Divergent discourse between protests and counter-protests:#BlackLivesMatter and #AllLivesMatter

Ryan J. Gallagher,∗ Andrew J. Reagan, Christopher M. Danforth, and Peter Sheridan DoddsDepartment of Mathematics and Statistics, Computational Story Lab,

Vermont Complex Systems Center, and Vermont Advanced Computing CoreThe University of Vermont, Burlington, VT 05405

(Dated: May 23, 2017)

Since the shooting of Black teenager Michael Brown by White police officer Darren Wilson inFerguson, Missouri, the protest hashtag #BlackLivesMatter has amplified critiques of extrajudicialkillings of Black Americans. In response to #BlackLivesMatter, other Twitter users have adopted#AllLivesMatter, a counter-protest hashtag whose content argues that equal attention should begiven to all lives regardless of race. Through a multi-level analysis of over 860,000 tweets, westudy how these protests and counter-protests diverge by quantifying aspects of their discourse.We find that #AllLivesMatter facilitates opposition between #BlackLivesMatter and hashtags suchas #PoliceLivesMatter and #BlueLivesMatter in such a way that historically echoes the tensionbetween Black protesters and law enforcement. In addition, we show that a significant portionof #AllLivesMatter use stems from hijacking by #BlackLivesMatter advocates. Beyond simplyinjecting #AllLivesMatter with #BlackLivesMatter content, these hijackers use the hashtag todirectly confront the counter-protest notion of “All lives matter.” Our findings suggest that BlackLives Matter movement was able to grow, exhibit diverse conversations, and avoid derailment onsocial media by making discussion of counter-protest opinions a central topic of #AllLivesMatter,rather than the movement itself.

I. INTRODUCTION

Protest movements have a long history of forcing diffi-cult conversations in order to enact social change, and theincreasing prominence of social media has allowed theseconversations to be shaped in new and complex ways.Indeed, significant attention has been given to how toquantify the dynamics of such social movements. Recentwork studying social movements and how they evolvewith respect to their causes has focused on Occupy WallStreet [1–3], the Arab Spring [4], and large-scale protestsin Egypt and Spain [5, 6]. The network structures ofmovements have also been leveraged to answer questionsabout how protest networks facilitate information diffu-sion [7], align with geospatial networks [8], and impactoffline activism [9–11]. Both offline and online activistshave been shown to be crucial to the formation of protestnetworks [12, 13] and play a critical role in the eventualtipping point of social movements [14, 15].

The protest hashtag #BlackLivesMatter has come torepresent a major social movement. The hashtag wasstarted by three women, Alicia Garza, Patrisse Cullors,and Opal Tometi, following the death of Trayvon Martin,a Black teenager who was shot and killed by neighbor-hood watchman George Zimmerman in February 2012[16]. The hashtag was a “call to action” to address anti-Black racism, but it was not until November 2014 whenWhite Ferguson police officer Darren Wilson was not in-dicted for the shooting of Michael Brown that #Black-LivesMatter saw widespread use. Since then, the hashtaghas been used in combination with other hashtags, such

∗ ryan.gallagher@uvm.edu

as #EricGarner, #FreddieGray, and #SandraBland, tohighlight the extrajudicial deaths of other Black Ameri-cans. #BlackLivesMatter has organized the conversationsurrounding the broader Black Lives Matter movementand activist organization of the same name. Some havelikened Black Lives Matter to the New Civil Rights move-ment [17, 18], though the founders reject the comparisonand self-describe Black Lives Matter as a “human rights”movement [19].

Researchers have only just begun to study the emer-gence and structure of #BlackLivesMatter and its as-sociated movement. To date and to the best of ourknowledge, Freelon et al. have provided the most com-prehensive data-driven study of Black Lives Matter [20].Their research characterizes the movement through mul-tiple frames and analyzes how Black Lives Matter hasevolved as a movement both online and offline. Otherresearchers have given particular attention to the be-ginnings of the movement and its relation to the eventsof Ferguson, Missouri. Jackson and Welles have shownthat the initial uptake of #Ferguson, a hashtag that pre-cluded widespread use of #BlackLivesMatter, was due tothe early efforts of “citizen journalists” [21], and Bonillaand Rosa argue that these citizens framed the story ofMichael Brown in such a way that facilitated its even-tual spreading [22]. Other related work has attemptedto characterize the demographics of #BlackLivesMatterusers [23], how #BlackLivesMatter activists affect sys-temic political change [24], and how the movement self-documents itself through Wikipedia [25].

#BlackLivesMatter has found itself contended by arelatively vocal counter-protest hashtag: #AllLivesMat-ter. Advocates of #AllLivesMatter affirm that equal at-tention should be given to all lives, while #BlackLives-

arX

iv:1

606.

0682

0v5

[cs

.CL

] 2

0 M

ay 2

017

2

Aug 2014

Sep Oct Nov Dec Jan 2015

Feb Mar Apr May Jun Jul Aug

Time (Days)

0

5000

10000

15000

20000

Num

ber

of

Tw

eets Darren Wilson

non-indictment

Daniel Pantaleonon-indictment

Death of Two NYPD Officers

Baltimore Protests

Charleston ChurchShootingWalter Scott Death

Beyonce, John LegendGrammy Performances and

Chapel Hill Shooting

Outrage OverSandra Bland Death

Number of #BlackLivesMatter and #AllLivesMatter Tweets (10% Gardenhose Sample)

#BlackLivesMatter

#AllLivesMatter

FIG. 1. Time series showing the number of #BlackLivesMatter and #AllLivesMatter tweets from Twitter’s 10% Gardenhosesample. The plot is annotated with several major events pertaining to the hashtags. Shaded regions indicate one-week periodswhere use of both #BlackLivesMatter and #AllLivesMatter peaked in frequency. These are the periods we focus on in thepresent study.

Matter supporters contend that such a sentiment derailsthe Black Lives Matter movement. The counter-hashtag#AllLivesMatter has received less attention in terms ofresearch, largely being studied from a theoretical angle.The phrase “All Lives Matter” reflects a “race-neutral”or “color-blind” approach to racial issues [26]. Whilethis sentiment may be “laudable,” it is argued that race-neutral attitudes can mask power inequalities that resultfrom racial biases [27]. From this perspective, those whoadopt #AllLivesMatter evade the importance of race inthe discussion of Black deaths in police-involved shoot-ings [28, 29]. To our knowledge, our work is the firstto engage in a data-driven approach to understanding#AllLivesMatter. This approach not only allows us tosubstantiate several broad claims about the use of #Al-lLivesMatter, but to also highlight trends in #AllLives-Matter that are absent from the theoretical discussion ofthe hashtag.

Given the popular framing of Black Lives Matter asthe New Civil Rights movement, the movement and itsoppositions are in and of themselves of interest to study.Furthermore, while qualitative study of #AllLivesMat-ter has provided a theoretical framing of the hashtag,it is still unclear to what extent, if any, the All LivesMatter movement hijacked the conversations of #Black-LivesMatter. Through the computational analysis of over860,000 tweets, in this paper we comprehensively quan-tify the ways in which the protest and counter-protestdiscourses of #BlackLivesMatter and #AllLivesMatterdiverge most significantly. Unlike previous studies of po-litical polarization that focus on hashtag trends or cu-rated lists of terms [31, 32, 34, 35], the methods we lever-

age take advantage of the entirety of textual data, thusgiving weight to all protest sentiments that were voicedthrough these hashtags.

Through application of these methods at the wordlevel and topic level, we first show that #AllLivesMat-ter diverges from #BlackLivesMatter through supportof pro-law-enforcement sentiments. This places #Black-LivesMatter and hashtags such as #PoliceLivesMatterand #BlueLivesMatter in opposition, a framing that isin line with the theoretical understanding of #AllLives-Matter and that mimics how relationships between blackprotesters and law enforcement have been historically de-picted. We then demonstrate that #AllLivesMatter ex-periences significant hijacking from #BlackLivesMatter,while #BlackLivesMatter exhibits rich and information-ally diverse conversations, of which hijacking is a muchsmaller portion. These findings, which augment the the-oretical discussion of the impact of the All Lives Mattermovement, suggest that the Black Lives Matter move-ment was able to avoid being derailed by counter-protestopinions on social media by relegating discussion of suchopinions to the counter-protest, rather than the move-ment itself.

II. MATERIALS AND METHODS

A. Data Collection

We collected tweets containing #BlackLivesMatterand #AllLivesMatter (case-insensitive) from the period

3

August 8th, 2014 to August 31st, 2015 from the Twit-ter Gardenhose feed. The Gardenhose represents a10% random sample of all public tweets. Our sub-sample resulted in 767,139 #BlackLivesMatter tweetsfrom 375,620 unique users and 101,498 #AllLivesMattertweets from 79,753 unique users. Of these tweets, 23,633of them contained both hashtags. When performing ouranalyses, these tweets appear in each corpus.

Previous work has emphasized the importance of view-ing protest movements through small time scales [20, 21].In addition, we do not attempt to characterize all ofthe narratives that exist within #BlackLivesMatter and#AllLivesMatter. Therefore, we choose to restrict ouranalysis to eight one-week periods where there were si-multaneous spikes in #BlackLivesMatter and #AllLives-Matter. These one-week periods are labeled on Figure 1and are as follows:

1. November 24th, 2014 : the non-indictment of Dar-ren Wilson in the death of Michael Brown. [37]

2. December 3rd, 2014 : the non-indictment of DanielPantaleo in the death of Eric Garner [38].

3. December 20th, 2014 : the deaths of New York Citypolice officers Wenjian Liu and Rafael Ramos [39].

4. February 8th, 2015 : the Chapel Hill shooting andthe 2015 Grammy performances by Beyonce andJohn Legend [40, 41].

5. April 4th, 2015 : the death of Walter Scott [42].

6. April 26th, 2015 : the social media peak of protestsin Baltimore over the death of Freddie Gray [43].

7. June 17th, 2015 : the Charleston Church shooting[44].

8. July 21st, 2015 : outrage over the death of SandraBland [45].

B. Entropy and Diversity

A large part of our textual analysis relies on tools frominformation theory, so we describe these methods hereand frame them in the context of the corpus. Given atext with n unique words where the ith word appearswith probability pi, the Shannon entropy H encodes “un-predictability” as

H = −n∑

i=1

pi log2 pi. (1)

Shannon’s entropy describes the unpredictability of abody of text, and so, intuitively, we say that a text withhigher Shannon’s entropy is less predictable than a textwith lower Shannon’s entropy. It can then be useful tothink of Shannon’s entropy as a measure of diversity,

where high entropy (unpredictability) implies high di-versity. In this case, we refer to Shannon’s entropy asthe Shannon index. We employ this raw diversity mea-sure in Section III A to examine the diversity of languagesurrounding individual words.

Of the diversity indices, only the Shannon index givesequal weight to both common and rare words [46]. Inaddition, even for a fixed diversity index, care must betaken in comparing diversity measures to one another[46]. In order to make linear comparisons of diversitybetween texts, we convert the Shannon index to an effec-tive diversity. The effective diversity D of a text T withrespect to the Shannon index is given by

D = 2H = 2

(−

∑ni=1 pi log2 pi

). (2)

The expression in Eq. 2 is also known as the perplexity ofthe text. The effective diversity allows us to accuratelymake statements about the ratio of diversity between twotexts. For example, unlike the raw Shannon index, theeffective diversity doubles in the situation of comparingtexts with n and 2n equally-likely words. We apply theeffective diversity in Section III C.

C. Jensen-Shannon Divergence

The Kullback-Leibler divergence is a statistic that as-sesses the distributional differences between two texts.Given two texts P and Q with a total of n unique words,the Kullback-Leibler divergence between P and Q is de-fined as

DKL(P ||Q) =

n∑i=1

pi log2

piqi, (3)

where pi and qi are the probabilities of seeing word i inP and Q respectively. However, if there is a single wordthat appears in one text but not the other, this divergencewill be infinitely large. Because such a situation is notunlikely in the context of Twitter, we instead leveragethe Jensen-Shannon divergence (JSD) [47], a smoothedversion of the Kullback-Leibler divergence:

DJS(P ||Q) = π1DKL(P ||M) + π2DKL(Q||M). (4)

Here, M is the mixed distribution M = π1P+π2Q whereπ1 and π2 are weights proportional to the sizes of P andQ such that π1 + π2 = 1. The Jensen-Shannon diver-gence has been previously used in textual analyses thatrange from the study of language evolution [48, 49] to theclustering of millions of documents [50].

