Shockwaves and turbulence across social media Pedro D. Manrique Frank Huo Sara El Oud Minzhang Zheng Lucia Illari Neil F. Johnson Physics Department George Washington University Washington DC 20052 U.S.A.

2025-05-03 0 0 2.71MB 6 页 10玖币

侵权投诉

Shockwaves and turbulence across social media

Pedro D. Manrique, Frank Huo, Sara El Oud, Minzhang Zheng, Lucia Illari, Neil F. Johnson

Physics Department, George Washington University, Washington, DC 20052, U.S.A.

(Dated: October 27, 2022)

Online communities featuring ‘anti-X’ hate and extremism, somehow thrive online despite mod-

erator pressure. We present a ﬁrst-principles theory of their dynamics, which accounts for the fact

that the online population comprises diverse individuals and evolves in time. The resulting equa-

tion represents a novel generalization of nonlinear ﬂuid physics and explains the observed behavior

across scales. Its shockwave-like solutions explain how, why and when such activity rises from

‘out-of-nowhere’, and show how it can be delayed, re-shaped and even prevented by adjusting the

online collective chemistry. This theory and ﬁndings should also be applicable to anti-X activity in

next-generation ecosystems featuring blockchain platforms and Metaverses.

Society is struggling with online anti-X hate and extrem-

ism, where ‘X’ can nowadays be any topic, e.g. religion,

race, ethnicity [1–5]. Recent research has conﬁrmed that

in-built online communities play a key role in develop-

ing support for a topic at scale [6] and anti-X sentiment

is no diﬀerent [1–5]. These in-built communities are re-

ferred to diﬀerently on diﬀerent platforms, e.g. Group

on VKontakte and on Gab, Page on Facebook, Channel

on Telegram, and are unrelated to community-detection

in networks. Each in-built community is a self-organized

aggregate of anywhere from a few to a few million users.

Such anti-X communities can grow quickly from out of

nowhere because of interested individuals or other com-

munities joining (fusing) with them (Fig. 1(a), empirical

fusion) [6–10]. Having content that violates platforms’

Terms and Conditions means that they can also suddenly

get shut down when discovered by moderators (Fig. 1(b),

empirical total ﬁssion). Therefore in contrast to commu-

nities such as pizza fans, there is a clear beneﬁt for such

anti-X communities to grow in a bottom-up way in or-

der to remain under moderators’ radar. Figure 2(a)(b)

illustrates the sea of erratic shark-ﬁn-shaped waves that

emerges: each shows an anti-X community’s size of mem-

bership as it suddenly appears and grows through fusion

and may then suddenly disappear via total ﬁssion. Some

social scientists [11] are suggesting that such volatility is

‘online turbulence’ which could – if proven true – open up

an important new ﬁeld for physics and also help bridge

the current gap between computational approaches to on-

line (mis)behavior and in-depth case-studies [12].

Unfortunately such physics does not yet exist, i.e.

there is no ﬁrst-principles theory that accounts for popu-

lations of objects (e.g. anti-X individuals) that (1) have

their own internal character that may evolve over time,

and (2) interact in a distance-independent way as allowed

by the Internet, and (3) have a changing total size (e.g.

Internet use jumped 13.2% in 2020), and (4) undergo

rapid fusion-ﬁssion dynamics as in Fig. 1, Fig. 2 (a)(b).

Here we propose this new physics via a ﬁrst-principles

dynamical theory of anti-X communities within and

across social media platforms. The resulting equation

(Eq. 2; derivation SM Sec. 2) provides a novel generaliza-

FIG. 1. Empirically observed (a) fusion and (b) total ﬁssion

of in-built communities featuring anti-U.S. hate on VKon-

takte between day t(yellow) and t+ 1 (blue). Red nodes are

anti-U.S. communities that later got shut down (total ﬁssion);

green nodes are those still not yet shut down; yellow links

point to individuals (white dots) removed from the anti-U.S.

community on day t+ 1; blue links point to individuals added

to the anti-U.S. community on day t+1. Spatial layout results

from (a) and (b) being closeups of a fuller network plotted

using ForceAtlas2, meaning that nodes appearing closer to-

gether are more interconnected. (b) also shows that very few

individuals are simultaneously also members of other com-

munities (SM shows further proof). (c) Empirically observed

clustering of anti-government communities across platforms

around U.S. Capitol attack. (d)-(f): The theory in this paper

incorporates (d) heterogeneous individuals aggregating (i.e.

fusion) based on character similarity, (e) total ﬁssion with

probability νf, (f) time-varying population size N(t).

tion of nonlinear ﬂuid physics, including shockwaves and

turbulence, and extends the physics of aggregation and

fragmentation [13–34]. Its solutions explain empirically

observed patterns within and across social media plat-

forms (Fig. 2), and predict how the rise of anti-X com-

munities can be delayed, re-shaped and even prevented

arXiv:2210.14382v1 [nlin.AO] 25 Oct 2022

(a) (b)

(d)

(f)

01/2015 03/2015 05/2015

6average

F(t)

01 03 05

N(t)/ 104

2015

theory

data

×103

size

001/2019 01/2020

theory

data

×103

6F(t)

average

F(t)

101102103104

= 2.46 ± 0.29

within

VKontakte

10−1

10−2

CCDF

10−2

= 2.5 ± 0.5

within

Facebook

102103104

10−1

total size

07/2019 01/2020

within

Facebook

×105

theory

data

across

platforms

100101102

10−1

10−2 = 2.55 ± 0.13

(1,2)

