma’ the Dutchbats experienced following the Srebrenica genocide.

Further ‘crawls’ within the cluster result in ‘unrelated’ videos about the topic. One notable branch within the cluster is the one separated from the cluster, which described the role of the Dutch military in Dutch Indonesia. These videos are directed by a small amount of videos. The major directories within this cluster is to videos from the same network or NGO, unrelated to the original subject.

By setting up an ego network with the biggest nodes, we are looking to see if we can find connections cross-lingual. The aim of this was to find the pathway certain users would find within the clusters and how that would shape their narrative.

Another noteworthy cluster is that of Cyrillic news media. Also detached from the centre of the network, it shows little informative videos about the actual genocide, although it does have a graphic video with executions.

In the Youtube Network, the languages (and therefore the perspectives) are segregated by several crawls. How can the user go from the videos from one cluster to the ones in another cluster, and get a different perspective into the genocide?

The Dutch, Serbian and Bosnian News Media Video Network

The relationship between the Bosnian and Serbian cluster

How the Serbian cluster in-degrees relate to the Bosnian cluster

The Serbian cluster bridges to the Bosnian with videos about the body identification process, the war atrocities and Serbian Propaganda. The bridges are very important to see how the Serbs are dealing with their role in the genocide, and it seems they are interested in the actual facts about the genocide.

How the Bosnian cluster in-degrees relate to the Serbian cluster

The Bosnian links to the Serb with a live stream of newsnetwork ‘NewsOne’, however the actual content of this live stream can not be traced. There are also news reports in Cyrillic about the Russia Veto of the acknowledgement of the genocide and a news report about the freedom march. It can be argued that the Bosnians are looking for accountability in Serbian Media, and there are videos that acknowledge the genocide.

The relationship between the Dutch and Serbian cluster

How the Dutch cluster in-degrees to the Serbian cluster

The node bridges in the Serbian Cluster show news reports about the soldiers who executed Bosnians in the genocide, where they are either convicted or portrayed. The Node Bridges show no relationship towards the Serbian Sentiment (however, if one would to translate the Cyrillic comments, there is a huge presence of national pride about the genocide and justification). Even though the Cyrillic audience justifies the behaviour of the soldiers during the events that led up to the Genocide, the videos with a high In-Degree in the cluster tell the narrative of the Bosnian victims.

How the Serbian cluster in-degrees to the Dutch cluster

The Nodes that are in the Dutch cluster bridging the Serbian one is actually just one video: “Srebrenica: The deadly summer of 95”, where the events leading towards the Genocide are told from the Dutch Perspective.

The relationship between the Dutch and Bosnian cluster

How the Dutch cluster in-degrees to the Bosnian cluster

Within the Bosnian cluster, there are several nodes bridging to the Dutch cluster. What is notable about the videos that are bridging between the Dutch and Bosnian, are that they are about the proceedings towards the genocide, as well as a video what has happened to Srebrenica after 1995. The Bosnian network offers they who are interested an non-curated list of videos that tell less about the actual events, but offer a narrative about the situation Srebrenica was in.

How the Bosnian cluster in-degrees to the Dutch cluster

The Bosnian network bridges to the Dutch Network with a few videos.

“Why Srebrenica HAD to falll”

“Srebrenica still hunts the Dutchbats”

“Standing at the gate “From Srebrenica with love”

“Dutchbat Veteran Kok disputes verdict Srebrenica”

Important to note here is that the bridging notes from the Dutch network are all about humanising the ‘mistake’ the Dutchbats made. And to justify their decision to leave Srebrenica. The Dutch network simply stretches out into unrelated videos from the same network or the Dutch military, offering little to none documentation of the genocide.

Understanding Youtube’s Recommendation Networks as Linguistic Search Point of Views

While the findings above are building on merging the 4 distinct databases into a single one, there is also value in attempting to maintain their distinctiveness. Parallel to the content analysis of the combined network, a more topological approach was undertaken to highlight the distinctiveness of the original databases within the whole. To do so, the following steps were undertaken:

1. Using YouTube tool, four searchers were performed to generate the initial datasets of recommendation networks of Srebrenica on YouTube across four languages and regions (BS; BA; RS; SR; EN, US; NL, NL) - these represent YouTube ’s Unique linguistic point of view (LPOV; Massa and Scrinzi, 2012). However, unlike more traditional LPOVs as in the case of wikipedia, this one combines the elements of search and personalisation as performed by YouTube ’s machine learning algorithims, based on the user behaviour of those with the corresponding interface language and location set in their YouTube (Google) profiles.

   1. A new column, language, was added to each dataset, to indicate from which original LPOV the file came.

   2. Following, an attempt was made to merge the files together, but it soon became evident that the existence of the same nodes (videos) in various recommendation networks causes issues. Since all information except the “language” field was shared for a node that appeared more than once, Gephi appended the information into a single one, but kept the linguistic designation of the first dataset. Thus, for example, as we started with SR and appended all other networks to it, 50% of the nodes in the final network ended up being “Serbian”, because those included both all unique nodes appearing only in the Serbian dataset and all nodes that appear in Serbian and any other language. A more refined approach was needed.