The JSD has the useful property of being bounded be-tween 0 and 1. The JSD is 0 when the texts have exactlythe same word distribution, and is 1 when neither texthas a single word in common. Furthermore, by the lin-earity of the JSD we can extract the contribution of anindividual word to the overall divergence. The contribu-

4

tion of word i to the JSD is given by

DJS,i(P ||Q) = −mi log2mi + π1pi log2 pi + π2qi log2 qi,(5)

where mi is the probability of seeing word i in M . Thecontribution from word i is 0 if and only if pi = qi. There-fore, if the contribution is nonzero, we can label the con-tribution to the divergence from word i as coming fromtext P or Q by determining which of pi or qi is larger.

III. RESULTS

A. Word-Level Divergence

For each of the eight time periods, the collectionsof #BlackLivesMatter and #AllLivesMatter tweets areeach represented as bags of words where user handles,links, punctuation, stop words, the retweet indicator“RT,” and the two hashtags themselves are removed. Wethen calculate the Jensen-Shannon divergence betweenthese two groups of text, and rank words by percent con-tribution to the total divergence. We present the re-sults of applying the JSD to the weeks following the non-indictment of Darren Wilson and the Baltimore protestsin Figures 2–5 and the remaining periods in AppendixFigures A1–A4.

All contributions on the JSD word shift graphs arepositive, where a bar to the left indicates the word wasmore common in #AllLivesMatter and a bar to the rightindicates the word was more common in #BlackLives-Matter. The bars of the JSD word shift graph are alsoshaded according to the diversity of language surround-ing each word. For each word w, we consider all tweetscontaining w in the given hashtag. From these tweets,we calculate the Shannon index of the underlying worddistribution with the word w and hashtag removed. Ahigh Shannon index indicates a high diversity of wordswhich, in the context of Twitter, implies that the wordw was used in a variety of different tweets. On the otherhand, a low Shannon index indicates that the word woriginates from a few popular retweets. We emphasizethat here we are using the Shannon index not to com-pare diversities between words, but to simply determineif a word was used diversely or not. By using Figure A2as a baseline (a period where #AllLivesMatter was domi-nated by one retweet), we determine a rule of thumb thata word is not used diversely if its Shannon index is lessthan approximately 3 bits.

By inspection of Figure 2, we find that #ferguson and#fergusondecision, both hashtags relevant to the non-indictment of Darren Wilson for the death of MichaelBrown, contribute to the divergence of #BlackLivesMat-ter from #AllLivesMatter. Similarly, in Figure 5 #fred-diegray emerges as a divergent hashtag during the Bal-timore protests due to #BlackLivesMatter. In each ofthese periods, #BlackLivesMatter diverges from #All-LivesMatter by talking proportionally more about the

1.5 1.0 0.5 0.0 0.5 1.0 1.5Contribution

(% of total JSD = 0.0855 bits)

50

45

40

35

30

25

20

15

10

5

1

Rank

murderedunarmedwomen

subtleerasure

#ripracism

insteadshot

identitiestimes

horizontallytraumaoppression

likebiased

retweettwitter

indicting#mainstreammedia

fuelingrepoing

workwrong

beatwannamakes

3000#williamsburg

agechanging

#ferguson#fergusondecision

#nycprotestpeoplereally

tweetwhite

structuralrelevant

everythingbrutality

epitomizedcountrysituationsingle

oppressioncommunities

stopignoring

facedthink

JSD Word ContributionsNovember 24, 2014 to November 30, 2014

10 5 0 5 10

#AllLivesMatter #BlackLivesMatter

FIG. 2. Jensen-Shannon divergence word shift graph for theweek following the non-indictment of Darren Wilson. Allword contributions are positive percentages of the total diver-gence, where bars to the left and right indicate the word wasmore common in #AllLivesMatter and #BlackLivesMatterrespectively. Shading indicate the Shannon index (in bits) oftweets containing the given word. Lighter shading indicatesthe contribution is due to one or several popular retweets.Darker shading indicates the contribution is due to the wordbeing used throughout many different tweets.

5

8.0 6.0 4.0 2.0 0.0 2.0 4.0Contribution

(% of total JSD = 0.2198 bits)

50

45

40

35

30

25

20

15

10

5

1

Rank

missionlet

#racist#shutitdown

#bluelivesmatter

choosenever

remainsprotesters#shutdown5thave

prayersampkeep

comeampgooursamemaintain

#nypdlivesmatterliu

officersmotivated

breathededicated

violenceramos

families#antoniomartin

colormind

job#blackxmas

yallgender

sexualityamerica

blowingclass

eitherheroes

protest#nypdguess

#ripericgarner#moa

#nypdstrong#icantbreathe

ukrip

brightonhove

mall#policelivesmatter

merica

JSD Word ContributionsDecember 20, 2014 to December 26, 2014

10 5 0 5 10

#AllLivesMatter #BlackLivesMatter

FIG. 3. Jensen-Shannon divergence word shift graph forthe week following the death of two NYPD police officers.See main text and the caption of Fig. 2 for an explana-tion of the word shift graphs. The diversity of “merica”appears to be high, however it is almost exclusively usedin one popular retweet. This high diversity is due to al-terations of the original retweet that added comments.Since these less popular, modified retweets contain dif-ferent language from one another, and the original tweethas no other words used in it, the diversity surrounding“merica” is artificially elevated. So, although uncommon,the diversity measure may be difficult to interpret in thecase of popular one-word retweets.

6.0 4.0 2.0 0.0 2.0 4.0Contribution

(% of total JSD = 0.3901 bits)

50

45

40

35

30

25

20

15

10

5

1

Rank

plzamen

legendsyria

seemworld

#jesuischarliefightselmacentersstrengthglory

#syriaperformance

mutesenselessappalled

media#grammmysqueen

brigadedoesnt

muslimspowerprince

tragedyvictims

hatemonths

placetalking

existsave

#fergusonblack6

silenceright

wouldnt#mikebrown

students3

#beyhivebeyonce

#muslimlivesmatterhill

chapel#handsupdontshoot

#grammys#chapelhillshooting

JSD Word ContributionsFebruary 08, 2015 to February 14, 2015

10 5 0 5 10

#AllLivesMatter #BlackLivesMatter

FIG. 4. Jensen-Shannon divergence word shift graph forthe week following the 2015 Grammy Awards and theChapel Hill shooting. See main text and the caption ofFig. 2 for an explanation of the word shift graphs. Thisperiod reflects a time in which usage of both #BlackLives-Matter and #AllLivesMatter diverged due to focus on dif-ferent events. Discussion within #AllLivesMatter reflectsthe Chapel Hill shooting, while discussion within #Black-LivesMatter reflects the 2015 Grammy Awards and theperformances of Beyonce and John Legend that alludedto Black Lives Matter.

6

1.0 0.5 0.0 0.5 1.0 1.5Contribution

(% of total JSD = 0.1225 bits)

50

45

40

35

30

25

20

15

10

5

1

Rank

leftgender

marchpain

#endslavery#govegan

leavealum

#wakeupamerica#whitelivesmatter

#farm365directing

talkedprotesting

squaregonna

someonesimage

barely#rednationrising

imkenya

massacre#blackspring

talkingunion

riotinggiven

147#ccot#tcotrace

wouldnt1960sviral#every28hrs

inspiredapril

#bluelivesmatter

original#policelivesmatter

#prayforbaltimore

#baltimoredismissive

undermines2015

sayinguncalled

#freddiegray#baltimoreuprising

JSD Word ContributionsApril 26, 2015 to May 02, 2015

10 5 0 5 10

#AllLivesMatter #BlackLivesMatter

FIG. 5. Jensen-Shannon divergence word shift graph for theweek encapsulating the peak of the Baltimore protests sur-rounding the death of Freddie Gray. See main text and thecaption of Fig. 2 for an explanation of the word shift graphs.In this period, the conservative alignment of #AllLivesMatteris particularly prevalent, as seen in the hashtags #tcot (TopConservatives on Twitter), #ccot (Conservative Christians onTwitter), and #rednationrising.

relevant deaths of Black Americans. Similar divergencesappear in the other periods as well, as evidenced by theappearance of #ericgarner, #walterscott, and #sandrab-land in the respective Appendix word shift graphs.

During important protest periods, the conversationwithin #AllLivesMatter diversifies itself around the livesof law enforcement officers. As shown in Figure 5, dur-ing the Baltimore protests in which #baltimoreuprisingand #baltimore were used significantly in #BlackLives-Matter, users of #AllLivesMatter responded with diverseusage of #policelivesmatter and #bluelivesmatter. Sim-ilarly, Figure 3 shows that upon the death of the twoNYPD officers, words such as “officers,” “ramos,” “liu,”and “prayers” appeared in a variety of #AllLivesMattertweets. In addition, pro-law enforcement hashtags suchas #policelivesmatter, #nypd, #nypdlivesmatter, and#bluelivesmatter all contribute to the divergence of #Al-lLivesMatter from #BlackLivesMatter. Such divergencecomes at the same time that the hashtags #moa and#blackxmas and words “protest,” “mall,” and “america”were prevalent in #BlackLivesMatter due to Christmasprotests, specifically at the Mall of America in Blooming-ton, Minnesota. So, in the midst of political protests byBlack Lives Matter advocates, we see a law-enforcement-aligned response from #AllLivesMatter.

During the period following the non-indictment of Dar-ren Wilson, there are some words, such as “oppression,”“structural,” and “brutality,” that seem to suggest en-gagement from #AllLivesMatter with the issues beingdiscussed within #BlackLivesMatter, such as structuralracism and police brutality. Since the diversities of thesewords are low, we can inspect popular retweets contain-ing these words to understand how they were used. Do-ing so, we find that the words actually emerge in #Al-lLivesMatter due to hijacking [36], the adoption of ahashtag to mock or criticize it. That is, these wordsappear not because of discussion of structural oppres-sion and police brutality by #AllLivesMatter advocates,but because #BlackLivesMatter supporters are specifi-cally critiquing the fact such discussions are not occur-ring within #AllLivesMatter. (We have chosen to notprovide direct references to these tweets so as to protectthe identity of the original tweeter.) Similarly, a “3-panelcomic” strip criticizing the notion of “All Lives Matter”circulated through #AllLivesMatter following the deathof Eric Garner (Figure A1), and after the Chapel Hill andCharleston Church shootings, #BlackLivesMatter propo-nents leveraged #AllLivesMatter to question why believ-ers of the phrase were not more vocal (Figures 4 and A3).We note that we are able to pick up on these instances ofhijacking by inspecting words with high divergence, butlow diversity (meaning the divergence comes almost en-tirely from the few retweets containing the word). Thishijacking drives the divergence of #AllLivesMatter from#BlackLivesMatter in many of these periods.

7

#BlackLivesMatter Nodes % Original Nodes Edges Clustering

Nov. 24–Nov. 30, 2014 243 7.39% 467 0.0605Dec. 3–Dec. 9, 2014 339 5.98% 794 0.0691Dec. 20–Dec. 26, 2014 187 5.96% 391 0.1635Feb. 8–Feb. 14, 2015 70 4.75% 94 0.1740Apr. 4–Apr. 10, 2015 80 4.12% 114 0.1019Apr. 26–May 2, 2015 234 5.54% 471 0.1068Jun. 17–Jun. 23, 2015 167 6.35% 246 0.0746Jul. 21–Jul. 27, 2015 216 6.18% 393 0.0914#AllLivesMatter

Nov. 24–Nov. 30, 2014 26 5.76% 35 0.1209Dec. 3–Dec. 9, 2014 31 3.92% 49 0.2667Dec. 20–Dec. 26, 2014 41 3.95% 70 0.2910Feb. 8–Feb. 14, 2015 18 3.50% 23 0.1894Apr. 4–Apr. 10, 2015 7 1.88% 6 0.0000Apr. 26–May 2, 2015 38 4.12% 62 0.1868Jun. 17–Jun. 23, 2015 22 4.56% 28 0.3571Jul. 21–Jul. 27, 2015 33 4.26% 44 0.1209

TABLE I. Summary statistics for topic networks created fromthe full hashtag networks using the disparity filter at the sig-nificance level α = 0.03 [53].

B. Topic Networks

Having analyzed the micro-level dynamics of word us-age within #BlackLivesMatter and #AllLivesMatter, weturn to focus on the broader topics of these movementsand how these topics coincide with the word-level anal-ysis. Previous work on political polarization has usedhashtags as a proxy for topics [30, 32, 34, 51, 52] and herewe use the same interpretation. However, not all hash-tags assist in understanding the broad topics. For exam-ple, #retweet and #lol are two such hashtags that fre-quently appear in tweets, but they provide no evidentlyrelevant information about the events that are being dis-cussed. Thus, we require a way of uncovering the mostimportant topics and how they connect to one another.