0510

CCDF CCDF

FIG. 2. Empirical data (symbols) and Eq. 2 theory predictions (lines) for in-built anti-X communities within and across

platforms. (a) Size (i.e. number of members) of foreign anti-U.S. (jihadi) communities on VKontakte. (b) Size of domestic

anti-U.S. government (pro-civil war) communities on Facebook. Insets: changing population size; time-averaged F(t) which

suggests that (b) reﬂects a heterophily fusion mechanism more than (a). (c-d) Complementary cumulative distribution (CCDF)

of individual community sizes sfrom (a) and (b). (e) Evolution of total size of all communities from (b). (f) CCDF at a higher

scale, i.e. sizes of clusters of interlinked communities. Inset: empirically inferred interaction kernel W(s1, s2) obtained from

data across all platforms; axes s1and s2are sizes of interacting aggregates.

(Fig. 3). Its approximate analytic solutions can also

explain the complex multi-community evolution around

the U.S. Capitol attack (Fig. 4(b)). Our empirical data

is obtained using a published methodology [35–37] that

we summarize in SM Sec. 1. We are not claiming that

all online anti-X activity will always exhibit the patterns

in Fig. 2 and Fig. 4(b), or that all non-anti-X activity

never does – however, the important cases shown here

do, while SM Sec. 1.1 shows that non-anti-X communi-

ties typically do not. Situations in which the observed

anti-X activity does not follow these patterns may there-

fore be indicative of other mechanisms being at play, e.g.

top-down coordination or state actor control.

Our theory considers N(t) heterogeneous individu-

als that are attracted online by such shocking content

and hence could aggregate over time depending on their

traits. Each aggregate (i.e. in-built community) then

totally fragments with some small probability at each

timestep (Fig. 1(d)-(f)). Following prior social sci-

ence and physics studies [13, 14, 27–32], each individual

i= 1,2, ..., N(t) can have an arbitrary number of traits,

expressed as a vector ~yi(t) where each component (trait

value) lies between 0 and 1; but for notational simplicity

we only consider one here. At each timestep, two (e.g.

randomly) chosen individuals iand jcan fuse together

with a probability that depends on the pair’s similarity

|yi(t)−yj(t)|(Fig. 1(d)). If iand jare already part

of an aggregate, their whole aggregates fuse. Hence the

mechanism accounts for loners (aggregates of size s= 1)

joining together, or a loner joining a community (aggre-

gate of any size s > 1), or two communities (aggregates

of any size s1, s2>1) joining together (Fig. 1(a)). We

can calculate a mean-ﬁeld fusion probability F(t) by av-

eraging over the population distribution at time t, e.g.

for a constant uniform distribution, pairing favoring sim-

ilarity (homophily) yields F(t) = 2/3 while dissimilarity

(heterophily) yields F(t)=1/3 (details in SM Sec. 2.1).

In this way, F(t) captures the online collective chemistry.

Master equations for the number ns(t) of aggregates of

size s, are for s > 1 and s= 1 respectively:

˙ns=F(t)

N(t)2X

s1+s2=s

s1ns1s2ns2−2F(t)sns

N(t)2

∞

s1=1

s1ns1−νfsns

N(t)

˙n1=−2F(t)n1

N(t)2

∞

s1=1

s1ns1+νf

N(t)

∞

s1=2

1ns1+˙

N(t) (1)

where the ﬁrst term(s) on each right-hand side are fusion,

the next are total ﬁssion, and the ﬁnal ˙n1(t) term is the

inﬂux of potential recruits. The fusion product-kernel

is justiﬁed empirically by studies of humans’ electronic

communications [23] and by the online anti-X data (see

Fig. 2(f) inset and SM Sec. 1.1). We made the reason-

able assumption that macrolevel quantities F(t) and N(t)

vary slowly compared to microlevel aggregation: the SM

shows this is justiﬁed by comparing to full microscopic

simulations. Deﬁning u(x, t) = Pssns(t)e−xs, Eq. 1

becomes

˙u(x, t) = −2F(t)

N2(t)u(x, t)u0(x, t) + 2[F(t) + νf/2]

N(t)u0(x, t)

+e−x[˙

N(t)−νf

N(t)u0(x, t)|x=0] (2)

where u0is the x-derivative. Equation 2 is a novel

generalization of nonlinear ﬂuid equations with shock-

wave solutions. An additional link mechanism, discussed

later, would add the diﬀusive term u00 (x, t) typical of

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

ShockwavesandturbulenceacrosssocialmediaPedroD.Manrique,FrankHuo,SaraElOud,MinzhangZheng,LuciaIllari,NeilF.JohnsonPhysicsDepartment,GeorgeWashingtonUniversity,Washington,DC20052,U.S.A.(Dated:October27,2022)Onlinecommunitiesfeaturing`anti-X'hateandextremism,somehowthriveonlinedespitemod-eratorpressur...

展开>> 收起<<

Shockwaves and turbulence across social media Pedro D. Manrique Frank Huo Sara El Oud Minzhang Zheng Lucia Illari Neil F. Johnson Physics Department George Washington University Washington DC 20052 U.S.A..pdf

共6页,预览2页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Shockwaves and turbulence across social media Pedro D. Manrique Frank Huo Sara El Oud Minzhang Zheng Lucia Illari Neil F. Johnson Physics Department George Washington University Washington DC 20052 U.S.A.

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: