On May 25th, 2020, a viral eleven-minute clip showing the murder of George Floyd sparked international outrage and solidarity, leading to the digital memorial event Blackout Tuesday on Instagram. We analyzed posts to compare Blackout Tuesday discourse with #blacklivesmatter movement conversations. Using topic modeling, we identified dominant themes and counter-narratives in Blackout Tuesday and #blacklivesmatter captions. Using hashtag co-occurrence analysis, we investigate hashtag networks to situate the discourses within spheres of Instagram activism. Our findings indicate that both corpora share themes like ”calls to action”, but Blackout Tuesday posts are shorter and solidarity-focused, while #blacklivesmatter posts are longer and address white privilege more explicitly. #blacklivesmatter is linked to anti-racist activism hashtags, while Blackout Tuesday connects more with popular culture and #Alllivesmatter. This supports qualitative research on Blackout Tuesday's performative allyship, adding a quantitative perspective to existing research.
The number of posts from the #blacklivesmatter movement (#blm) is estimated to be 28 million, which exempliifes the movement’s impact on society [ 1, 2]. However, the popularity of the movement reached a peak when “Blackout Tuesday” (BT) took place, a digital memorial day on Instagram [1, 2]. BT was caused by a wave of outrage about the murder of George Floyd, an African American who was killed on May 25, 2020 by two white police oficers in Minneapolis. The outrage was sparked by an 11-minute clip of the murder which went viral in social media. The video was posted in the context of the #blm movement and in a cultural setting where African Americans perceived law enforcement as agents of brutality [3]. To postulate solidarity with African Americans in their fight for racial justice, social media users posted a black square and added a post caption with the hashtag #blackouttuesday. Given the cultural context, the video supported the perception of white police brutality, 2. BT and #blm on Instagram white supremacy, and systemic injustice against African Americans. While #blm needed years to gain a large #blm is a movement started by the three African Amerinternational audience, BT reached millions within a day. ican women Alicia Garza, Patrisse Cullors, and Opal
The #blm movement has received significant attention Tomet. It is described as: ”an ideological and political inin research and has had a strong impact on discussions tervention in a world where Black lives are systematically in mainstream media; however, little attention has been and intentionally targeted for demise. It is an afirmation of Black folks’ contributions to this society, our humanity, and our resilience in the face of deadly oppression” [4]. The number of posts from the #blm movement is estimated to be 28 million, which exemplifies the movement’s impact on society [1, 2]. #blm is a new-new social movement: hierarchical, participatory and decentralized, paid to BT to date. Most existing research on BT has been conducted using interview studies and hermeneutic methodologies. In addition, there is one quantitative study by Chang et al. [7] which examines the contents of images, focusing on visual and geographic analyses.
Because a large share of the posts related to BT feature a black tile rather than an image, it is also important to examine the text in the post, the caption.
To fill this gap, this study used text mining to investigate the content of posts on BT. We examined the interrelations between BT and #blm by applying topic modeling and hashtag co-occurrence analysis to both discourses within the same period of time. Our aim was to understand how BT has impacted the #blm discourse. We also wanted to compare the diferent types of discourse, as BT is a digital memorial day, whereas #blm is a new social movement. repertoires, combining connective with collective action 4. Digital memorial day versus new social [5, 6]. In the summer of 2020, 20 million people joined movement: BT and #BlackLivesMatter constiprotests to address racial injustice and demand account- tute diferent forms of political participation in ability for the murder of countless Black people at the social media. Is this visible in the posts, topical hands of law enforcement [7, 8]. clusters and networks of hashtags? If so, can we
BT was part of the campaign #TheShowMustBePaused, draw conclusions about how discussions are led spreading from the music industry to social media [9, 10]. diferently depending on the type of participation, It was started by Jamila Thomas and Brianna Agyemang, in this case a memorial day versus a new social both female music executives at major record companies movement? in the U.S. [10]. The campaign was opened in response to the murders of George Floyd, Breonna Taylor, Ahmaud Arbery and other Black citizens at the hands of police 3. Data Collection and Description [11]. Thomas and Agyemang encouraged music industry professionals to halt business operations for the duration Building on the framework of Cultural Analytics, we base of June 2, 2020 to prompt conversation about financial em- our analysis on cultural sampling [16]. Because hashtags powerment of Blacks in the music industry [10, 11]. The structure online discourse [12], we used the hashtags campaign involved posting a black square on Instagram, #blackouttuesday and #blm to create our cultural sample. while refraining other social media activity that day [10]. Manovich highlights that current humanities research Soon, the hashtag morphed and circulated, with posts follows ”Close Reading” [17] as the dominant paradigm being diferent than intended [ 9, 10]: the hashtag was of textual analysis which puts single artifacts by profesused to postulate solidarity with George Floyd, instead of sional authors at the center of research [16, 18]. Using encouraging financial empowerment of Blacks in the mu- cultural samples allows to investigate nonprofessional sic industry. This also impacted the #blm conversation vernacular created by ”regular” people [16]. We add a on Instagram. distant reading perspective to existing research on racial
Often, users posted the hashtag #blm in #blackouttues- justice movements by extracting a cultural sample of day posts. Hashtags allow users to find specific infor- #blm and BT from Instagram posts. mation or to monitor a situation because they serve to We collected our corpus retrospectively using Crowdstructure discourse centered around a specific topic in Tangle1. We used the two search terms #blacklivesmatsocial media platforms like Instagram [12, 13]. In table 2, ter and #blackouttuesday to find public Instagram posts we show that nearly 10% of #blackouttuesday posts were published within a three month period following the tagged with the hashtag #blm. Using #blm in BT posts death of George Floyd. Our corpus comprises posts from concealed #blm posts with social justice information, hin- 05/24/2020 11:59 pm to 08/24/2020 12:00 am. We exdering critical collective organization [10, 13]. Therefore, ported the data in two batches, as CrowdTangle’s search #blm called not to use the hashtag along with BT posts. limits exports to a maximum of 300,000 posts. Using Next to concealing #blm posts, BT critics called it “white this method, we collected 548,249 posts for #blm, and guilt day” [14]. Posts were considered as empty gestures 305,344 posts for #blackouttuesday. The CrowdTangle [15, 10]. This accusation was not made in the context of dataframes contain posts as a link to one image per post #blm. and the ”description” column that contains the caption
With a bird’s eye view on hashtag relations and topics, of the post. In addition, the exported data includes metawe hope to gain insight into diferences between BT posts data for each post: The timestamp when the post was and #blm. Thus, we investigated the following research published, the username and name of the creator, and questions: interaction metrics for likes and comments. Our analysis concentrates on captions, the textual part of Instagram 1. Topic prevalence and significance: What are posts. Additionally, we use the timestamps for a descripthe dominant themes and topics in the #black- tive analysis of trends within our corpus. outtuesday and #blm Instagram captions? Com- We collected the top hashtags and sorted them by usparing the two, what are the core concerns? age in Table 2. Expectedly, #blm posts referenced BT, with 2. Counter speech: Which counter narratives ap- #georgefloyd, #blackouttuesday and #icantbreathe in the pear? top twenty hashtags. Unlike #blackouttuesday, #blm con3. Hashtag co-occurences Which hashtags co- tains no references to the original campaign #theshowoccurring in the #blackouttuesday and #blm feeds mustbepaused in the top hashtags. #blackouttuesday also are most prevalent, and do their interconnections includes #alllivesmatter, a hashtag signaling counterdisform distinct clusters? What do these hashtag course. Breonna Taylor, another victim of police brutality, networks suggest about the broader contexts and is referenced in the #blm corpus. intersections within the sphere of black activism?
The type-token density is a measure for language complexity. It allows insight into the complexity of both discourses, which we present in Table 1. Expectedly, Omena introduces hashtags as natively digital objects type-token density is high in both discourses when mea- that enable users to join debates on the local and global sured including hashtags. We attribute this to special scale through their indexing function [19]. Following characteristics of social media language, such as frequent Roger’s digital methods approach [20], we use these hashuse of hashtags, non-standardized spelling and the use of tags as digital traces [21] to study the #blm movement in emojis. While BT has a higher type-token density than the light of BT. Co-occurrence analysis allows to extract #blm when hashtags are included (0,96 vs 0,88), we can a network of hashtags, which gives insight into the movesee that #blm has a higher type-token density when hash- ments’ relations to other activist discourses indexed by tags are excluded (0.79 vs 0.56 without hashtags). This hashtags. shows that more diferent hashtags are used in #black- For the co-occurence analysis, we preprocessed both outtuesday posts than in #blm posts. At the same time, corpora in the same way. We extracted the hashtags from the text without hashtags is less lexically diverse than each caption using regex, lowercased the hashtags and in #blm posts. The type-token density in #blackouttues- counted the occurences of each hashtag. We selected day is only slighty above average (0.56), while #blm has the top 1,000 hashtags for each corpus and created a a type-token ratio that is quite high (0.79). #blm posts co-occurrence network, counting the co-occurence for are also longer than #blackouttuesday posts. On aver- each hashtag pair. Each network was imported to Gephi, age, #blm post lengthis three times longer than that of where we used the modularity algorithm [22] to find #blackouttuesday posts (69 tokens incl. hashtags to 19 hashtag clusters [23]. In a last step, we plotted the nettokens). This ratio remains consistent when substracting work for each modularity class within each of the two hashtags (60 vs 16 tokens). networks. These plots were the basis for our qualitative exploration of the hashtag clusters. Through this exploration, we were able to name each cluster based on the hashtags they contain. We extracted the modu6. Results 6.1. Core concerns and narratives larity classes associated with each hashtag to conduct a quantitative assessment of the hashtag clusters. We excluded the search hashtag from each network during this mapping process to mitigate potential biases (#blackouttuesday for the one, #blm for the other). Each post was then assigned to a specific class based on a majority rule approach which considered the hashtags present in the post. We labeled cases as ’ambigious’ where a clear majority for a particular modularity class was not evident.
The hashtag clusters are saved in a digital repository.2 We visualize the most frequent themes occuring in the datasets in figure 1 and figure 2, with the bubble size representing the relative frequency within the 100 most frequent topics. The colors for shared topics in both discourses correspond. The datasets share many common themes. We identify that both datasets contain many ”calls to action”. Apparently, many posts aim to activate readers politically, for example by joining protests or 5. Topic Modeling signing petitions. Other calls to action are more generic, manifesting in topic words like fight or change. Other Topic modeling is a method to cluster themes in large posts ask readers to become conscious of racism and corpora that is widely applied in the digital humanities. white privilege. In the #blm dataset, 30% of posts are Typical for social media data, our posts are quite short, ”calls to action”, while only 13 % of the BT posts fall into with post length ranging from ¯= 19 token for BT posts this category. to ¯ = 69 token for #blm posts. Therefore, we chose to We identified the theme ”voice-of-color” in the topemploy BERTopic [24] for the topic modeling due to its ics. The voice-of-color is an established thesis from critability to handle sparse data. ical race theory. It holds that alleged minority status
We applied only minimal preprocessing. We removed brings with it a presumed competence to speak about @mentions for privacy protection and deleted 29 post race and racism [28]. The speech act of ”breaking the duplications that are a result of the scraping process. silence” appears in both corpora. It is more present in the Next to this, we removed words with two or less letters, #blm dataset (for a comparison, see figure 1 and figure 2). stopwords and the hashtags. Within the BT dataset, this becomes visible with topics
