January 10, 2016
1
Social Media in Disaster Management
Roman Fekolkin [email protected]
Trends in IS and Information Security, Luleå University of Technology
Abstract – Provision of an accurate and timely disaster
relief effort requires an insightful informational
coverage of a crisis situation. Social media has a
potential of providing a large amount of information
generated by people about different aspects of a certain
crisis situation. However, emergency-oriented social
media utilization is still at an embryonic stage of its
understanding. So, in this report 10 research studies
were reviewed in an attempt to gain some understanding
about this approach. As it appeared, crowdsourced
information allows to get a versatile and insightful
understanding of a crisis situation, but, as with any Big
Data, there are major challenges of making use of high
volumes of diverse information, in terms of extraction
and representation of relevant and accurate
information. Furthermore, other research studies
proposed extraction/representation techniques which
were primarily aimed at thematic and geographical
clustering of information. As it appeared from the real
cases of social media application during emergencies,
there is still no standardized approach on how such
information can be handled, but in those cases,
emergency responders still managed to successfully
apply their own methods of primarily geographical and
thematic clustering of information, while still being
greatly challenged by overwhelming amount of user-
generated information. Thus, the success could perhaps
be explained by social media being used primarily as a
complement/enhancement of other already proved to be
effective techniques. Further exploration would be
needed in order to gain a more complete understanding
of the current state emergency-oriented social media
utilization, which could later be used as a theoretical
framework for a new research study in this area.
Index Terms – Big Data, content filtering, disaster
management, information extraction, social media
INTRODUCTION
Social media nowadays has become a widely used
communicational medium and generally essential
component of many people’s lives. Social media offer a
range of possibilities to many people when it comes to not
only consumption, but also creation of content, establishing
convenient opportunities for socialization (Yates &
Paquette, 2011; Yin, Lampert, Cameron, Robinson, &
Power, 2012). However, social media can be used for more
than just communication and sharing information between
acquaintances. Users of social networks directly or
indirectly share their opinions, experiences and interests
about different things and events, the information about
which from the perspective of those users can be very useful
for various organizations, including the ones that deal with
disaster relief efforts, when it comes to making appropriate
decisions in an objective, contextually accurate and
productive manner (Yates & Paquette, 2011; Kavanaugh,
Fox, Sheetz, Yang, Li, Shoemaker & Xie, 2012; Lohr,
2012).
In particular, Big Data obtained from social
networks can be used for such serious purposes as raising
social awareness among a large group of individuals on a
national or even international level about various
phenomenon and different world events, such as natural or
human-inflicted disasters (Yin, Lampert, Cameron,
Robinson, & Power, 2012; Imran, Elbassuoni, Castillo, Diaz
& Meier, 2013) and thereby provide continuous support for
a more productive and accurate disaster response (Ehnis &
Bunker, 2012; McClendon & Robinson, 2013). Accuracy of
a disaster response can be a particularly challenging thing to
achieve, because in case of disasters, usefulness of such
response depends on actual understanding of dynamically
changing contextual conditions of an affected environment
(Yates & Paquette, 2011; Ehnis & Bunker, 2012). That is
why, considering potential usefulness of user-generated
content/knowledge for gaining a versatile or, in other words,
multi-perspectival understanding of a particular emergency
situation and taking into account unacceptability of wasting
time and resources in vain when people’s lives are at stake,
there is a need to properly understand how to extract
relevant content and put it to good use.