2. The recommendations network were then exported as CSV files (one for nodes and one for edges). The nodes files were combine (see 3 below) while the edges were kept for future reintegration into the complete network. This was crucial since we wanted to maintain the recommendations (edges) and LPOV in the videos (nodes) in the final graph.

3. We then added to each dataset 4 linguistic binary columns, to indicate whether the video is found within the recommendation network of that language. For instance, the image below indicates that this video appears for the Bosnian and Serbian searches for [Srebrenica] and [Сребреница] respectively, but not when searching for [Srebrenica] on YouTube with English or Dutch as the interface language.

4. Using VLOOKUP function, the 4 tables were then compared to each other 4 times (once per each starting LPOV table). This was done to include all unique videos per language. The resulting 4 tables were merged, duplicates were removed and the data was otherwise cleaned.

5. Finally, a new network graph was created from the combined data table, and the previously exported edge CSV files were added on top of it. The resulting network was spatialised using FA2. The graph allowed us to do several things:

1. Using the total number of times a video appears in a certain language showcase the common (canonical) and separate recommendations networks existing within and across languages.

2. Finding the “negative spaces” of the networks - for instance, which videos exist in all but one language, and what such videos can tell us about YT’s conceptualisation of that language’s LPOV.

Unfortunately, we did not have enough time to go in-depth into the different ramification for this specific dataset.

6. Discussion

Mapping and analysing the issue sphere around war atrocities of recent conflicts, we draw attention to the complexities of how issues are perceived and commemorated by different factions involved, and how these are intimately tied to local, national, and transnational politics. The Srebrenica commemoration we found similarities in images and framings across platforms related to solidarity, such as the flower of hope, or the number of genocide victims in the dominant framing on TW and Insta. Conversely the marginal framing of the antiprogram, draws attention to the suffering of Others, on TW and Insta. Another significant insight comes from the YouTube recommendation algorithm and the potential for peace builders to map what videos are bridge nodes between one cluster and another. To recognize the Serbian casualties during 1991-1995, similar iconography is used. The significance for reconciliation efforts is the political question who is allowed to commemorate and who should we remember? Similar discussions emerge around the suffering of families of Nazi supporters during WWII. To what extent can the voices of marginal Others be recognized if those others are blamed to be culprits of violent atrocities.

7. Conclusion

Research on transitional justice is concerned with how people deal with their violent past in the present. Transitional justice, a concept emerged in the 1990s, focuses on how people with a shared violent past deal with these traumas in the present. The aim is to recognize the victims suffering, focusing on forgiveness and eventually leading to reconciliation between people in postwar societies. The commemoration of Srebrenica, is one of these forms of transitional justice, which strives to recognize the suffering of the Bosniak victims. But, this event is not recognized or felt by Bosnians equally across ethnicities, or transnationally between Bosnians or the Dutch for example. The different views and perspectives materialised in the different framings. But in the recommendation network, we saw bridging nodes where Bosnian were watching videos that explained the Dutchbats role in the conflict more specifically, and vice versa, we saw Dutch recommendations bridged with videos that focused on the present-day perspective on Srebrenica and the postwar situation in Bosnia. This tells us that there are possibilities for clusters to move outside their own “echo chamber” (Chun 2018), however these were very marginal and further research on the dynamics and strategies for intervention is needed.

Further sociological research could zoom into the role of NGOs and political campaigning during commemoration in general or try to incorporate a comparative perspective by incorporating the Holocaust remembrance or the Rwandan genocide. Second, further technical research could focus on how NGOs or other political activists can retrain recommendation algorithms to break free from digital segregation to foster empathy and understanding between civilians in postwar ethnically divided societies.

8. References

Chun, W.H. (2017, April 17) “We're all living in virtually gated communities and our real-life relationships are suffering”. In Wired. Accessed 18-7-2018 from:http://www.wired.co.uk/article/virtual-segregation-narrows-
our-real-life-relationships

Halilovich, H. (2013). Places of pain: Forced displacement, popular memory and trans-local identities in Bosnian war-torn communities (Vol. 10). Berghahn Books.

Massa, P., & Scrinzi, F. (2012, August). Manypedia: Comparing language points of view of Wikipedia communities. In Proceedings of the Eighth Annual International Symposium on Wikis and Open Collaboration (p. 21). ACM.

Mehler, D. (2017). The last ‘never again’? Srebrenica and the making of a memory imperative. European Review of History: Revue européenne d'histoire, 24(4), 606-630.

Rogers, R. (2018). Otherwise Engaged: Social Media from Vanity Metrics to Critical Analytics. International

Journal of Communication : IJoC, 12, 450-472.

Rogers, R., & Sendijarevic, E. (2012). Neutral or National Point of View? A Comparison of Srebrenica articles across Wikipedia's language versions. Proc. Wikipedia Academy.