After preprocessing and topic modeling, we assigned that include words such as voice, heard, voices, use, space. labels to the 100 most frequent topics of each dataset Within the #blm dataset, this becomes even more clear, (postprocessing). For a human eye, it becomes clear that with topics such as Silent, silence, quiet, staying, silenced, many topics follow broader themes. Therefore, we add an Voice, voices, heard, amplify, use. additional step by identifying broader themes consisting Both corpora share themes, but we identify two big of similar topics after the postprocessing. diferences. Common themes are mentions of other anti
Firstly, we apply the baseline application of BERTopic racist movements, references to African American artists [24], using UMAP for dimensionality reduction and the and musicians, or references to platform afordances. A HBDSCAN minimal cluster size to 30 [25, 26]. This re- diference lies in internationality. 21% of BT topics are duced the amount of topics drastically, which is why we written in other languages than English, such as Spanish, set the minimal cluster size to 150 items. Although the German, French or Russian which points to the internabaseline application of BERTopic [24] yields good results tional character of BT (considering the 100 most frequent in terms of readibility and topic diversity, we conducted topics). In contrast, #blm is rooted in the English speakan experimental study for finetuning on samples of both ing countries U.S., Canada, and Australia [29]. Another datasets to increase topic diversity (n=0,1%). Maximal diference between the corpora is the perspective of soliMarginal Relevance considers the similarity of tokens darity which is prevalent in BT posts. 7% of topics relate with the document, along with the similarity of already to solidarity, using hashtags like #icare or #togetherforselected keywords and keyphrase [27, 24]. We found that change. In contrast, the #blm dataset thematizes equality topic words consisted of two words instead of one in and privilege, calling out white privilege (3% of topics). many cases, but would sometimes contain the same word While solidarity expresses the perspective of an outsider, twice. The application of MMR did not increase topic the corresponding #blm theme expresses a deeper underdiversity. We found that the right preprocessing is more standing of racism and systems of oppression. important to obtain high topic diversity. 2https://osf.io/cu2bj/?view_only=bc770f9539c64682a0bd477d5bd6bb99
The modularity algorithm discovered five communities within the hashtag co-occurences for #blm, and six for BT. In case of BT, hashtags were split unevenly between the classes: Classes 2–4 contained between 3.6% (=33, related to of-topic hashtags. For example, two hashclass 3) and 5.8% (=53, class 4) of all hashtags, while tag networks contain mundane content related to the the classes 0, 1, and 5 contained between 16.6% (=152, food blogging and wildlife, such as #animalsofinsta, #aniclass 0) and 35.4% (=325, class 1) of all hashtags. The mallover or #wildlifeonearth. Another network revolves smallest cluster contains hashtags referring to food and purely around U.S. sports men and basketball. Unlike animals. Class 4 contains references to sport and class the comparable co-occurence network at #blm, these net5 the #blackgirlmagic and #blackbusiness theme. The works do not contain any other hashtags from Africanclasses 0, 1, and 5 contain political hashtags, multilin- American communities except for #blm and #blackexcelgual hashtags, and content-related hashtags (like #por- lence within the animal topics. This shows the maintrait). The hashtag #blackouttuesday appears in class 0, stream character of BT compared to #blm. This is supthe multi-lingual class, possibly as the unique bridge to ported by the wide political spectrum visible in solidarthe other clusters. When mapping posts to modularity ity hashtags, referencing conservative and republican classes based on the top 1000 hashtags, 71.6% (=56783) hashtags alongside hashtags on the political left such as of all posts were identified to belong to class 1, the clus- #socialist or contents related to the democratic party. ter that contains most hashtags related to the movement, A shared topic are support networks for black busilike #theshowmustbepaused and #justiceforgeorgefloyd. nesses and empowerment content of black women, re10.6% of posts (=8371) were mapped to the multilin- lated to hashtags such as #BlackGirlMagic, #BlackEmgual cluster (0), 8.21% (=6507) to cluster 5, and 6.28% powerment or #melaninpoppin. Scholars have estab(=4975) of posts were classified as ambiguous, as they lished that empowerment occurs within the realm of did not show a clear majority for the one or the other social media. For example, #BlackGirlMagic, introduced cluster. A minority of posts was mapped to classes 2–4. by CaShawn Thompson, negotiates societal presentations
The hastags co-occuring in the #blm network are more of Black women [30]. Black women-centered discourse evenly split into four clusters: from 37.3% (=373) in class achieves empowerment by highlighting their experiences 3 to 13.7% (=137) in class 2. Hashtags contained in the in ways that are often neglected or distorted in traditional largest cluster (3) are mostly non-political and mundane media outlets [31]. Within the #blm corpus, these hash(#family, #food, #college, #followme). The smallest class tags are closely connected to mental health content. In (2) contains #black+x hashtags, like #blackqueen, #black- the BT context, hashtags are connected to support hashnews, #blackbloggers. Class 0 (24.4%, =244) contains tags for Black businesses. hashtags revolving around justice in combination with diferent topics, as well as allyship hashtags (e.g. #istand- 6.3. Counter Speech withyou), while class 1 (24.6%, =24.6) clusters hashtags related to politics and policy issues (e.g #notmypresident, Another hashtag shared by both #blm and BT is the #guncontrol). The classification based on the hashtag colorblind hashtag #alllivesmatter. #Alllivesmatter is a mapping for posts shows a more even distribution for the counter-protest hashtag whose content argues that equal #blm hashtag which is congruent with the hashtag distri- attention should be given to all lives regardless of race bution across clusters. Most posts were mapped to cluster [32]. The ”All Lives Matter” movement is, ”one of the pri1 (31.1%, =127410), with the ambiguous classification mary ways in which people resisted the #blm movement coming second (19.2%, =78706). 17.5% ( =71907) [...] in the form of [...] a counterslogan to undermine posts were mapped to the mundane class 3, and 17.3% the purpose and message of the #blm call to action” [33]. ( =70947) to class 0. Finally, the smallest amount of Powell et al. have shown that the use of #blm or #Allposts (15.0%, =61308) were classified to cluster 2. LivesMatter are signals of political identity [34].
In general, the #blm co-occurence network shows that the social movement is closely related to other hashtags 6.4. Political Hashtags of Black activism, while also containing links to popular culture that are common to Instagram, such as art or Several political hashtags appear close to #blm, like photography. In the aftermath of George Floyd’s death, #berniesanders, #NeverTrump, #NeverBiden and #Prothe hashtags #justiceforgeorgefloyd, #protest, #policebru- gressives. A number of city hashtags are close to these tality, #justiceforbreonnataylor and the German hashtag hashtags, namely #LosAngeles, #Hollywood, #Brooklyn, #gegenrassismus (”against racism”), are closest to dif- #Atlanta, and #Chicago. A study analyzing Twitter proferent spellings of #blm. An example of the hashtag’s ifle information found that the #blm movement is largely relatedness to other movements of Black activism are ignored in places with a large percentage of white or empowerment hashtags, such as #blackexellencexx, #un- Hispanic populations, compared to places with smaller apologeticallyBlack and #BlackGirlMagic. percentages of these groups [35]. Published in 2019, the
While BT co-occurence networks contain hashtags study was conducted before BT which gave the #blm from a wider political spectrum than #blm, they are also movement a new spark both in the U.S. and internationally. A geospatial analysis could provide insight into whether this finding remains true after BT and if it is true for both #blm and BT posts. intersectional discrimination of African Americans and carry unconscious biases that are potentially harmful. 7. Discussion: Digital memorial day versus new social movement [1] S. Harlow, Journalism’s change agents: Black lives matter,# blackouttuesday, and a shift toward activist doxa, Journalism & Mass Communication Quarterly 99 (2022) 742–762.