I. Social Media and Big Data
In order for Big Data, generated by social networks, to make
sense and facilitate its accurate usage in problem-solving
processes without making conclusion based on misleading
findings with potentially disastrous consequences, it would
have to be analysed accurately. However, extraction and
structurally understandable representation of tremendous
amounts of non-homogeneous data collected from social
networks is far from being a trivial task, making it
absolutely essential to apply highly sophisticated
technological solutions and qualified experts in order to
analyse and actually make practical use of all that data
(Lohr, 2012; Mayer-Schönberger & Cukier, 2013). The
importance of extraction of relevant information and its
analysis was raised in numerous research studies (Yin,
Lampert, Cameron, Robinson, & Power, 2012; Imran,
Elbassuoni, Castillo, Diaz & Meier, 2013; Middleton,
January 10, 2016
2
Middleton & Modafferi, 2014), where the authors proposed
various techniques; including geographical clustering and
topic-based filtering of user entries, some of which are
discussed later in this paper. In general, depending on the
context, by making sense of such large and complex data
sets and converting them into knowledge, there would be a
possibility not only to make predictions about potentially
beneficial business moves, but also get a valuable insight
into a certain situation and even make potentially life-saving
decisions, when it comes to using social media-originated
Big Data analysis for emergency management (Tene &
Polonetsky, 2012; Cavoukian & Jonas, 2012).
However, analysis of Big Data is, in its turn,
challenged by technological, intellectual and thereby
financial resources that would have to be invested into it,
which might simply be unfeasible to handle for many
organizations. Therefore, when it comes to making sense of
data retrieved from social networks for disaster
management, government agencies should be involved, not
only because it would be feasible or realistic for them to
invest the necessary financial and intellectual resources, but
also because it should be their responsibility to ensure safety
of their citizens during emergencies (Yates & Paquette,
2011; Kawtrakul, Mulasastra, Chanlekha, Rajbhandari,
Prathumchai, Nagai & Khunthong, 2012).
II. Purpose of this paper
Usage of social media has been rapidly gaining recognition
as a powerful tool for emergency management, but its
capabilities, just as with any new technology trend, are not
yet fully explored, which is understandable because
maturity of approaches evolves over time. So, the purpose
of this paper was to analyse research studies in the field of
social media application in emergency management. This
was done in order to get an insight about how this rapidly
growing in popularity technique has been used by
emergency agencies for the last several years, exploring its
benefits and challenges, and how those challenges were
dealt with. To be more precise, this paper sought to provide
answers to the following questions:
What advantages and challenges arise from usage
of social media in emergency management?
What types of approaches can be used for
extraction/representation of relevant information
from user-generated content on social networks?
How is social media used in emergency
management in real emergency situations?
METHODOLOGY
As it was stated by Denscombe (2014), a document study
could be useful for getting acquainted with relevant research
studies conducted by others and then using that theoretical
framework for designing a full-scale research and assist
researchers in interpretation of their findings. So, in this
study, a document study in a form of a literature review was
employed and even though this paper was merely about
studying relevant literature about the chosen topic, the
constructed theoretical framework could still be used for
designing a full-scale research study about social media
application in emergency management and interpreting its
findings.
Furthermore, in this paper, the research studies
were reviewed in such a way that would build a coherent
and logical discussion of the main arguments that the
authors have attempted to communicate to the reader. So, in
this report, relevant papers were analysed and their main
points were extracted into a coherent summary. By being
relevant, it is implied that the reviewed papers were related
to the field of emergency-oriented social media utilization.
Through the literature review, this paper presents not only
potential challenges that can arise from using social media
during disaster events, but also the ways how those
challenges can be dealt with.
The requirements for this assignment stated that the
analysis had to include 8 to 10 relevant research studies.
That’s exactly have been done. Furthermore, as it has been
stated in one of the research studies (Yates and Paquette,
2011) that was published in 2011, at that period of time
there were no previous research studies that explored social
media utilization by emergency agencies, which was
actually the primary interest of this particular report.
Therefore, the year 2011 was used as the earliest publication
date a paper could have in order to be chosen for analysis.
The literature review section was divided into three
sub-sections, with each one being dedicated different
aspects of social media usage in emergency management.
So, the first sub-section is dedicated to discussion of
benefits and challenges with emergency-oriented social
media utilization. The second sub-section discusses some of
the approaches that were proposed by other research studies
that were aimed to deal with some of those challenges.
Finally, the last sub-section presents a discussion of some of
the real case examples of how social media was used in
crisis situations. To maintain the balance between the sub-
sections, each of them presented a discussion of three
research studies relevant to their respective theme/aspect
that they had to cover. In the following table 1 there is an
overview of the number of papers that were discussed in
each sub-section.
TABLE I TOTAL NUMBER OF PAPERS REVIEWED IN EACH SUB-SECTION
Topic Number of papers
Introduction to Literature
review 1
Benefits and challenges with
usage of social media in
disaster management 3
Approaches for extraction and
representation of information 3
Examples of real cases of social
media application during
emergencies 3
January 10, 2016
3
Each sub-section was concluded with a table that
summarized the main points of the reviewed research
papers, either they the identified challenges/benefits of
social media or types of methods used for
extraction/representation of information.
To search for relevant research papers for analysis,
Google Scholar was used. The main search criterion,
besides the aforementioned year of publication, was its
relation to usage of social media in emergency management.
LITERATURE REVIEW
As it was mentioned previously, it would take time before
effective ways of social media application during
emergencies could be fully understood. So, the authors of
one of the not so recent research studies (Merchant, Elmer
& Lurie, 2011), which was actually published when usage
of social media for emergency management was still a “new
thing” and this concept was still at an embryonic stage of its
development with very few explorative research studies
being published at that time, and associatively limitations of
that approach were not yet properly understood. So, the
authors gave somewhat predictive but, as it turned out from
the upcoming discussion, reasonable statements that
crowdsourced information collected from social networks
could serve as a valuable knowledge base for supporting
emergency agencies in designing contextually more
accurate and generally relevant plans for
preparation/prevention, response and recovery with a greater
community involvement and thereby greater response
network, all of which could at the very least minimize risks
of uncontrollably catastrophic development of crisis events.
However, the authors were not blinded by potential power
of social media in emergency management, but rather they
were realistic about challenges that would have to be faced
before social media could realize its potential in such
serious area of application. Namely, the authors have
importantly pointed out the issue with inaccurate or even
deliberately falsified information posted by users on social
networks, which could potentially mislead decision-making
processes, which in case of crisis situations could have
severe consequences, when people in need might simply be
deprived of relevant or any assistance/aid at all. Thus,
understandably, in order to avoid taking risks with full
reliance on social media during emergencies, it was
suggested by the authors that it would preferably have to be
employed in a strategically aligned combination with
traditional (more common) approaches used by a particular
emergency agency for their enhancement, due to social
media being unable, at that time, to serve as a superior or
even an equally effective standalone alternative to those
already tested approaches.
So, even though, at the time the aforementioned
research study was published there were still significant
gaps in knowledge about social media application for
emergency management, the authors still managed to give a
comprehensive overview of this approach, realistically
discussing not only its potential benefits, but also
emphasizing its indisputably critical shortcomings, which
would undoubtedly have to be accounted for when putting
such approach to use. The arguments presented by that
study were later discussed in other research studies that
aimed to present scientifically valuable arguments based on
extensive studies of real cases and actual experiments with
practical application of such approach.