To find these topics, we first construct hashtag net-works for each of #BlackLivesMatter and #AllLivesMat-ter, where nodes are hashtags and weighted edges denoteco-occurrence of these hashtags. To extract the coreof this hashtag network, we apply the disparity filter,a method introduced by Serrano et al. that yields the“multiscale backbone” of a weighted network [53]. Thisbackbone extracts statistically significant edges com-pared to a null model where all weights are uniformlydistributed. We take the topic network to be the largestconnected component of the backbone. For significancelevel α < 0.03, the disparity filter begins to force drasticdrops in the number of nodes removed from the originalhashtag network, as shown in Figure B1. For this reason,we analyze the backbones only for α ≥ 0.03.

Visualizations of the topic networks following the weekof the death of the two NYPD officers are presented inFigures 6 and 7, and networks of the remaining periodsare shown in the Appendix. Node sizes are proportionalto the square root of the number of times each hashtagwas used, and node colors are determined by the Louvainstructure detection method [54]. The exact assignmentsof topics to communities is not critical, but rather theyprovide a visual guide through the networks.

Table I and Appendix Tables B1 and B2 report thenumber of nodes, edges, clustering coefficients, and per-centages of nodes maintained from the full hashtag net-works per each significance level α = 0.03, 0.04, and 0.05respectively. We see that across all time periods of in-terest and significance levels, the number of topics in#BlackLivesMatter is higher than that of #AllLivesMat-ter. We also note that the clustering of the #BlackLives-Matter topics is less that of #AllLivesMatter almost al-ways. Thus, not only are there more topics presented in#BlackLivesMatter, but they are more diverse in theirconnections. In contrast, the stronger #AllLivesMatterties, as measured by their clustering, suggest that the#AllLivesMatter topics are more tightly connected andrevolve around similar themes. We see the clusteringis less within #AllLivesMatter during the week of Wal-ter Scott’s death, where the topic network has a star-likeshape with no triadic closure across all significance levels.This low clustering is not indicative of diverse conversa-tion, as the central node #BlackLivesMatter connectsseveral disparate topics. This is in line with our word-level analysis, where we concluded from Figure A2 thatthe discussion within #AllLivesMatter was dominated bya retweet not pertaining to the death of Walter Scott, theevent of that time period.

In order to extract the most central topics of #Black-LivesMatter and #AllLivesMatter during each time pe-riod, we compare the results of three centrality measures,betweenness centrality [55], random walk betweeness cen-trality [56], and PageRank [57] on the topic networksat significance level α = 0.03. Through inspection ofthe rankings of each list, we find the relative rankings ofthe most central topics in #BlackLivesMatter and #Al-lLivesMatter are robust to the centrality measure used.Table II shows the rankings according to random walkcentrality. Rankings from betweenness centrality andPageRank are reported in Appendix Tables B3 and B4.

We see that for both #BlackLivesMatter and #All-LivesMatter, the top identified topics are indicative ofthe relative events occurring in each time period. In#BlackLivesMatter for instance, #ferguson and #mike-brown are top topics after the non-indictment of DarrenWilson, #walterscott is a top topic after the death ofWalter Scott, and #sandrabland and #sayhername area top topics during the time period following the death ofSandra Bland. However, these major topics rank differ-ently in both #BlackLivesMatter and #AllLivesMatter.For instance, while #mikebrown, #ericgarner, #icant-breathe, #freddiegray, #baltimore, and #sandrablandall consistently rank higher in #BlackLivesMatter thanin #AllLivesMatter. The significant presence of thesehashtags within #BlackLivesMatter is consist with ourfindings from the JSD word shift graphs.

The most prominent discussion of non-Black lives inthe topic networks of #AllLivesMatter is discussion ofpolice lives. We see that in #AllLivesMatter, #nypd,#policelivesmatter, and #bluelivesmatter are rankedhigher as topics in #AllLivesMatter than in #Black-

8

houston

justiceforall

michaelbrown

oaklandprotests

alivewhileblack

tamirrice

asianlivesmatter

protest

poltwt

tgdn

thinkmoor

oakland

milwaukee

prince

policethepolice

mallofamerica

police

lajustice

achristmasstory

minnesota

merrychristmasinheaven

policestate

black

coplivesmatter

blackchristmas

justiceforericgarner

silentnight

selma

shutitdown

every28hours

ferguson

anonymous

wakeup

trayvonmartin

moa

handsupdon

ccot

christiantwitter

dayofpurpose

torturereport

christmas

2014in5words

christmaseve

policelivesmatter

millionsmarchnycwecantbreathe

xmas

sotfbrutality

akaigurley

antoniomartin

blacktwitterfoxnews

brownlivesmatter

mmla traintakeover

cleveland

fergusondecision

ftp

chargemetoodontrehamilton

jailkillercops

chicago

cancelchristmas

icantbreathe

missing

nojusticenowings

racisminamerica

abortjesus

deblasioresignnow

cnnismaaiylbrinsley

hispaniclivesmatter

blacklivesmpls

losangeles

tcot

policeshootings

newyork

tanishaanderson

jordanbaker

tlot

riptrayvonmartin

lgbt

ericgarner

berkeley

nativelivesmatter

thismustend

mynypd

nypdlivesmatter

nojusticenopeace

uppers

millionsmarch

farrakhan

uniteblue

johncrawford

fuck12

deblasio

thisstopstoday

america

ripericgarner

whiteprivilege

blackxmas

icantsleep

alllivesmatter

ows

acab

womenslivesmatter

mke

mikebrown

libcrib

usa

racism

indictthesystem

princess

occupy

racist

teaparty

dontehamiltonnyc2palestine

ferguson2nyc

stoptheviolence

shutdownptree

stolenlives

otg

cops

dontshoot

dcferguson

rip

anarquistaslibertad

atlanta

nypd

stlouis

pjnet

obama

p2topprog

iris_messenger

isis

itstopstoday

oaklandprotest

stl

mlk

bluelivesmatter

justice4dontre

crimingwhilewhite

whitelivesmatter

every28

shutitdownatl

shutdown5thave

wewillnotbesilenced

ripmikebrown

icantbreath

policebrutality

minneapolis

wearentsafe

idlenomore

tcshutitdown

seattle

phillyblackout

prolife

blacklivesmatterminneapolis

amerikkka

dontshootpdx

justiceforeric

ferguson2cle

icanbreathe

nyc

millionsmarchla

handsup

ny

wufam

handsupdontshoot

justiceformikebrown

brooklyn

blacktruthmatters

unionstation

sadday

copslivesmatter

blacklivesmattermoa

winterofresistance

antonio

justice4all

dayton

missouri

harlem

nypdshooting

ripantoniomartin

FIG. 6. #BlackLivesMatter topic network for the week following the death of two NYPD officers.

nypdlivesmattermichaelbrown

rafaelramos

icantbreath

deblasio

rip

brownlivesmatterantoniomartin

nba

handsupdontshoot

racebaiters

mikebrown

police

icantbreathe

merrychristmas

coplivesmatter

wenjianliu

nyc

mayorresign

stoptheviolence

tcot

nypdstrong

stolenlives

holdermustgo

brooklyn

shutitdown

fergusonnotallcopsarebad

nypd

nativelivesmatter

p2

whitelivesmatter

copslivesmatterpolicelivesmatter

bluelivesmatter

alsharpton

nojusticenopeace

ericgarnerthinblueline

nypdshooting

blacklivesmatter

FIG. 7. #AllLivesMatter topic network for the week following the death of two NYPD officers.

9

#BlackLivesMatter #AllLivesMatterTop Hashtags Betweenness Top Hashtags Betweenness

Nov. 24–Nov. 30, 2014

1. ferguson2. mikebrown3. fergusondecision4. shutitdown5. justiceformikebrown6. tamirrice7. blackoutblackfriday8. alllivesmatter9. boycottblackfriday10. blackfriday

0.89690.23030.11590.08540.06520.05830.05770.05120.04420.0439

1. blacklivesmatter2. ferguson3. fergusondecision4. mikebrown5. nycprotest6. williamsburg7. brownlivesmatter8. sf9. blackfridayblackout10. nyc

0.72290.60790.17910.17300.03030.01920.01580.01500.01370.0123

Dec. 3–Dec. 9, 2014

1. ericgarner2. icantbreathe3. ferguson4. shutitdown5. mikebrown6. thisstopstoday7. handsupdontshoot8. nojusticenopeace9. nypd10. berkeley

0.62210.50350.28230.11170.07450.05050.04690.04490.03990.0352

1. blacklivesmatter2. ericgarner3. ferguson4. icantbreathe5. tcot6. shutitdown7. handsupdontshoot8. crimingwhilewhite9. miami10. mikebrown

0.65890.43960.36840.27430.19160.15290.07970.06660.06660.0645

Dec. 20–Dec. 26, 2014

1. icantbreathe2. ferguson3. antoniomartin4. shutitdown5. nypd6. alllivesmatter7. ericgarner8. nypdlivesmatter9. tcot10. handsupdontshoot

0.57420.32340.28260.24730.14960.13240.12650.11470.10820.0986

1. blacklivesmatter2. nypd3. policelivesmatter4. nypdlivesmatter5. ericgarner6. nyc7. bluelivesmatter8. icantbreathe9. shutitdown10. mikebrown

0.67910.44860.29260.19900.15650.14610.14090.12540.09920.0860

Feb. 8–Feb. 14, 2015

1. blackhistorymonth2. grammys3. muslimlivesmatter4. bhm5. alllivesmatter6. handsupdontshoot7. mikebrown8. ferguson9. icantbreathe10. beyhive

0.65060.56140.55150.49870.49320.41500.25880.17540.16140.1325

1. muslimlivesmatter2. blacklivesmatter3. chapelhillshooting4. jewishlivesmatter5. butinacosmicsensenothingreallymatters6. whitelivesmatter7. rip8. hatecrime9. ourthreewinners

0.80120.42960.41920.22790.04310.01120.00730.00710.0058

Apr. 4–Apr. 10, 2015

1. walterscott2. blacktwitter3. icantbreathe4. ferguson5. p26. ericgarner7. alllivesmatter8. mlk9. kendrickjohnson10. tcot

0.91180.22830.17790.15550.10220.10030.09060.07170.06850.0617

1. blacklivesmatter 1.0000

Apr. 26–May 2, 2015

1. freddiegray2. baltimore3. baltimoreuprising4. baltimoreriots5. alllivesmatter6. mayday7. blackspring8. tcot9. baltimoreuprising10. handsupdontshoot

0.58060.47320.26250.24150.10530.09700.07370.05810.05000.0458

1. blacklivesmatter2. baltimoreriots3. baltimore4. freddiegray5. policelivesmatter6. baltimoreuprising7. tcot8. peace9. whitelivesmatter10. wakeupamerica

0.72270.43390.34630.18690.11060.10140.06630.04510.04440.0281

Jun. 17–Jun. 23, 2015

1. charlestonshooting2. charleston3. blacktwitter4. tcot5. unitedblue6. racism7. ferguson8. usa9. takedowntheflag10. baltimore

0.88490.25510.14890.13790.13400.13230.10850.10110.07900.0788

1. charlestonshooting2. blacklivesmatter3. bluelivesmatter4. gunsense5. pjnet6. 2a7. wakeupamerica8. tcot9. gohomederay10. ferguson

0.66660.62380.49000.25710.22920.18570.14940.12890.09520.0952

Jul. 21–Jul. 27, 2015

1. sandrabland2. sayhername3. justiceforsandrabland4. unitedblue5. blacktwitter6. alllivesmatter7. tcot8. defundpp9. p210. m4bl

0.78020.31750.19940.18700.17880.16480.10810.08270.07560.0734

1. blacklivesmatter2. pjnet3. tcot4. uniteblue5. defundplannedparenthood6. defundpp7. sandrabland8. justiceforsandrabland9. prolife10. nn15

0.84040.26890.26830.24400.24370.16920.13860.13860.08810.0625

TABLE II. The top 10 hashtags in the topic networks as determined by random walk betweeness centrality for each time period.Some #AllLivesMatter topic networks have less than 10 top nodes due to the relatively small size of the networks.