Previous work has investigated visual aspects of BT, stud- [2] S. Ho, A social media ”blackout” enthralled instaied posting motivations and the the role of celebrities, gram. but did it do anything?’, NBC News (2020). while we studied Instagram post captions [36, 37, 14]. URL: https://shorturl.at/8RYmA. We contrasted topics and hashtag co-occurences of the [3] C. Chaney, R. V. Robertson, Racism and police digital memorial event BT and the impactful movement brutality in america, Journal of African American #blm. We found that they share many similar topics, Studies 17 (2013) 480–505. such as calls to action, mentions and thoughts of George [4] A. Garza, A herstory of the #blacklivesmatter moveFloyd, and connections to other antiracist movements. ment, ProudFlesh: New Afrikan Journal of CulHowever, BT posts were posted from the solidarity per- ture, Politics and Consciousness (2014). URL: https: spective, while #blm discourse broaches the issue of white //api.semanticscholar.org/CorpusID:141982276. privilege. Moreover, #blm is more closely related to other [5] M. Castells, Networks of Outrage and Hope - Social hashtags of Black activism, while BT posts are more fre- Movements in the Internet Age, John Wiley Sons, quently connected to posts related to popular culture, New York, 2015. underscoring its place in mainstream micro-activism. [6] B. Cammaerts, The new-new social movements: Nevertheless, topic modeling results show that many Are social media changing the ontology of social BT posts seek to mobilize people or express solidarity to- movements?, Mobilization: An International Quarwards the murder or police brutality (see figure 2, figure terly 26 (2021) 343–358. 1). [7] L. Buchanan, Q. Bui, J. K. Patel, Black lives matter
We gain insight into networks of Black activism on may be the largest movement in us history, The Instagram. #blm is embedded in a network of anti-racist New York Times 3 (2020) 2020. activism. Posts with the hashtag are on average more [8] M. M. Francis, L. Wright-Rigueur, Black lives matter than twice as long and have a higher type-token ratio. in historical perspective, Annual Review of Law In contrast, BT posts are shorter and contain many dif- and Social Science 17 (2021) 441–458. ferent hashtags. Posts in various languages characterize [9] L. Bakare, C. Davies, Blackout tuesday: black the memorial day as an international event. BT is an squares dominate social media and spark debate, international spark of outrage – and in its nature more The Guardian 2 (2020). superficial than #blm. We point to Wellman’s study, who [10] K. Blair, Empty gestures: Performative utterances investigates BT in the light of performative allyship [37]. and allyship, Journal of Dramatic Theory and CritNext to this, future work should compare the contents of icism 35 (2021) 53–73. #alllivesmatter and BT posts. [11] TheShowMustBePaused - theshowmustbepaused.com, https://www. 8. Ethics theshowmustbepaused.com/, ???? [Accessed 12-07-2024].
This paper is based on a poster created for the 8th an- [12] S. J. Jackson, M. Bailey, B. F. Welles, # HashtagAcnual conference of the association Digital Humanities im tivism: Networks of race and gender justice, Mit deutschsprachigen Raum, which called for papers with Press, 2020. the topic ”Kulturen des digitalen Gedächtnisses”, engl. [13] A. Willingham, Why posting a black image with the Cultures of digital memory [38]. We researched #Black- “black lives matter” hashtag could be doing more outtuesday due to the actuality of the topic and the true harm than good, CNN (2020). interest in the memorial culture of Blackout Tuesday, [14] S.-S. Duvall, N. Heckemeyer, #BlackLivesMatter: an international memorial day to the African American black celebrity hashtag activism and the discursive victims of white police brutality in the U.S.. This paper formation of a social movement, Celebrity Studis limited due to the authors’ outsider perspective. As ies 9 (2018) 391–408. URL: https://doi.org/10.1080/ white Europeans, we can in no way comprehend the 19392397.2018.1440247. doi:10.1080/19392397. 2018.1440247. [15] A. Valen Levinson, Ambivalent action: Recognizing Edition) - An Introduction, NYU Press, New York, bothness in the narratives of blackout tuesday 1, 2017. in: Sociological Forum, volume 38, Wiley Online [29] L. C. Hillstrom, Black Lives Matter: From a moment Library, 2023, pp. 553–574. to a movement, Bloomsbury Publishing USA, 2018. [16] L. Manovich, Cultural Analytics, MIT Press, Cam- [30] C. J. Porter, J. A. Byrd, Juxtaposing# blackgirlmagic bridge, 2020. as “empowering and problematic:” composite nar[17] B. H. Smith, What was “close reading”? a century of ratives of black women in college., Journal of Dimethod in literary studies, The Minnesota Review versity in Higher Education 16 (2023) 273. 2016 (2016) 57–75. [31] M. Erigha, A. Crooks-Allen, Digital communities of [18] F. Moretti, Distant Reading, Verso Books, London, black girlhood: New media technologies and online 2013. discourses of empowerment, The Black Scholar 50 [19] J. J. Omena, E. T. Rabello, A. G. Mintz, Digital (2020) 66–76.
Methods for Hashtag Engagement Research, Social [32] R. J. Gallagher, A. J. Reagan, C. M. Danforth, P. S. Media + Society 6 (2020) 2056305120940697. URL: Dodds, Divergent discourse between protests and https://doi.org/10.1177/2056305120940697. doi:10. counter-protests:# blacklivesmatter and# alllives1177/2056305120940697. matter, PloS one 13 (2018) e0195644. [20] R. Rogers, Digital Methods, MIT Press, Cambridge, [33] N. Carney, All lives matter, but so does race: Black 2015. lives matter and the evolving role of social media, [21] D. Lazer, A. Pentland, L. Adamic, S. Aral, A.-L. Humanity & society 40 (2016) 180–199.
Barabasi, D. Brewer, N. Christakis, N. Contractor, [34] M. Powell, A. D. Kim, P. E. Smaldino, Hashtags as J. Fowler, M. Gutmann, T. Jebara, G. King, M. Macy, signals of political identity:# blacklivesmatter and# D. Roy, M. Van Alstyne, Social science. Compu- alllivesmatter, Plos one 18 (2023) e0286524. tational social science, Science 323 (2009) 721– [35] M. Hafner, A place-based analysis of# blacklives723. URL: http://dx.doi.org/10.1126/science.1167742. matter and counter-protest content on twitter, Geodoi:10.1126/science.1167742. Journal 84 (2019) 1257–1280. [22] V. D. Blondel, J.-L. Guillaume, R. Lambiotte, E. Lefeb- [36] H.-C. H. Chang, A. Richardson, E. Ferrara, vre, Fast unfolding of communities in large net- #JusticeforGeorgeFloyd: How Instagram facilworks, Journal of statistical mechanics 2008 (2008) itated the 2020 Black Lives Matter protests, P10008. URL: https://iopscience.iop.org/article/ PloS one 17 (2022) e0277864. URL: http://dx.doi. 10.1088/1742-5468/2008/10/P10008/meta. doi:10. org/10.1371/journal.pone.0277864. doi:10.1371/ 1088/1742-5468/2008/10/P10008. journal.pone.0277864. [23] M. Bastian, S. Heymann, M. Jacomy, Gephi: an open [37] M. L. Wellman, Black squares for black lives? persource software for exploring and manipulating formative allyship as credibility maintenance for networks, in: Third international AAAI conference social media influencers on instagram, Social Meon weblogs and social media, 2009. dia+ Society 8 (2022) 20563051221080473. [24] M. Grootendorst, Bertopic: Neural topic modeling [38] M. Geierhos, DHd2022: Kulturen des digiwith a class-based tf-idf procedure, 2022. URL: https: talen Gedächtnisses. Konferenzabstracts, Zenodo, //arxiv.org/abs/2203.05794. arXiv:2203.05794. 2022. URL: https://doi.org/10.5281/zenodo.6304590. [25] L. McInnes, J. Healy, J. Melville, UMAP: Uni- doi:10.5281/zenodo.6304590. form Manifold Approximation and Projection for Dimension Reduction, ArXiv e-prints (2018).
arXiv:1802.03426. [26] L. McInnes, J. Healy, S. Astels, hdbscan: Hierarchical density based clustering, The Journal of Open
Source Software 2 (2017) 205. [27] J. Carbonell, J. Goldstein, The use of mmr, diversitybased reranking for reordering documents and producing summaries, in: Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR ’98, Association for Computing Machinery, New York, NY, USA, 1998, p. 335–336.
URL: https://doi.org/10.1145/290941.291025. doi:10.
1145/290941.291025. [28] R. Delgado, J. Stefancic, Critical Race Theory (Third