III. Benefits and challenges with usage of social media
in disaster management
One of the relatively early research studies (Gao, Barbier &
Goolsby, 2011) presented enlightening arguments about
benefits and challenges with usage of social media’s
crowdsourced information for disaster relief efforts. The
authors did state that production of information on social
networks is based on its participatory creation by a variety
of different individuals with different opinions and
experiences, which makes that information highly versatile
and thereby particularly valuable for making accurate
emergency management planning. However, according to
the authors, this was not the main advantage of social
media. They looked at social media usefulness from the
perspective of its more specific capabilities that could bring
solid benefits to those who utilizes them for emergency
management. So, they interestingly outlined that users on
social networks tend to respond almost immediately to
emergencies, by posting information about those
emergencies on social networks. That information could in
its turn be used by emergency agencies for getting a timely
insight into a particular situation. This timely information
acquisition is indeed critical especially because in those
situations the lack of timeliness could result in unacceptably
severe and life-threatening consequences. Another benefit
that was emphasized by the authors was that information
posted by users on social networks could be categorized by
their tags, which could help emergency agencies to more
easily identify the most important issues and address their
resolution. Information contained in users’ entries could
even be categorized by geo-tags. Such geographical
mapping was argued to be capable of providing emergency
agencies with a possibility to more effectively locate
requests for help and get a generally more comprehensive
understanding of emergency’s big picture. Understandably,
there are certain shortcomings with analysis of large
amounts of highly variant information, just as with any Big
Data. Those shortcomings and their role in prevention of
social media from becoming a productive tool for
emergency management were carefully discussed by the
authors of that paper. Some of the emphasized shortcomings
included potential inaccuracy or insufficiency of
information posted by users on social networks information
during emergencies for emergency agencies to be able to
use it for emergency management or, even worse, use it
with contextually relevant precision. No surprise was the
fact that users’ entries might not necessarily have geo-tags
assigned to them or duplicate entries with the same geo-tags
could exist greatly distorting real distribution of emergency
January 10, 2016
4
“reports”, all of which would practically render the idea of
geographical mapping of crowdsourced information
infeasible. Naturally, one would think how to tame power of
social media and make it beneficial. Therefore, the authors
understandably stated that in order for crowdsourced
information to be usefully applied in emergency
management, there would be a need for it to be properly
processed and analysed by mechanisms for verification of
its relevance and accuracy, with a follow-up categorization
and visual representation, making it readily-available for
exploitation by emergency agencies. The authors’
arguments are very relevant, because accuracy and time-
efficiency are absolutely essential in emergency
management.
There was a more recent research study
(Alexander, 2014), which, regardless of the almost three
year difference in the time of publication, presented a lot of
similar arguments about benefits and shortfalls of
emergency-oriented social media utilization as the ones
presented by the authors of the aforementioned paper.
Namely, the author also outlined the fact that collaboratively
created and thereby versatile information collected from
social networks could allow emergency agencies to act
proactively for crisis risk reduction and reactively for
responding to crisis events, having contextual characteristics
properly understood and regular people’s opinions
accounted for when making decisions about how to assist
them during crises. Furthermore, the author even pointed
out an important but, at the same time, somewhat obvious
fact that usage of collaboratively created information for
emergency management would make those people
contributors to emergency management, making it socially
inclusive. Interestingly, what has not been mentioned in the
previous research study is the fact that public availability of
information broadcasted on social networks could be used
for attracting people’s attention and boosting fundraising,
for example, for assisting people with emergency recovery.
However, just like in the aforementioned study by Gao et al.
(2011), the author here did not explicitly mention social
media being useful for recovery from those events, which
was actually emphasized by the previously discussed
Merchant et al. (2011). On the other hand, recovery phase of
emergency management could perhaps be considered as one
of the components of risk reduction phase, where dealing
with consequences of a crisis would include taking
measures for prevention of those factors from failing to
maintain stability again during next crisis events. As for the
challenges with social media application for emergency
management, they author’s arguments were pretty aligned
with the ones presented by the aforementioned Gao et al.
Namely, the author unsurprisingly pointed out dangers of
inaccurate or deliberately falsified and thereby misleading
information posted by users on social networks, which
would not only make it useless but even dangerous when
people’s lives are at stake. So, expectedly, another noted
challenge was people perceiving information generated by
other people on social networks being less trustworthy than
information provided by official media outlets and
authorities. The author’s arguments about social media
being potentially misused by “bad guys” for carrying out
their sinister plans and causing crisis situations are not
exactly related to challenges with using social media in
emergency management, but rather it’s a general threat that
can emerge from social media, which could however be
accounted for when designing a contingency planning. The
absolutely essential downside of social media that was
unacceptably overlooked by the previous research study was
the fact that without electricity and the Internet people
would simply be unable to use technology for accessing
social media, considering that there is no guarantee that
availability of those aspects would be maintained during
disaster events. This problem could indeed prevent people
from submitting critical information and associatively
emergency agencies would be unable to provide necessary
assistance, which supports the arguments of that same paper
by Gao et al. about importance of using social media in
strategically aligned combination with traditional methods
used by a particular emergency agency. Last but not least, it
is worth to note that the three year difference in publication
date between this research and the previously discussed one
presented by Gao et al. did not change the fact that, the
author here still pointed out the existing lack of sufficient
number of research studies in this important field of
research. This is a particularly interesting fact considering
that both authors explicitly emphasized the rapidly growing
interest in emergency-oriented social media utilization,
which raises a question about why this interest did not
trigger a more active research of this field that would help to
facilitate its greater understanding.
Advantages and challenges with social media
utilization in emergency management have also been
discussed in one of the recently published research studies
(Anikeeva, Steenkamp & Arbon, 2015), which was largely
aligned with the arguments of the previous two research
studies. Namely, the authors pointed out that, when it comes
to accessing emergency-related information during disaster
events, the increasing popularity of social media has
resulted in people’s preference for informationally
diversified and socially inclusive social media rather than
traditional but somewhat one-sided means of information
sources (e.g. TV, radio, telephone). At the same time, just as
it was also argued by the aforementioned Alexander (2014),
people might be hesitant to trust information posted by other
users on social networks. However, unlike Alexander, the
authors here have suggested that official media outlets and
emergency agencies should keep up with current trends and
establish continuous social media presence, by allocating
financial and properly trained human resources for
maintaining sustainability and productiveness of that
presence as part of a sound strategy. Their arguments are
understandable and more than relevant, because considering
amount of circulated on social networks information,
significant amount of which might be speculative and
inaccurate, it could dangerously distort people’s
January 10, 2016
5
understanding of an emergency situation, unless reliable
sources of verified information are to be found on
conveniently accessible social media. Furthermore, similarly
to Alexander, this research study also outlined dependency
between people’s ability to access social media and such
factors as quality of network coverage and accessibility of
electricity, which might be very poor in remote areas or
severely damaged respectively. This could indeed leave out
people in certain regions unaccounted for when planning
emergency relief efforts and thereby lead to a life-
threatening delay in aid provision. Last but not least, the
authors outlined the fact that, unlike regular websites
employed by emergency organizations, social media
platforms tend to be capable of handling much higher
volumes of traffic. Inability to communicate or obtain
information from affected regions on time could endanger
many people’s lives, because as it was previously discussed
by Gao et al. (2011), timeliness is absolutely critical in fast
changing situations. That is why, once again supporting the
arguments of that same Gao et al it is important to carefully
consider both advantages and shortfalls of social media and
other emergency management techniques, combining them
in a strategically aligned and mutually strengthening
manner, making emergency management strategy with as
few potential weak spots as possible.
a. Overview of challenges and advantages of
social media in disaster management
In general, the reviewed research studies in this section
agreed on some absolutely essential benefits (e.g.
information versatility/availability). However, the authors
also agreed on the fact that while one of the major
advantages of information posted by users on social
networks being its diversity, there is a great challenge with
extraction of information that is actually relevant and
accurate. In the table 2 below, there is a summary of
advantages with using social media in emergency
management that were discussed in this section of the
literature review. TABLE II
SUMMARY OF ADVANTAGES WITH USING SOCIAL MEDIA IN EMERGENCY
MANAGEMENT
Advantages Papers mentioning them
Tag-based categorization of
information about crisis
events
Gao, Barbier & Goolsby,
2011; Alexander, 2014
Geographical clustering of
information about crisis
events
Gao, Barbier & Goolsby,
2011; Alexander, 2014
Immediate generation of
information about crisis
events
Gao, Barbier & Goolsby,
2011; Alexander, 2014
Making emergency
management socially
Alexander, 2014;
Anikeeva, Steenkamp &
inclusive (people have a say
in it)
Arbon, 2015
Versatility of collaboratively
created information
Gao, Barbier & Goolsby,
2011; Alexander, 2014;
Anikeeva, Steenkamp &
Arbon, 2015
Public availability of
information => raising
awareness about crises among
people
Gao, Barbier & Goolsby,
2011; Alexander, 2014;
Anikeeva, Steenkamp &
Arbon, 2015
Social networks being more
robust and prone to failure
during periods of high traffic
Anikeeva, Steenkamp &
Arbon, 2015
Attracting public attention for
launching appeals for
donations
Alexander, 2014
In the table 3 below, there is a summary of shortfall or
challenges with using social media in emergency
management that were discussed in this section.