10

800

1000

1200

1400

1600

1800Eff

ect

ive L

exic

al D

ivers

ity

Effective Lexical Diversity Subsamples#BlackLivesMatter #AllLivesMatter

Nov 2014

Dec Jan 2015

Feb Mar Apr May Jun Jul Aug

Month

0

50

100

150

200

250

300

350

400

450

Eff

ect

ive H

ash

tag D

ivers

ity

Effective Hashtag Diversity Subsamples

FIG. 8. To control for how volume affects the effective diversity of #BlackLivesMatter and #AllLivesMatter, we break thetime scale down into months and subsample 2,000 tweets from each hashtag 1,000 times. The plot shows notched box plotsdepicting the distributions of these subsamples for effective lexical and hashtag diversity. The notches are small on all theboxes, indicating that the mean diversities are significantly different at the 95% confidence level across all time periods.

LivesMatter during December 20th and April 26th pe-riods, similar to what we found in the JSD word shiftgraphs. On the other hand, hashtags depicting stronganti-police sentiment such as #killercops, #policestate,and #fuckthepolice appear almost exclusively in #Black-LivesMatter and are absent from #AllLivesMatter. Thealignment of #AllLivesMatter with police lives coincideswith a broader alignment with the conservative sphereof Twitter that is apparent through the topic networks.In several periods for #AllLivesMatter, #tcot is a cen-tral topic, as well as #pjnet (Patriots Journal Network),#wakeupamerica, and #defundplannedparenthood. Thehashtag #tcot also appears in several of the #BlackLives-Matter periods as well. This is to be expected, as Freelonet al. found that a portion of #BlackLivesMatter tweetswere hijacked by the conservative sphere of Twitter [20].

However, the hijacking of #BlackLivesMatter and con-tent injection of conservative views is a much smallercomponent of the #BlackLivesMatter topics as comparedto the respective hijacking of the #AllLivesMatter top-ics. As evidenced both by the network statistics andthe network visualizations themselves, the #BlackLives-Matter topic networks show that the conversations arediverse and multifaceted while the #AllLivesMatter net-works show conversations that are more limited in scope.Furthermore, #BlackLivesMatter is consistently a morecentral topic within the #AllLivesMatter networks than#AllLivesMatter is within the #BlackLivesMatter net-works. Thus, hijacking is more prevalent within #All-LivesMatter, while #BlackLivesMatter users are able tomaintain diverse conversations and delegate hijacking to

only a portion of the discourse.

C. Conversational Diversity

Having quantified both the word-level divergences andthe large-scale topic networks, we now measure the in-formational diversity of #BlackLivesMatter and #All-LivesMatter more precisely. We do this through twoapproaches. First, we measure “lexical diversity,” thediversity of words other than hashtags. Second, we mea-sure the hashtag diversity. We measure these diversitiesusing the effective diversity described in Eqn. 2 in Sec-tion II B. Furthermore, to account for the different vol-ume of #BlackLivesMatter and #AllLivesMatter tweets,we break the time scale down into months and subsample2,000 tweets from each hashtag 1,000 times, calculatingthe effective diversities each time. The results are shownin Figure 8.

The lexical diversity of #BlackLivesMatter is largerthan #AllLivesMatter in eight of the ten months withan average lexical diversity that is 5% more than thatof #AllLivesMatter. Interestingly, the two cases where#AllLivesMatter has higher lexical diversity are in thetwo periods when there were large non-police-involvedshootings of people of color and #AllLivesMatter wasused as a hashtag of solidarity. However, the more strik-ing differences are in terms of hashtag diversity. On av-erage, the hashtag diversity of #BlackLivesMatter is sixtimes that of #AllLivesMatter. This is in line with ournetwork analysis where we found expansive #BlackLives-

11

Matter topic networks and tightly clustered, less diverse#AllLivesMatter topic networks.

The low hashtag diversity of #AllLivesMatter is rela-tively constant. One could imagine that the lack of di-versity in the topics of #AllLivesMatter is a result of afocused conversation that does not deviate from its mainmessage. However, as we have demonstrated through theJSD word shift graphs and topic networks, the conver-sation of #AllLivesMatter does change and evolve withrespect to the different time periods. We see mentionsof the major deaths and events of these periods within#AllLivesMatter, even if they do not rank as highly interms of topic centrality. So, even though both protesthashtags have overlap on many of the major topics, thediversity of topics found within #BlackLivesMatter farexceeds that of #AllLivesMatter, even when accountingfor volume.

IV. DISCUSSION.

Through a multi-level analysis of #BlackLivesMatterand #AllLivesMatter tweets, ranging from the word levelto the topic level, we have demonstrated key differencesbetween the two protest groups. Although one of thesedifferences has been proportionally higher discussion ofBlack deaths in #BlackLivesMatter, it is important tonote that #AllLivesMatter is not completely devoid ofdiscussion about these deaths. For instance, #riper-icgarner is prominent within #AllLivesMatter follow-ing the death of Eric Garner, #iamame (“I am AfricanMethodist Episcopal”) contributes more to #AllLives-Matter following the Charleston Church shooting, andthe names of several Black Americans appear in the #Al-lLivesMatter topic networks. However, many of thesesigns of solidarity are associated with low diversity. Inlight of this, it is also important to note that there isa lack of discussion of other deaths within #AllLives-Matter. That is, in examining several of the main peri-ods where #AllLivesMatter spikes, only the Chapel Hillshooting period (Figures 4 and B7) shows discussion ofnon-Black deaths.

The word shift graphs and topic networks reveal thatthe only other lives that are significantly discussed within#AllLivesMatter are the lives of law enforcement officers,particularly during times in which there is heavy protest-ing. Although the notions of “Black Lives Matter” and“Police Lives Matter” are not necessarily mutually exclu-sive [29], we see that the conversations within #AllLives-Matter often frame Black protesters versus law enforce-ment with an“us versus them” mentality. This framingechoes the ways in which media outlets have historicallyframed the tension between Black protesters and law en-forcement [58–60], where police officers and protesters areseen as “enemy combatants” [28] and such movements

appear to “jeopardize law enforcement lives” [29]. So,by facilitating this opposition, #AllLivesMatter becomesthe center of upholding historically contentious views inthe midst of what some consider the New Civil RightsMovement.

In line with Freelon et al.’s findings, we have uncoveredhijacking of #BlackLivesMatter by #AllLivesMatter ad-vocates [20]. Such hijacking is similar to the contentinjection described by Conover et al. [30], where onegroup adopts the hashtag of politically opposed group inorder to inject their ideological beliefs. Such content in-jection has also between found in the work of Egyptianpolitical polarization [34]. However, a significant por-tion #AllLivesMatter hijacking by #BlackLivesMattersupporters is not simple content injection. Rather, advo-cates of #BlackLivesMatter often use #AllLivesMatterto directly interrogate the stance of “All Lives Matter”and the worldview implied by that phrase. Furthermore,such discussions have largely been relegated to #All-LivesMatter, allowing #BlackLivesMatter to exhibit di-verse conversations about a variety of topics. Althoughpast research has expressed concern that #AllLivesMat-ter would derail from the movement started by #Black-LivesMatter [26, 28, 29, 61], our data-driven approachhas allowed us to uncover that #BlackLivesMatter hascountered #AllLivesMatter content injection.

We were largely able to unpack these narratives bylooking not just as hashtag trends and curated term lists,as in previous studies, but at the entirety of the text andhow it is structured at both the word and topic levels. Inthis sense, our work falls closer to that of researchers whohave constructed networks of these protest movements toinform largely qualitative research [5, 21, 36], rather thanresearch that focuses solely on the topological propertiesof the networks. Our methods generalize to studyinghow the discourse of two or more protest movements di-verge from one another and how these differences can becaptured at the word and topic levels. Furthermore, byintegrating diversities within our divergence analysis, wehave shown how we can measure word-level differencesbetween protests and counter-protests in such a way thatfacilitates follow-up qualitative investigation of how ex-actly those words contribute to the divergence. Overall,these tools provide an avenue for exploring more of thedata narratives between #BlackLivesMatter and #All-LivesMatter, and other instances of political polarization.

ACKNOWLEDGEMENTS

All authors would like to thank the members of theComputational Story Lab for their feedback and sup-port. They would also like to thank James Bagrow for hishelpful suggestions. PSD and CMD acknowledge supportfrom NSF Big Data Grant #1447634.

12

[1] M. D. Conover, E. Ferrara, F. Menczer, and A. Flammini.The digital evolution of Occupy Wall Street. PloS ONE,8(5):e64679, 2013.

[2] N. Caren and S. Gaby. Occupy online: Facebook andthe spread of Occupy Wall Street. Available at SSRN1943168, 2011.

[3] K. M. DeLuca, S. Lawson, and Y. Sun. Occupy WallStreet on the public screens of social media: The manyframings of the birth of a protest movement. Communi-cation, Culture & Critique, 5(4):483–509, 2012.

[4] P. N. Howard, A. Duffy, D. Freelon, M. M. Hussain,W. Mari, and M. Mazaid. Opening closed regimes: Whatwas the role of social media during the Arab Spring?Available at SSRN 2595096, 2011.

[5] Z. Papacharissi and M. de Fatima Oliveira. Affectivenews and networked publics: The rhythms of newsstorytelling on #Egypt. Journal of Communication,62(2):266–282, 2012.

[6] E. Anduiza, C. Cristancho, and J. M. Sabucedo. Mo-bilization through online social networks: The politicalprotest of the Indignados in Spain. Information, Com-munication & Society, 17(6):750–764, 2014.

[7] S. Gonzalez-Bailon and N. Wang. Networked discontent:The anatomy of protest campaigns in social media. Socialnetworks, 44:95–104, 2016.

[8] M. D. Conover, C. Davis, E. Ferrara, K. McKelvey,F. Menczer, and A. Flammini. The geospatial charac-teristics of a social movement communication network.PLoS ONE, 8(3):e55957, 2013.

[9] S. Gonzalez-Bailon, J. Borge-Holthoefer, A. Rivero, andY. Moreno. The dynamics of protest recruitment throughan online network. Scientific reports, 1(197), 2011.

[10] Z. C. Steinert-Threlkeld, D. Mocanu, A. Vespignani, andJ. Fowler. Online social networks and offline protest. EPJData Science, 4(1):1–9, 2015.

[11] J. Larson, J. Nagler, J. Ronen, and J. Tucker. Socialnetworks and protest participation: Evidence from 93million twitter users. In Political Networks Workshops &Conference, 2016.

[12] H. S. Christensen. Political activities on the Internet:Slacktivism or political participation by other means?First Monday, 16(2), 2011.

[13] P. Barbera, N. Wang, R. Bonneau, J. T. Jost, J. Na-gler, J. Tucker, and S. Gonzalez-Bailon. The criticalperiphery in the growth of social protests. PLoS ONE,10(11):e0143611, 2015.

[14] J. Borge-Holthoefer, N. Perra, B. Goncalves,S. Gonzalez-Bailon, A. Arenas, Y. Moreno, andA. Vespignani. The dynamics of information-drivencoordination phenomena: A transfer entropy analysis.Science Advances, 2(4), 2016.

[15] H. Qi, P. Manrique, D. Johnson, E. Restrepo, and N. F.Johnson. Open source data reveals connection betweenonline and on-street protest activity. EPJ Data Science,5(1):1, 2016.

[16] A. Garza. A herstory of the #blacklivesmat-ter movement. The Feminist Wire (2014).http://www.thefeministwire.com/2014/10/blacklivesmatter-2/. Accessed 20 June 2016.

[17] F. C. Harris. The next civil rights movement? Dissent,62(3):34–40, 2015.

[18] J. C. Kang. our demand is simple: Stop killingus: How a group of Black social media ac-tivists built the nations first 21st-century civilrights movement. New York Times (2015).http://www.nytimes.com/2015/05/10/magazine/our-demand-is-simple-stop-killing-us.html. Accessed 20 June2016.

[19] O. Tometi and G. Lenoir. Black Lives Mat-ter is not a civil rights movement. Time (2015).http://time.com/4144655/international-human-rights-day-black-lives-matter/. Accessed 20 June 2016.

[20] D. G. Freelon, C. D. McIlwain, and M. D. Clark. Be-yond the hashtags: #ferguson, #blacklivesmatter, andthe online struggle for offline justice. Available at SSRN,2016.

[21] S. J. Jackson and B. Foucault Welles. #ferguson is every-where: Initiators in emerging counterpublic networks. In-formation, Communication & Society, pages 1–22, 2015.

[22] Y. Bonilla and J. Rosa. #ferguson: Digital protest, hash-tag ethnography, and the racial politics of social mediain the united states. American Ethnologist, 42(1):4–17,2015.

[23] A. Olteanu, I. Weber, and D. Gatica-Perez. Charac-terizing the demographics behind the #blacklivesmattermovement. arXiv preprint arXiv:1512.05671, 2015.

[24] D. Freelon, C. McIlwain, and M. Clark. Quantifying thepower and consequences of social media protest. NewMedia & Society, page 1461444816676646, 2016.

[25] M. Twyman, B. C. Keegan, and A. Shaw. Black livesmatter in Wikipedia: Collaboration and collective mem-ory around online social movements. arXiv preprintarXiv:1611.01257, 2016.

[26] M. Orbe. #alllivesmatter as post-racial rhetorical strat-egy. Journal of Contemporary Rhetoric, 5(3/4):90–98,2015.

[27] T. Husband Jr. i dont see color: Challenging assump-tions about discussing race with young children. EarlyChildhood Education Journal, 39(6):365–371, 2012.

[28] R. Rickford. Black lives matter toward a modern practiceof mass struggle. In New Labor Forum, volume 25, pages34–42. SAGE Publications, 2016.

[29] C. L. Langford and M. Speight. #blacklivesmatter: Epis-temic positioning, challenges, and possibilities. Journalof Contemporary Rhetoric, 5(3/4):78–89, 2015.

[30] M. Conover, J. Ratkiewicz, M. R. Francisco,B. Goncalves, F. Menczer, and A. Flammini. Po-litical polarization on Twitter. ICWSM, 133:89–96,2011.

[31] M. D. Conover, B. Goncalves, A. Flammini, andF. Menczer. Partisan asymmetries in online political ac-tivity. EPJ Data Science, 1(1):1–19, 2012.

[32] L. Bode, A. Hanna, J. Yang, and D. V. Shah. Candidatenetworks, citizen clusters, and political expression strate-gic hashtag use in the 2010 midterms. The ANNALS ofthe American Academy of Political and Social Science,659(1):149–165, 2015.

[33] M. Tremayne. Anatomy of protest in the digital era: Anetwork analysis of twitter and occupy wall street. SocialMovement Studies, 13(1):110–126, 2014.

[34] J. Borge-Holthoefer, W. Magdy, K. Darwish, and I. We-ber. Content and network dynamics behind egyptian

13

political polarization on twitter. In Proceedings of the18th ACM Conference on Computer Supported Cooper-ative Work & Social Computing, pages 700–711. ACM,2015.

[35] I. Weber, V. R. K. Garimella, and A. Batayneh. Secularvs. islamist polarization in egypt on twitter. In Proceed-ings of the 2013 IEEE/ACM International Conference onAdvances in Social Networks Analysis and Mining, pages290–297. ACM, 2013.

[36] S. J. Jackson and B. Foucault Welles. Hijacking#myNYPD: Social media dissent and networked coun-terpublics. Journal of Communication, 65(6):932–952,2015.

[37] E. McClam. Ferguson cop Darren Wilson not indictedin shooting of Michael Brown. NBC News (2014).http://www.nbcnews.com/storyline/michael-brown-shooting/ferguson-cop-darren-wilson-not-indicted-shooting-michael-brown-n255391. Accessed 20 June2016.

[38] R. Sanchez and S. Prokupecz. Protests after N.Y. cop notindicted in chokehold death; Feds reviewing case. CNN(2014). http://www.cnn.com/2014/12/03/justice/new-york-grand-jury-chokehold/. Accessed 20 June 2016.

[39] B. Muller and A. Baker. 2 N.Y.P.D. of-ficers killed in Brooklyn ambush; Suspectcommits suicide. New York Times (2014).http://www.nytimes.com/2014/12/21/nyregion/two-police-officers-shot-in-their-patrol-car-in-brooklyn.html.Accessed 20 June 2016.

[40] S. Ahmed and C. E. Shoichet. 3 students shot todeath in apartment near UNC Chapel Hill. CNN(2015). http://www.cnn.com/2015/02/11/us/chapel-hill-shooting/. Accessed 20 June 2016.

[41] E. Ramirez. Grammys 2015: PharrellWilliams, Beyonce, Prince pay tribute to BlackLives Matter movement. Billboard (2015).http://www.billboard.com/articles/events/grammys-2015/6465687/grammys-2015-pharrell-williams-beyonce-prince-black-lives-matter-hands-up-dont-shoot. Accessed20 June 2016.

[42] C. Elmore and D. MacDougall. Man shot and killedby North Charleston police officer after traffic stop;sled investigating. The Post and Courier (2015).http://www.postandcourier.com/article/20150404/PC16/150409635/1180/. Accessed 2016/06/20.

[43] D. A. Graham. The mysterious deathof Freddie Gray. The Atlantic (2015).http://www.theatlantic.com/politics/archive/2015/04/the-mysterious-death-of-freddie-gray/391119/. Accessed 20June 2016.

[44] T. Berenson. Everything we know aboutthe Charleston shooting. Time (2015).http://time.com/3926112/charleston-shooting-latest/.Accessed 20 June 2016.

[45] J. Schuppe. The death of Sandra Bland:What we know so far. NBC News (2015).http://www.nbcnews.com/news/us-news/death-sandra-bland-what-we-know-so-far-n396036. Accessed 20 June2016.

[46] L. Jost. Entropy and diversity. Oikos, 113(2):363–375,2006.

[47] J. Lin. Divergence measures based on the shannonentropy. Information Theory, IEEE Transactions on,37(1):145–151, 1991.

[48] E. A. Pechenick, C. M. Danforth, and P. S. Dodds. Char-acterizing the Google Books corpus: strong limits to in-ferences of socio-cultural and linguistic evolution. PLoSONE, 10(10):e0137041, 2015.

[49] E. A. Pechenick, C. M. Danforth, and P. S. Dodds. Islanguage evolution grinding to a halt?: Exploring thelife and death of words in english fiction. arXiv preprintarXiv:1503.03512, 2015.

[50] K. W. Boyack, D. Newman, R. J. Duhon, R. Klavans,M. Patek, J. R. Biberstine, B. Schijvenaars, A. Skupin,N. Ma, and K. Borner. Clustering more than two mil-lion biomedical publications: Comparing the accuraciesof nine text-based similarity approaches. PLoS ONE,6(3):e18029, 2011.

[51] M. A. Smith, L. Rainie, B. Shneiderman, and I. Himel-boim. Mapping Twitter topic networks: From polarizedcrowds to community clusters. Pew Research Center, 20,2014.

[52] D. M. Romero, B. Meeder, and J. Kleinberg. Differ-ences in the mechanics of information diffusion acrosstopics: Idioms, political hashtags, and complex conta-gion on Twitter. In Proceedings of the 20th InternationalConference on World Wide Web, pages 695–704. ACM,2011.

[53] M. A. Serrano, M. Boguna, and A. Vespignani. Extract-ing the multiscale backbone of complex weighted net-works. Proceedings of the National Academy of Sciences,106(16):6483–6488, 2009.

[54] V. D. Blondel, J.-L. Guillaume, R. Lambiotte, andE. Lefebvre. Fast unfolding of communities in large net-works. Journal of statistical mechanics: theory and ex-periment, 2008(10):P10008, 2008.

[55] M. E. Newman and M. Girvan. Finding and evaluat-ing community structure in networks. Physical review E,69(2):026113, 2004.

[56] M. Girvan and M. E. Newman. Community structurein social and biological networks. Proceedings of the na-tional academy of sciences, 99(12):7821–7826, 2002.

[57] L. Page, S. Brin, R. Motwani, and T. Winograd. Thepagerank citation ranking: Bringing order to the web.Technical report, Stanford InfoLab, 1999.

[58] T. A. Van Dijk. Race, riots and the press an analysisof editorials in the british press about the 1985 disor-ders. International Communication Gazette, 43(3):229–253, 1989.

[59] T. A. Van Dijk. New (s) racism: A discourse analyticalapproach. Ethnic minorities and the media, pages 33–49,2000.

[60] J. A. B. Lackey. Framing Social Protest: How LocalNewspapers Covered the University of Georgia Desegre-gation Protests in January, 1961. 2005.

[61] N. Carney. All lives matter, but so does race black livesmatter and the evolving role of social media. Humanity& Society, 40(2):180–199, 2016.

14

Appendix A: JSD Word Shift Graphs

3.0 2.0 1.0 0.0 1.0 2.0 3.0Contribution

(% of total JSD = 0.0738 bits)

50

45

40

35

30

25

20

15

10

5

1

Rank

partsolidarity

changedont

#whereisjusticeenough

vssphincter

livesnycbreakingshut#thisstopstodayprotest

#rednationrisingpointunity

students#shocking

#shutitdownerasure

diein#policelivesmatter

awaymakes

#seattlegoingfocus

gogot

total#ferguson

missing#nypd

tagequalitystarted

#womblifematterssaying

#icantbreathe#ericgarner

nailsfeed

problemkris

straubsimple3panel

patreoncomic

JSD Word ContributionsDecember 03, 2014 to December 09, 2014

10 5 0 5 10

#AllLivesMatter #BlackLivesMatter

FIG. A1. Jensen-Shannon divergence word shift graph forthe week following the non-indictment of Daniel Pantaleo.See main text and the caption of Fig. 2 for an explanationof the word shift graphs. The leading contributors tothe divergence on the side of #AllLivesMatter are due toa popular retweet discussing a “3-panel” sketch “comic”that describes what the “problem” with #AllLivesMatteris.

8.0 6.0 4.0 2.0 0.0 2.0 4.0Contribution

(% of total JSD = 0.7811 bits)

50

45

40

35

30

25

20

15

10

5

1

Rank

rantwingtimeracistcopsanyoneconvictedsee#ericgarner#trayvonmartinmakes#blackgirlsrockvictoryviaoldchris29dontpauldoesntturnedyrsrelevantampanotherchargedtonightmanmurderbackcoverkilledvideolet#fergusonstillpoliceblackunarmed

#walterscottkenyadidntdead

left147

universitymassacre

happenbarelytalked

JSD Word ContributionsApril 04, 2015 to April 10, 2015

10 5 0 5 10

#AllLivesMatter #BlackLivesMatter

FIG. A2. Jensen-Shannon divergence word shift graphfor the week following the death of Walter Scott. Seemain text and the caption of Fig. 2 for an explanationof the word shift graphs. The leading contributors tothe divergence on the side of #AllLivesMatter are due toa popular retweet that asks why people are not payingattention to a school massacre that occurred in Kenya.This retweet is persistent, which is why the words “kenya”and “massacre” appear in later dates, such as in Fig. 5

.

15

3.0 2.0 1.0 0.0 1.0 2.0Contribution

(% of total JSD = 0.1655 bits)

50

45

40

35

30

25

20

15

10

5

1

Rank

wantedseven

resistwave

usehope

#cdcwhistleblowerterrorism

gives#wewillshootback

redsave

#wakeupamericawheres#pjnet

didntarrestconfederate

failedbleeduseful

flagsdivide

#itsaracething#policelivesmatter

conquer#whiteprivilege

dangerspilled

replyingeasy

#charlestonunitychaingovernments

okall#alllivesmatter

conquersmarxist

unites#racism

#charlestonstrongwhose

#yulindogmeatfestivalheartrendinghorrifying

messagefoolsim

#iamameequalityrooting

JSD Word ContributionsJune 17, 2015 to June 23, 2015

10 5 0 5 10

#AllLivesMatter #BlackLivesMatter

FIG. A3. Jensen-Shannon divergence word shift for theweek following the Charleston church shooting. See maintext and the caption of Fig. 2 for an explanation of theword shift graphs.

4.0 2.0 0.0 2.0 4.0 6.0Contribution

(% of total JSD = 0.1806 bits)

50

45

40

35

30

25

20

15

10

5

1

Rank

facereassured

viralexercising

simpleswear

demandblacknoise

usekilling

amazingjustice#justiceforsandrabland

matterwould

trollinginvestigate

#prolife#whathappenedtosandrablandfolks#doj

picturevs

rathercirculatingimages#blackwomenmatterwrongheartmadeconfirmationknewcoldwtf#justiceformonroebird#monroebird

saywell

definemovie

#sayhername#sandrabland

arentsaying

shootingchurchesschools

theatersyall

JSD Word ContributionsJuly 21, 2015 to July 27, 2015

10 5 0 5 10

#AllLivesMatter #BlackLivesMatter

FIG. A4. Jensen-Shannon divergence word shift for theweek encapsulating outrage over the death of SandraBland. See main text and the caption of Fig. 2 for anexplanation of the word shift graphs.

