Improving information spread by spreading groups
| Pages | 24-42 |
| DOI | https://doi.org/10.1108/OIR-08-2018-0245 |
| Published date | 15 November 2019 |
| Date | 15 November 2019 |
| Author | Alon Sela,Orit Milo,Eugene Kagan,Irad Ben-Gal |
| Subject Matter | Library & information science,Information behaviour & retrieval,Collection building & management,Bibliometrics,Databases,Information & knowledge management,Information & communications technology,Internet,Records management & preservation,Document management |
Improving information spread by
spreading groups
Alon Sela
Department of Industrial Engineering, Ariel University, Ariel, Israel;
Department of Industrial Engineering, Tel Aviv University, Tel Aviv, Israel and
Ariel Cyber Innovation Center (ACIC), Ariel University, Ariel, Israel
Orit Milo
Department of Finance, Hebrew University of Jerusalem, Jerusalem, Israel
Eugene Kagan
Department of Industrial Engineering, Ariel University, Ariel, Israel, and
Irad Ben-Gal
Department of Industrial Engineering, Tel Aviv University, Tel Aviv, Israel
Abstract
Purpose –The purpose of this paper is to propose a novel method to enhance the spread of messages in
social networks by “Spreading Groups.”These sub-structures of highly connected accounts intentionally echo
messages between the members of the subgroup at the early stages of a spread. This echoing further boosts
the spread to regions substantially larger than the initial region. These spreading accounts can be actual
humans or social bots.
Design/methodology/approach –The paper reveals an interesting anomaly in information cascades in
Twitter and proposes the spreading group model that explains this anomaly. The model was tested using an
agent-based simulation, real Twitter data and questionnaires.
Findings –The messages of few anonymous Twitter accounts spread on average more than well-known
global financial media groups, such as The Wall Street Journal or Bloomberg. The spreading groups (also
sometimes called BotNets) model provides an effective mechanism that can explain these findings.
Research limitations/implications –Spreading groups are only one possible mechanism that can explain
the effectiveness of spread of tweets from lesser known accounts. The implication of this work is in showing
how spreading groups can be used as a mechanism to spread messages in social networks. The construction
of spreading groups is rather technical and does not require using opinion leaders. Similar to the case of “Fake
News,”we expect the topic of spreading groups and their aim to manipulate information to receive growing
attention in public discussion.
Practical implications –While harnessing opinion leaders to spread messages is costly, constructing
spreading groups is more technical and replicable. Spreading groups are an efficient method to amplify the
spread of message in social networks.
Social implications –With the blossoming of fake news, one might tend to assess the reliability of news by
the number of users involved in its spread. This heuristic might be easily fooled by spreading groups.
Furthermore, spreading groups consisting of a blend of human and computerized bots might be hard to
detect. They can be used to manipulate financial markets or political campaigns.
Originality/value –The paper demonstrates an anomaly in Twitter that was not studied before. It proposes
a novel approach to spreading messages in social networks. The methods presented in the paper are valuable
for anyone interested in spreading messages or an agenda such as political actors or other agenda
enthusiasts. While social bots have been widely studied, their synchronization to increase the spread is novel.
Keywords Information spread, Viral marketing, Social networks, Spreading groups, Fake news, BotNets
Paper type Research paper
1. Introduction
Modern social network platforms provide a simple and efficient way to spread messages and
increase their influence (Teng et al., 2014) among individuals or communities of users. The
results of such message spread (the terms “messages”and “information”are used
Online Information Review
Vol. 44 No. 1, 2020
pp. 24-42
© Emerald PublishingLimited
1468-4527
DOI 10.1108/OIR-08-2018-0245
Received 20 August 2018
Revised 1 March 2019
14 June 2019
Accepted 12 August 2019
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1468-4527.htm
Ariel Cyber Innovation Center (ACIC), Ariel University, Ariel, 40700, Israel.
24
OIR
44,1
interchangeably) can vary in scale, from serving the needs of an individual user to having a
widespread influence on political (Harvey, 2013), social and economic phenomena. Examples of
large-scale social influences where social networks played a major role include the “Arab
Spring”(Howard et al.,2011),the“Occupy Wall Street”movement (Bennett and Segelberg, 2011)
and the elections in the USA (Allcott and Gentzkow, 2017; Kim and Hastak, 2018). In fact, in
several areas, the role of social networks and their smart use by a small group of users can lead
to an outcome that can be as influential as that of traditional media.
In the financial markets, the role of social networks as an information spreader is critical.
Research has established that the news spread through a social network is often published
before it reaches the traditional media (Leskovec et al., 2009). Trust is also a critical aspect,
affecting the likelihood of a message’s spread (Gorman, 2014). Nevertheless, the blossoming
of fake news (Allcott and Gentzkow, 2017; Newman et al., 2017) might suggest that
sometimes trust can be misleading, and fake rumors can seem to be trustworthy sources.
Our investigation was motivated by the observation that there are individual Twitter
accounts with especially high levels of average retweet rates that are substantially higher than
those of well-known global media companiessuchasTheNewYorkTimesorTheWallStreet
Journal. We define “well-known users”to be international media outlets who are traditionally
“information broadcasters.”More specifically, we define well-known users as international
media companies with more than 1m followers. We expect such organizations to play a more
significant role in information dissemination compared to a private account or small companies.
To explain these anomalies, we constructed a novel theoretical model that demonstrates
the boosting of a message’s spread by spreading groups. Such spreading groups are
organized sub-structures of interconnected users that intentionally amplify the exposure of a
message that they wish to spread. To accomplish this goal, they utilize a ping-pong-like
interchange of the message within the group’s members in its early spreading stages.
Through this strategy of transmitting a message overand over within the group in the early
stages of its spread,these groups ignite a larger information cascade, and help the message’s
spread to other parts of the social network that are external to the group itself at laterstages.
The possible benefits gained from using spreading groups are clear. In contrast to other
message-spreading methods, such as using opinion leaders, the construction of spreading
groups is mainly a technical issue. The formation of spreading groups can be accomplished
with real Twitter accounts or with the accounts of bots (computer programs that imitate a
social network user). Thus, although the initial construction of spreading groups requires
some initial technical skill, its later use to gain influence in a social network is easy.
The paper is organized as follows. Section 2 provides a short history of rumor spread in
stock-related domains and suggests a motivation for the creation of spreading groups.
Section 3 presents the theoretical model and, using simulations, demonstrates how by using
only spreading groups a message can be spread to a larger audience. This section defines
the model’s characteristics and formulates different criteria for the intentional spread of
information. Section 4 describes the methods of Twitter data analysis, which is followed by
the experimental results in Section 5. The Results section begins with the theoretical model
and then details the Twitter data set analysis and its results. Section 6 discusses the
findings, and Section 7 concludes by describing the implications of bot technology in
message spread and its relevance to spreading groups.
2. Background
Studies of stock-related news rumors and their influence on financial markets can be traced
back to the rise of internet message boards and forums. In early 1998, for example, the
correlation between the number of published messages about companies on the Yahoo
Finance Board and their shares’values was noticed and reported (US Securities and
Exchange Commission, 1998). In their study of the social influence on the stock market,
25
Spreading
groups
To continue reading
Request your trial