TABLE III
SUMMARY OF SHORTFALLS WITH USING SOCIAL MEDIA IN EMERGENCY
MANAGEMENT
Shortfalls Papers mentioning
them
Lack of mechanisms for using
social media in
coordinated/synchronized manner
by multiple emergency agencies
Gao, Barbier &
Goolsby, 2011
Inaccurate, falsified or irrelevant
information
Gao, Barbier &
Goolsby, 2011;
Alexander, 2014;
Anikeeva, Steenkamp
& Arbon, 2015
Public availability of information
=> potential endangerment of
those involved in emergency
management on location.
Gao, Barbier &
Goolsby, 2011
Access to social networks is
dependent on availability of
electricity and the Internet in
particular
Alexander, 2014;
Anikeeva, Steenkamp
& Arbon, 2015
Weak social media presence of
trustworthy emergency agencies
and official media outlets =>
information on social networks
being untrustworthy
Anikeeva, Steenkamp
& Arbon, 2015
Disinformation on social
networks being legally
punishable => no one to be held
responsible for deliberate
disinformation
Anikeeva, Steenkamp
& Arbon, 2015
January 10, 2016
6
IV. Approaches for extraction and representation of
information
Practically all of the previously discussed research studies
have pointed out challenges with extraction of relevant and
accurate information from social media. There were,
however, several research studies that have proposed
peculiar and claimed to be effective approaches on how
emergency agencies could filter out masses of large chunks
of useless and potentially misleading information and rather
focus on information that can actually be used for
facilitation of decision making processes with contextual
relevance to a particular case.
An interesting approach was proposed in a not so recent
research study (Yin, Lampert, Cameron, Robinson, &
Power, 2012), where the issue with overwhelming number
of relevant and irrelevant user entries on social networks,
specifically Twitter as a widely popular information
broadcasting platform, could drastically decrease usefulness
of such information for its inclusion in highly sensitive
decision-making processes performed by emergency
agencies. So, the authors proposed a Twitter-based
emergency detection system, that would extract and
visualise information in real-time, making it readily
available for emergency agencies to make timely decisions
about appropriate disaster relief efforts, the importance of
which was previously outlined by Gao et al. (2011). The
proposed system would include data first being fetched from
social media pool, then processed by using numerus data-
mining methods, with each serving a specific purpose. After
data processing, the relevant information constituted by
filtered out tweets would be visually represented to end-
users or, in other words, emergency agencies. The data-
mining was made possible due to the fact that Twitter is
based on open-source software, where developers were free
to use its functionally, for example for retrieval of user
entries, for their own purposes, such as emergency
management. The processing of information included it
being classified by importance, which could assist
emergency agencies when prioritizing where their
emergency response would be needed most, based on
importance of an affected entity and severity of its
condition. This particular classification is indeed critical,
because functionality of vital infrastructure such as electric
stations, roads/bridges might be absolutely essential for
supporting people throughout a crisis situation, however,
information about such problems might be easily
overshadowed by less significant information that does not
have the same importance to a large number of people.
Another thing is that during crisis events people tend post
information about those event to social networks. This
would normally be reflected by a sudden increase in the
number of event-specific entries on Twitter. Detection of
such activities could help emergency organisations to react
to such crisis events in a timelier manner. So, the authors
proposed a burst-detection method that would deal with
such detection and alert when those bursts occur.
Furthermore, the authors also proposed a method which
would process information by applying a clustering
algorithm that would group information by event-specific
and currently relevant topics, constituted by pre-filtered
data, bringing out the ones that are actually accurate and
important to be considered when planning relief efforts.