16

Appendix B: Hashtag Topic Networks

0.00

0.02

0.04

0.06

0.08

0.10

Perc

ent

of

Ori

gin

al H

ash

tag N

etw

ork

Hashtag Network vs. Significance Level - #BlackLivesMatter

α = 0.05 α = 0.04 α = 0.03 α = 0.02 α = 0.01

Dec 2014

Jan 2015

Feb 2015

Mar 2015

Apr 2015

May 2015

Jun 2015

Jul 2015

Aug 2015

Time (Periods of Interest)

0.00

0.02

0.04

0.06

0.08

0.10

Perc

ent

of

Ori

gin

al H

ash

tag N

etw

ork

Hashtag Network vs. Significance Level - #AllLivesMatter

FIG. B1. Percent of original hashtag network maintained for #BlackLivesMatter (top) and #AllLivesMatter (bottom) at eachof the periods of interest for varying levels of the disparity filter significance level. We wish to filter as much of the network aspossible, while avoiding sudden reductions in the number of nodes in the network. Note, when going from α = 0.03 to α = 0.02,the February 8th #BlackLivesMatter and July 21st #AllLivesMatter networks fall in size by a factor of approximately onehalf. Therefore, we choose α = 0.03.

#BlackLivesMatter Nodes % Original Nodes Edges Clustering

Nov. 24–Nov. 30, 2014 270 8.21% 526 0.0598Dec. 3–Dec. 9, 2014 389 6.86% 912 0.0645Dec. 20–Dec. 26, 2014 209 6.66% 439 0.1537Feb. 8–Feb. 14, 2015 76 5.16% 107 0.1744Apr. 4–Apr. 10, 2015 85 4.38% 121 0.0893Apr. 26–May 2, 2015 266 6.30% 547 0.1030Jun. 17–Jun. 23, 2015 176 6.70% 269 0.07690Jul. 21–Jul. 27, 2015 233 6.67% 435 0.0926#AllLivesMatter

Nov. 24–Nov. 30, 2014 29 6.43% 40 0.1200Dec. 3–Dec. 9, 2014 31 3.92% 51 0.2801Dec. 20–Dec. 26, 2014 49 4.72% 86 0.2626Feb. 8–Feb. 14, 2015 24 4.67% 30 0.1521Apr. 4–Apr. 10, 2015 8 2.15% 7 0.0000Apr. 26–May 2, 2015 41 4.45% 71 0.1967Jun. 17–Jun. 23, 2015 23 4.77% 33 0.4909Jul. 21–Jul. 27, 2015 45 5.82% 61 0.1010

TABLE B1. Summary statistics for topic networks created from the full hashtag networks using the disparity filter at thesignificance level α = 0.04 [53].

17

#BlackLivesMatter Nodes % Original Nodes Edges Clustering

Nov. 24–Nov. 30, 2014 298 9.07% 583 0.0590Dec. 3–Dec. 9, 2014 444 7.83% 1025 0.0595Dec. 20–Dec. 26, 2014 227 7.24% 485 0.1494Feb. 8–Feb. 14, 2015 82 5.57% 119 0.1946Apr. 4–Apr. 10, 2015 99 5.10% 139 0.0905Apr. 26–May 2, 2015 294 6.96% 607 0.0976Jun. 17–Jun. 23, 2015 190 7.23% 305 0.0795Jul. 21–Jul. 27, 2015 250 7.15% 475 0.0898#AllLivesMatter

Nov. 24–Nov. 30, 2014 29 6.43% 40 0.1200Dec. 3–Dec. 9, 2014 37 4.68% 62 0.2456Dec. 20–Dec. 26, 2014 54 5.21% 94 0.2500Feb. 8–Feb. 14, 2015 24 4.67% 30 0.1521Apr. 4–Apr. 10, 2015 10 2.69% 9 0.0000Apr. 26–May 2, 2015 44 4.77% 76 0.1944Jun. 17–Jun. 23, 2015 23 4.77% 33 0.4909Jul. 21–Jul. 27, 2015 47 6.08% 66 0.1151

TABLE B2. Summary statistics for topic networks created from the full hashtag networks using the disparity filter at thesignificance level α = 0.05 [53].

18

#BlackLivesMatter #AllLivesMatterTop Hashtags Betweenness Top Hashtags Betweenness

Nov. 24–Nov. 30, 2014

1. ferguson2. mikebrown3. standup4. tamirrice5. truth6. fergusondecision7. arrestdarrenwilson8. justiceformikebrown9. shutitdownatl10. alllivesmatter

0.83560.15840.08450.08350.08040.79350.0716.0.07040.06450.0488

1. blacklivesmatter2. ferguson3. policebrutality4. mikebrown5. fergusondecision

0.61160.45500.17330.15660.1566

Dec. 3–Dec. 9, 2014

1. icantbreathe2. ericgarner3. ferguson4. civilrights5. shutitdown6. lebronjames7. rip8. cantbreathe9. mikebrown10. nojusticenopeace

0.44340.41260.27060.09520.09390.08230.06540.06240.06110.0514

1. blacklivesmatter2. shutitdown3. ferguson4. ericgarner5. policebrutality6. tcot7. icantbreathe8. miami9. crimingwhilewhite10. handsupdontshoot

0.51530.34900.28270.27500.20450.19080.16930.06670.06670.0114

Dec. 20–Dec. 26, 2014

1. icantbreathe2. ferguson3. antoniomartin4. shutitdown5. ny6. handsupdontshoot7. racism8. obama9. mlk10. justice

0.52290.29380.21050.15750.11740.10120.09480.94110.07020.0640

1. nypd2. policelivesmatter3. blacklivesmatter4. icantbreathe5. bluelivesmatter6. ferguson7. handsupdontshoot8. nyc9. nojusticenopeace10. ericgarner

0.48840.33070.30640.24610.20120.18330.16790.14610.09870.0910

Feb. 8–Feb. 14, 2015

1. blackhistorymonth2. grammys3. bhm4. muslimlivesmatter5. alllivesmatter6. bluelivesmatter7. ferguson8. icantbreathe9. ericgarner10. racist

0.65040.55490.49870.49610.47690.39380.32260.27790.18410.1445

1. muslimlivesmatter2. blacklivesmatter3. chapelhillshooting4. jewishlivesmatter5. whitelivesmatter6. ourthreewinners

0.56610.37500.31610.22790.017640.1323

Apr. 4–Apr. 10, 2015

1. walterscott2. blacktwitter3. icantbreathe4. alllivesmatter5. ripwalterscott6. p27. tcot8. kendrickjohnson9. ferguson10. racism

0.80830.23070.15020.14880.14670.10610.06450.06450.05920.0558

1. blacklivesmatter 1.0000

Apr. 26–May 2, 2015

1. freddiegray2. baltimore3. blackpower4. baltimoreriots5. baltimoreuprising6. ferguson7. solidarity8. baltimore9. mayday10. alllivesmatter

0.45340.34410.22580.21200.18010.15180.11590.10020.08420.0833

1. blacklivesmatter2. peace3. baltimoreriots4. baltimore5. justice6. policelivesmatter7. freddiegray8. bluelivesmatter9. baltimoreuprising10. asianlivesmatter

0.65310.31980.25150.22890.13210.11180.09750.07200.05400.0270

Jun. 17–Jun. 23, 2015

1. charlestonshooting2. charleston3. tcot4. confederateflag5. racheldolezal6. blacktwitter7. usa8. baltimore9. love10. staywoke

0.63210.43360.17270.15470.11680.10470.09430.09040.07940.0792

1. charlestonshooting2. blacklivesmatter3. bluelivesmatter4. gunsense5. 2a6. pjnet7. gohomederay8. ferguson9. wakeupamerica10. tcot

0.66670.62380.49680.25710.18570.18090.09520.09520.00150.0015

Jul. 21–Jul. 27, 2015

1. sandrabland2. sayhername3. justiceforsandrabland4. blacktwitter5. alllivesmatter6. wewantanswers7. blackpower8. sandra9. m4bl10. blm

0.70380.19850.15330.13850.11540.11530.11280.08910.07310.0656

1. blacklivesmatter2. tcot3. policebrutality4. uniteblue5. sandrabland6. pjnet7. justiceforsandrabland8. defundplannedparenthood9. wakeupamerica10. whathappenedtosandrabland

0.78420.36890.21770.17940.17540.14110.12090.11890.10280.0625

TABLE B3. The top 10 hashtags in the topic networks as determined by betweenness centrality for each time period. Some#AllLivesMatter topic networks have less than 10 top nodes due to the relatively small size of the networks. For example, inthe period of April 4th, the #AllLivesMatter network consists of #blacklivesmatter as a hub with six edges connecting it tosix other hashtags. Thus, #blacklivesmatter is the only node visited when calculating paths for betweenness.

19

#BlackLivesMatter #AllLivesMatterTop Hashtags PageRank Top Hashtags PageRank

Nov. 24–Nov. 30, 2014

1. ferguson2. mikebrown3. fergusondecision4. shutitdown5. michaelbrown6. blackoutblackfriday7. boycottblackfriday8. justiceformikebrown9. notonedime10. america

0.22990.07780.05040.03430.02580.02380.02030.01970.01760.0161

1. blacklivesmatter2. ferguson3. fergusondecision4. mikebrown5. nycprotest6. williamsburg7. brownlivesmatter8. whitelivesmatter9. sf10. blackfridayblackout

0.26470.23710.05750.05090.03970.03170.02390.02350.02240.0210

Dec. 3–Dec. 9, 2014

1. ericgarner2. icantbreathe3. ferguson4. mikebrown

5. nypd6. shutitdown7. handsupdontshoot8. thisstopstoday9. seattle10. policebrutality

0.19740.15480.08600.02960.02860.02510.01830.01510.01420.0140

1. blacklivesmatter2. icantbreathe3. ericgarner4. ferguson5. tcot6. wecantbreathe7. mikebrown8. handsupdontshoot9. rednationrising10. shutitdown

0.20520.13970.12950.07280.04040.03790.03710.03550.02680.0267

Dec. 20–Dec. 26, 2014

1. icantbreathe2. ferguson3. shutitdown4. antoniomartin5. ericgarner6. nypdlivesmatter7. alllivesmatter8. moa9. nypd10. mikebrown

0.10430.05330.04500.03970.03330.03150.02560.02530.02470.0227

1. blacklivesmatter2. nypdlivesmatter3. nypd4. policelivesmatter5. ericgarner6. mikebrown7. bluelivesmatter8. icantbreathe9. nyc10. stolenlives

0.20420.09550.09450.06080.05690.04710.03280.03100.02910.0278

Feb. 8–Feb. 14, 2015

1. muslimlivesmatter2. alllivesmatter3. handsupdontshoot4. grammys5. mikebrown6. blackhistorymonth7. beyhive8. ferguson9. blacktwitter10. chapelhillshooting

0.14180.07300.06860.05750.04850.04670.04470.03380.02730.0244

1. muslimlivesmatter2. blacklivesmatter3. chapelhillshooting4. butinacosmicsensenothingreallymatters5. jewishlivesmatter6. christianslivesmatter7. buddhistlivesmatter8. whitelivesmatter9. rip10. hatecrime

0.31710.18150.17910.05820.05340.02180.02180.02140.01970.0181

Apr. 4–Apr. 10, 2015

1. walterscott2. ferguson3. ericgarner4. trayvonmartin5. blacktwitter6. icantbreathe7. mikebrown8. ftp9. alllivesmatter10. black

0.21410.08500.07080.06190.04920.03810.02310.02170.01750.0148

1. blacklivesmatter2. walterscott3. muslimlivesmatter4. whitelivesmatter5. icantbreathe6. ripwalterscott7. ifnotnowwhen

0.47100.22530.07050.06670.06290.05160.0516

Apr. 26–May 2, 2015

1. freddiegray2. baltimore3. baltimoreuprising4. baltimoreriots5. alllivesmatter6. ericgarner7. ferguson8. mayday9. blackspring10. michaelbrown

0.13620.11540.08500.07870.01530.01320.01280.01150.01080.0094

1. blacklivesmatter2. baltimoreriots3. baltimore4. freddiegray5. policelivesmatter6. baltimoreuprising7. tcot8. whitelivesmatter9. wakeupamerica10. prayforbaltimore

0.21040.14160.11150.07990.04270.04190.03270.03040.02050.0192

Jun. 17–Jun. 23, 2015

1. charlestonshooting2. racism3. charleston4. whiteprivilege5. blacktwitter6. ferguson7. itsaracething8. usa9. southcarolina10. alllivesmatter

0.16700.03610.03490.02170.02160.02030.01980.01970.01890.0186

1. blacklivesmatter2. iamame3. wakeupamerica4. pjnet5. bluelivesmatter6. tcot7. 2a8. charlestonshooting9. ohhillno10. cosproject

0.20590.14230.06990.06560.06480.06340.06020.03920.03780.0378

Jul. 21–Jul. 27, 2015

1. sandrabland2. sayhername3. justiceforsandrabland4. blackwomenmatter5. doj6. blacktwitter7. unitedblue8. alllivesmatter9. whathappenedtosandrabland10. tcot

0.20130.14490.04220.03520.02380.02010.01690.01630.01590.0132

1. blacklivesmatter2. sandrabland3. pjnet4. defundpp5. justiceforsandrabland6. uniteblue7. sayhername8. defundplannedparenthood9. tcot10. prolife

0.23020.09960.06610.05260.05050.04860.04720.04700.04560.0344

TABLE B4. The top 10 hashtags in the topic networks as determined by PageRank for each time period. Some #AllLivesMattertopic networks have less than 10 top nodes due to the relatively small size of the networks.

20

houston

blackdollarsmatter

justiceforall

michaelbrown

everylifematters

tamirrice

nopeace

ferguson2louisville

blackoutfriday

protest

london

thinkmoor

thanksgiving

repostfuckthepolice

police

prayforferguson

justice

staywoke

riots

oak2stl

policestate

blackfriday

black

dearferguson

justiceformichaelbrownlastwords

texas

mikebrownverdict

toronto

neverforget

umdfergusondc

blacklivesdontmatter

shutitdown

every28hours

ferguson

anonymous

occupystamp

fergusonoakland

trayvonmartin

boycottblackfriday

equality

marissaalexander

unionsquare

shutitdownnyc

michealbrowndetroit

anonymousopferguson

race

latinolivesmatter

truth

pdmfnb

londontoferguson

bartlockdown

blackout

noindictment

revolution

fight4justice

acab

alllivesmatter

brownlivesmatter

shutup

us

cleveland portland

clarkatlanta

equality4all

fergusondecision

brazil

ftp

onelove

rp

chicago

welcometony

ayotzinapa

immigrationaction

yamecanse

injustice

smh

cnn

brownfriday

akaigurley

losangeles

tcot

newyorkwewantjustice

nashville xula

riptrayvonmartin

ericgarner

nojusticeforblacks

fergsuon

opferguson

ohio

humanrights

obamathugs

changeopblackfriday

sanfrancisco

plannedparenthood

pdx

la

notonedime

nojusticenopeace

anon

walmart

demilitarizethepolice

umd

ferguson2oakland

austin

racismkills

atx

fergusonaction

uniteblue

johncrawford

love

streetart

vonderritmyers

spelman

racism

endpolicebrutality

whiteprivilege

soracist

blacktwitter

blackchildren

oakland

arrestdarrenwilson

ows

bart

fergusonphl

grandjury

libcribusa

peacefulprotest

taneshaanderson

onestruggle

indictthesystem

stoptheparade

nyc2ferguson

palestine

mikebrown

asianlivesmatter

blmlaitwasthestate

manchester

furgeson

ripmikebrown

mtvstars

morehouse

nomore

stoppolicebrutality

opdarrienhunt

nycprotest

danielholtzclaw

opkkk

dcferguson

standup

blackboysmatter

atlanta

nypd

stlouis

obama

p2

nonviolence

isis

sandiego

stl

mlk

abortion

shutitdownformikebrown

lmnopi

blackpower

ferguson2to

peaceinferguson

whitelivesmatter

blackfridaydeals

shutitdownatl

fergusiondecision

nojustice

yellowlivesmatter

peace

boston

policebrutality

rip

boycott

act2saveourlives

misuri

nerdland

civilrights

philly

fergusonsolidarity

wilson

seattle

prolife shaw

dallas

protestpeacefully

breaking

ny

indictamerica

auspol

nyc

personalprotestcrashcybermonday

handsup

walmartstrikers

itsbiggerthanyou

fueelestado

evangelicals4justice

whiteonwhitecrime

handsupdontshoot

blackoutblackfriday

blackfridayblackout

amerikkka

reprojustice

justiceformikebrown

fb

indictdarrenwilson

stolenlives

nojusticeformikebrownroxbury

justice4mikebrown

america

chi2ferguson

grandjurydecision

solidarity

fergusonverdict

darrenwilson

brownfridaychi

indictboston

don

endmilitarystylepolicing

fergusonprotest

standwithferguson

dontshoot

handsupdontspend

dayton

missouri

renishamcbrideericgardner

sf

solidaritywithferguson

FIG. B2. #BlackLivesMatter topic network for the week following the non-indictment of Darren Wilson.

michaelbrownpolicebrutality

brownlivesmatter

oakland

fergusondecision

mikebrownprayforferguson

justice

williamsburg

palestine

nyc

personalprotest

akaigurley

ripmikebrown

shutup

blackfridayblackout

nycprotest

justiceformikebrown

shutitdown

blacklivesmatter

trayvonmartin

boycottblackfriday

whiteprivilege

whitelivesmatter

sf

ferguson

FIG. B3. #AllLivesMatter topic network for the week following the non-indictment of Darren Wilson.

21

houston

torturereport

uchicagodiein

justiceforall

michaelbrown

jnu

berkeleyprotests

alivewhileblack

tamirrice

nopeace

asianlivesmatter

decriminalizeblack

protest

livingwhileblack

london

dontshootpdx

thinkmoor

anthros4ferguson

protests

chi2ferguson

wehearyoufuckthepolice

secchampionship

police

racialequality

la

paris

ferguson2miami

fergusonphl

oak2stl

policestate

foleysquare

kcwemustspeak

furgeson

nycprotest

kajiemepowell

lastwords

obama

naacp

texaskendrickjohnson

studentpower

royalvisitusa

lebronjames

justiceforericgarner

indictamerica

delhi

dc indigenous

ican

whereisjustice

shutitdown

timessquare

ferguson

anonymous

eric

trayvonmartin

aiyanajonesezellford

danielpantaleo

nerdland

ccot

equality

miami

policelivesmatter

shutitthefuckdownnfl

shutitdownnyc

michealbrown

detroit

eeuu

blackpower

handsup

truth

pdmfnb

toomanytoname

ohio

wecantbreathe

blackout

foley

revolution

cogic

acab

espn

alllivesmatter

artbasel

indigenouslivesmatter

foxnews

brownlivesmatter

ferguson2cal

crimingwhilewhite

us

thisendstodaywestsidehighway

englandnews

equality4all

d9diein

fergusondecision

rt

handsupdontshoot

bethechange

d6resist

maddow

chicago

justiceforreefa

icantbreathe

travonmartin

racisminamerica

unitebiue

injustice

cnn

whodoyouprotect

repost

christmas

hispaniclivesmatter

akaigurley

cbus2ferguson

tcot

tokyo4ferguson

newyork

berk2ferguson

boycottchristmas

retweet

charlesbarkley

lgbt

nojusticenoprofit

harvard

fergsuon

berkeley

local4

postracialamerica

mynypd

handsupwalkout

grandcentral

justicefortamirrice

downforbrown

sanfrancisco

whitesupremacy

unapologeticblack

pdx

notonedime

ericgarner

against

socialjustice

justice

anon

brooklynbridge

genocide

berkley

amadoudiallo

stopracism

itendstoday

baltimore

icouldbenext

duke

uniteblue

johncrawford

love

nojustice

enoughisenough

charlotte

egypt

cleveland

thisstopstodaypgh

icantbreatheuntil

ripericgarner

endpolicebrutality

nojusticenopeacenoracistpolice

diein

shutdown

allblacklivesmatter

vote

respect

humanrights

oakland

notonemore

ows

wechargegenocide

shutitdowndc

grandjury

usc

artbaselmiami

usa

rumainbrisbon

libcrib

racism

indictthesystem

nojusticeforericgarner

occupy

live

berkeleyprostests

racist

nyc2ferguson

gaza

palestine

mikebrown

cantbreathe

blmla

icantbreath

barclayscenter

bospoli

garner

ripmikebrown

stoptheviolence

punjabi

renishamcbride

stoppolicebrutality

abortion

uiuc

phillydiein

nomorebusinessasusual

cops

dontshoot

dcferguson

thetimeisnow

palestinianlivesmatter

standup

unt

atlanta

nypd

columbuscircle

nojusticenotree

tampa

blackouthollywood

p2

staywoke

legalizedlynching

itstopstoday

againstpolicebrutality

stl

mlk

changethenypd

royalshutdown

bluelivesmatter

seanbell

knowyourrights

noelnight

nojusticeinamerica

nmos

whiteprivilege

whitelivesmatter

statenisland

atlferguson

shutitdownatl

regram

boulder

noindictment

newyorkcity

ferguson2nyc

portland

boston

journeyforjustice

prochoice

india

policebrutality

rip

alaska

imatter

noracistpolice

nba

fergusoneverywhere

riverwalk

thisstopsnow

tcshutitdown

civilrights

philly

sikh

riptamirrice

rockcenterxmas

seattle

prolife

shaw

nojusticenopeace

art

mapoli

amerikkka

teargas

larryjackson

unite

aaa2014

ny

thesystemisbroken

ercigarner

rekiaboyd

auspol

nyc

boycottnypost

neworleans

ourlivesmatter

ramarleygraham

tsu

berkeleyprotest

pain

phoenix

ftp

memphis

blackoutblackfriday

peace

breaking

tbt

justiceformikebrown

brooklyn

peacefulprotest

peterpanlive

yamecanse2

blacktwitter

rp

america

handsupdontchoke

knowjusticeknowpeace

solidarity

darrenwilson

tokyo

plannedparenthood

oscargrant

inners

idlenomore

isis

jordandavis

thursdaynightfootball

dallas

nola

justice4reefa

taneshaanderson

every28hours

bodycameras

ayotzinapa

economicjustice

women

protestnyc

nomore

ericgardner

pjnet

livesmatter

sf

chokehold

fb

FIG. B4. #BlackLivesMatter topic network for the week following the non-indictment of Daniel Pantaleo.

justiceforall

michaelbrown

alivewhileblack

tamirricepolicebrutality

ripericgarner

brownlivesmattercrimingwhilewhite

kellythomas

handsupdontshoot

mikebrown

justice

icantbreathe

policestate

womblifematters

nyc

ourlivesmatter

tcot

rednationrising

ericgarner

shutitdown

blacklivesmatter

nypd

ferguson2miami

ccot miami

policelivesmatter

nojusticenopeace

whitelivesmatter

wecantbreathe

ferguson

FIG. B5. #AllLivesMatter topic network for the week following the non-indictment of Daniel Pantaleo.

22

bhm

livesmatter

itsonus

everylifematters

bhm2015

tamirrice

islamophobia

asianlivesmatter

mourningmonday

muslimlivesmatter

brownlivesmatter

christianlivesmatter

blacktwitter

oakland

blackbusiness

alllivesmatterprince

jewishlivesmatter

handsupdontshoot

mikebrown

blackhistoryyoudidntlearninschool

nojusticenopeace

justice

grammys2015butinacosmicsensenothingreallymatters

blackandproud

beyonce

racism

blackhistorymonth

humanlivesmatter

policestate

blackfriday14

racist

keepitmenace

withmuslims

black

islam

chapehillshooting

grammys

kobejordann

icantbreathe

muslimslivesmatter

darrenwilson

tcot

moorhistorymonth

selma

justiceformuslims

hoodiesup

ericgarner

muslims

translivesmatter

ferguson

nyc

chapelhillshooting

nativelivesmatter

p2

grandcentral

iris_messenger

glory

opicantbreathe

wakeupamerica

bluelivesmatter

blackfuturemonth

beyhive

sobu

shutitdownthisstopstoday

whitelivesmatter

visionsofablackfuture

moorishlivesmatter

FIG. B6. #BlackLivesMatter topic network for the week following the Chapel Hill shooting.

buddhistlivesmatter

chapelhillhatecrime

jewishlivesmatter

christianslivesmatter

islamophobia

butinacosmicsensenothingreallymatters

charliehebdo

rip

muslimlivesmatter

translivesmatter

blacklivesmatter

blacktwitter

chapelhillshooting

ourthreewinners

whitelivesmatter

chapehillshootingblackhistorymonth

FIG. B7. #AllLivesMatter topic network for the week following the Chapel Hill shooting.

23

uniteblue

michaelbrown

enoughisenough

tamirrice

feidinsantana

queen

thisstopstoday

foxnews

michaelslager

police

muslimlivesmatterfilmthepolice

blacktwitter

blackbusiness

alllivesmatter

walterscott

handsupdontshoot

inspiring

mikebrown

ifnotnowwhen

usa

policeshooting

justice

libcrib

icantbreathe

bringintheun

reclaimholyweekracism

princess

charleston

black

pjnet

professu

cnn

nycmorintoon

farrakhanatl

kendrickjohnson

murder

blackchurch

ramseyorta

tcot

ftp

march4kj

chsnews

selma

ripwalterscott

rednationrising

acab

cops

ericgarner

blackoutday

southcarolina

policebrutality nonewnypd

trayvonmartin

ferguson

nightofthelongknivesfreemumia

maddow

p2

stoppolice

whitelivesmatter

rip

ftw

blacktradelines

boycottindiegogo

northcharleston

bluelivesmatter

southcarolinashootingknowyourrights

nojusticenopeace

s

mlk

sc

every28hours

racismisreal

changetheworld

missing

shutdowna14

FIG. B8. #BlackLivesMatter topic network for the week following the death of Walter Scott.

ifnotnowwhen

walterscott

icantbreathe

muslimlivesmatter

blacklivesmatter

ripwalterscott

whitelivesmatter

FIG. B9. #AllLivesMatter topic network for the week following the death of Walter Scott.

24

houston

baltimoreuprising

michaelbrown

oaklandprotests

happeningnowtamirrice

cut50phillyisbaltimore

fergusontobaltimore

muslimlivesmatter

balitmoreuprising

protest

justiceforfreddiegray

tgdn

malcolmx

princepolicethepolice

repost

fuckthepolice

police

enoughisenough

la

justice

staywoke

natashamckenna

socialmedia

unity

denver

nyc

riots

policestate

undarated

justiceforfreddie

marilynmosby

maryland

swastika

texas

kendrickjohnson

selma

dc

whereistheleadershipinbaltimore

whereisjosephkent

shutitdown

asianlivesmatter

anonymous

freddiegrey

trayvonmartin

gunsense

baltimoreriots

ccot

phillipwhite

policelivesmatter

trayvonscott

baltimorenyc

abanks

rightnow

dvnlln

nigger

blackpower

salute

truth

blacktradelines

millionsmarchnyc

myahall

revolution

acab

baltimore2nyc

prayersallblacklivesmatter

foxnews

bluelivesmatter

every28hrs

detroit

fail

mayday2015

retweet

justice4freddie

rt

ftp

what

oakland

freddiegraychicago

prayforbaltimore

icantbreathe

oakland2baltimoreumes

onebaltimore

baltimorerebellion

bullcity

balitmoreriots

mayday

iceland

cnn

gop

blackgirlmagic

2a

tcot

terrencekellum

pdmfnb

tlot

terrancekellom

lgbt

ericgarnerbaltimoreprotest

news

peacebe

nyc2baltimore

ripfreddiegray

botl

pdx

wakeupamerica

nojusticenopeacebaltimorerising

baltimoreriot

stopracism

baltimore

fergusonaction

uniteblue

johncrawford

love

walterscott

boston2baltimore

baltimoreisrising

racism

whiteprivilege

nycriseup

blacktwitter

humanrights

alllivesmatter

ebony

jews

mayweatherpacquiao

mn2bmoremikebrown

libcrib

jew

usabmore

blacktranslivesmatter

philadelphia

baltimoreprotests

maydaysolidarity

occupy

quotes

baltimoremayor

palestine

fmg

orioles

newyork

israel

ainf

riot

stoppolicebrutality

tidalfacts

nycprotest

nyctobaltimore

dcferguson

minneapolis

atlanta

nypd

argentina

freediegray

p2

topprog

tidalforall

stl

mlk

chasenlife

soul

chi2baltimore

ebay

uk

whitelivesmatterows

americanlivesmatter

prayersforbaltimore

blacklives

boston

secondamendment

policebrutality

rip

whereisjusticeinbaltimore

rap

racist

blackbusiness

baltimorelootcrew

blackwomenslivesmatter

ericharris

seattle

hispaniclivesmatter

jerusalem

daca

peace

ourprotest

freddygray

killercops

bailtimoreriots

evencopslivesmatter

rekiaboyd

wakeup

shutdownthetword

dapa

eyery28hours

pjnet

blackspring

handsupdontshoot

uprising

amerikkka

hiphop

peacefulprotest

solidaritywithbaltimore

baltimorepolice

bcpd

america

restinpower

solidarity

laboragainstpoliceterror

harlem

oscargrant

nazi

philly

bmoreunited

makeachange

josephkent

fightfor15

ayotzinapa

ncat

media

nomore

balitmore

ericsheppard

unionsquareferguson

FIG. B10. #BlackLivesMatter topic network for the week encapsulating the peak of the Baltimore protests.

baltimoreuprising

love

baltimore2nyc

prayers

phillyriots

rednationrising

asianlivesmatter

muslimlivesmatter

protest

justiceforfreddiegray

freddiegray

bmore

staywoke

nepal

riots

evencopslivesmattertogetherwecan

tcotamericanlivesmatter

stoppolicebrutality

peace

policebrutality

ericgarner

blacklivesmatter

justice

peacebe

pjnet

ripfreddiegray

baltimoreriots

ccot

policelivesmatter

wakeupamerica

bluelivesmatter

prayforbaltimore

truth

whitelivesmatter baltimore

ferguson

FIG. B11. #AllLivesMatter topic network for the week encapsulating the peak of the Baltimore protests.

25

fifa

baltimoreuprising

amemassacre

tamirrice

gunsense

protest

unitedblue

civilrightsmovement

hatepeoplesmonday

whitefolkwork

motheremmanuel

sayhername

staywoke

socialmedia

ainf

takedowntheflag

black

boycottisraelneverforget

dylannroof

ame

patriot

ferguson

nonewnypd

trayvonmartin

haiti

hatecrime

blackempowerment

domesticterrorism

propheticgrief

bluekluxklan

ccot

apartheid

unionsquare

guns

transracial

race

2a

blacktradelines

charlestonchurchshooting

fairfax

gohomederay

england

prayers

mycity

foxnews

hoorayderay

ohhillno

freddiegray

takeitdown

undersiege

icantbreathe

ameshooting

icc4israel

racisminamerica

charlestonmassacre

prayersforcharleston

cnn

repost

gop

blackgirlmagic

notonemore

itsaracething

tcot

missing

chsshooting

freepalestine

emanuelameconfederatetakedown

biblestudy

saytheirnames

ericgarner

arneshabowers

morningjoe

usa

parenting

wakeupamerica

nojusticenopeace

bronx charleston

baltimore guncontrol

uniteblue

wewillshootback

love

walterscott

enoughisenough

notinournames

mediafail

whiteprivilege

lnybt

blacktwitter

notconservative

tntvote

alllivesmatter

tonyrobinson

jonstewart

mikebrown

libcrib

whitesupremacy

prince

racism

fairfield

occupy

racist

gaza

iamame

mckinney

churchmassacre

actualblacklivesmatter israel

confederateflag

youarehomederay

charelstonshooting

crushjimcrow

everywhere

southcarolina

policelivesmatter

whiteentitlement

melanin

nypd

stillbreathin

pjnet

p2

haitianlivesmatter

word

johncrawford

bluelivesmatter

us

muslimlivesmatter

buyblack

whitelivesmatter

potus

history

charlestonterrorism

clementapinckney

policebrutality

rip

derayishome

blackbusiness

prayforcharleston

historicblackchurch

terrorist

yourcity

breaking

charlestonshootingchurchshooting

nyc

terrorism

ericcasebolt

dominicanrepublic

blackspring

endracism

peace

dc

cbid

america

9candles

takedownthatflag

hillary2016

isis

standwithcharleston

sccharlestonstrong

harlem

racheldolezal

FIG. B12. #BlackLivesMatter topic network for the week following the Charleston church shooting.

gunsense

hilaryclinton

cosproject

ohhillno

charlestonshooting

deray

prayersforcharleston

gunfreezoneiamame

tcot

endracism

nra

ferguson

pjnet

muslimlivesmatter

wakeupamerica

bluelivesmatter

2a

charleston

whitelivesmatter

gohomederay

blacklivesmatter

FIG. B13. #AllLivesMatter topic network for the week following the Charleston church shoting.

26

nyc

michaelbrown tamirrice

millionmanmarch

muslimlivesmatter

facts

rexdalehenry

repost

fuckthepolice

suicide

police

sayhername

justice

staywoke

natashamckenna

portugalstandswithsandrabland

whitesupremacy

black

msnbc

texas

kendrickjohnson

neverforget

wallercounty

michaelsabbie

shutitdown

ferguson

anonymous

wakeup

trayvonmartin

lafayette

feelthebern

ccot

equality

ralkinajones

unionsquare

makeamericagreatagain

gop

jonathansanders

1u

indiaclarke

race

truth

justiceforsandrabland

kendrachapman

tntweeters

gohomederay

bernie

wakeupamerica

someonetellcnn

sandraspeaks

lnyhbt

arizonadefundpp

cleveland

brianencinia

retweetblackgirlsmatter

rt

stopkillingus

oakland

kimberleerandleking

freddiegray

maddow

newjersey

icantbreathe

defundplannedparenthood

rip

newark

racisminamerica

unitebiue

injustice

justiceforsandra

cnn

notallmen

justiceorelse

feminism

sandyspeaks

hispaniclivesmatter

akaigurley

hillaryclinton

cdcwhistleblower

tcot

newyork

sandrablandquestions

tlot

saytheirnames

lgbt

ericgarner

mustread

california

killedbycops

humanrights

millionpeoples

babyparts

nativelivesmatter

change

grandcentral

whathappenedtosandybland

usa

nn15

darriusstewart

richardlinyard

justiceforsandy

youth4bernie

systematicracism

nojusticenopeace

genocide

weneedjustice

justicefordarrius

charleston

blm

bernie2016

baltimoreuniteblue

blackgirlsrock

5t3f4n

enoughisenough

godbless

whitepeople

justiceforkindrachapman

youmatter

whiteprivilege

emmetttill

whathappenedtosandrabland

blacktwitter

allblacklivesmatter

m4bl

alllivesmatter

berniesoblack

blackouttwitter

libcrib

blackpower

deathbydemocrat

blacktranslivesmatter

racism

blackwomenmatter

babylivesmatter

teaparty

mikebrown

ppsellsbabyparts

wearegreen

brandonclaxton

assatashakur

notonmydime

weneedanswers

stoppolicebrutality

tentoomany

translivesmatter

justiceforblacklives

standup

nypd

memphis

pjnet

women

p2

sandrabland

lovewins

ripsandrablandstl

shaunking

bluelivesmatter

endofthegop

samueldubose

buyblack

whitelivesmatter

kindrachapman

history

sandystillspeaks

gamergate

710pepp

handsupdontshoot

policebrutality

kimberleeking

blackbusiness

sandrawasmurdered

nerdlandwewantanswers

handsupdontcrush

policelie

civilrights

racistuneducatednarcissist

blacklove

ifidieinpolicecustody

unity

thisiswhywefight

rekiaboyd

pepp

hearus

sandarabland

blackspring

ftp

endracism

iamsandrabland

sandybland

saveouryouth

blackoutday

ff

sandra

selfcare

america

restinpower

solidarity doj

hillary2016

plannedparenthood

justice4sandy

justice4sandrabland

dayofrage

austin

hillary

blackpeople

berniesanders

yesallwomen

ohhillno

myblackisbeautiful

m4blfree

FIG. B14. #BlackLivesMatter topic network for the week encapsulating outrage over the death of Sandra Bland.

unitebluepolicebrutality

muslimlivesmatter

defundppwhathappenedtosandrabland

blacktwitter

handsupdontcrush

prolife

racisminamerica

justice

badthingsarebad

sayhername

unity

imwithhuck

babylivesmatter

notallmen

ppsellsbabyparts

tcot

defundplannedparenthood

tlot

embarrassment

blacklivesmatter

pjnet

obama

sandrabland

ccot

nn15

wakeupamerica

bluelivesmatter

whathappened

justiceforsandrabland

stopracism

whitelivesmatter

FIG. B15. #AllLivesMatter topic network for the week encapsulating outrage over the death of Sandra Bland.