This clustering could give a comprehensive and, at the same
time, usable understanding of important aspects of a
particular crisis event. Further exploration and thereby
deeper understanding of information collected from social
media could be achieved by analysing users’ entries by geo-
tags and mapping them onto a real map of an affected
region, giving a geographical distribution of tweets. Such
method was also proposed by the authors; however,
understandably, usefulness of such representation would
depend on whether those tweets actually had geo-tags
assigned to them, but it would certainly be a useful tool to
complement other information processing methods. Finally,
after processing, information would be visualised with
respect to how that information was processed. In general,
the proposed system offered a combination of different data-
mining methods for processing user-generated information
collected from Twitter and making it visually readable for
enhancing versatility of situation awareness of emergency
agencies and thereby allowing to make timely and relevant
decisions.
Another more recent research study has also proposed
an information extraction and representation approach
(Middleton, Middleton & Modafferi, 2014), but it was not
quite as advanced, in terms of its multi-layerdness, as the
one presented in the aforementioned research study.
Namely, the authors here focused on geographical
representation of entries made by users on Twitter.
However, aside from lower methodological versatility of
information representation, the precision and reliability of
representation of geographical distribution of tweets was
addressed on a more thorough level. So, rather than being
reliant on tweets being geotagged, which both in this and in
the previous research study was outlined as a fundamental
issue that could render geographical representation of user-
generated information highly unreliable and practically
useless, the authors here introduced a multi-layerdness into
extraction of geographical data about tweets. So, since only
1% of tweets were claimed to have geo-tags, the
information about location of an event would be extracted
even from text of each tweet. It was assumed that
information contained in those tweets would also include
location-specific references that after additional processing
could help to pin-point geographical location of an
emergency all the way to a street-level or a particular
building. So, to describe it abstractly without going too
much into technicalities, in the proposed approach the
tweets would be fetched, their potential geographic tokens
or names of entities would be extracted and associated with
their respective location by using multiple open-source
APIs, after which those tweets with now specified locations
(place, street or region) would be geospatially clustered and
mapped over a map of a chosen area, displaying in real-time
January 10, 2016
7
easily identifiable places with an increased activity of
emergency “signals” submissions by users. In general, the
effectiveness of the proposed approach is heavily depend on
the ability to interpret location-specific references contained
in tweets and match them to real location coordinates, where
some of those references could be expressed in local “slang”
by using non-formal denominations for streets and places.
Therefore, it would be essential for this approach to be used
only by emergency organizations that poses proper
understanding of the actual context of a region that they
operate in. Furthermore, full reliance on tweets having
location-specific references might not be the right way to do
it, because absence of such information could lead to crisis
events being potentially overlooked. This could be
especially problematic in small areas with only few people
who could submit information about a possible crisis or in
remote regions with a limited mobile network coverage and
thereby, once again, a smaller number of potential
information submitters. For that purpose, as it has been
already stated numerous times in this report, it’s essential to
think of this approach as a means to complement other
emergency management techniques in a strategically
aligned and mutually contributing manner, in which case it
would have a chance to actually produce benefits with lower
probability of causing undesirable or even life-threatening
outcomes.
While the aforementioned research study focused on
tweets’ geospatial data analysis with its consequent
geographical visualization, there was another recent
research study (Parsons, Atkinson, Simperl & Weal, 2015)
that focused on thematic analysis of user-generated
information collected from Twitter. Similarly to one of the
methods presented by the previously discussed Yin et al.
(2012), the authors here attempted to classify tweets by their
main theme, which in this case was defined by the type of
information that each tweet was purposed to communicate.
The type of communicated information was determined
based on keywords and the main piece of information
contained in a tweet as well as on the way a tweet was
conveyed. The thematic analysis of the sampled event-
related tweets performed by the proposed method further
indicated that types of tweets were associated with stages of
crisis development. So, as a crisis progressed from its
inception phase, peak of magnitude and till the point when it
started to settle, the information posted by users changed
from being about expression of condolences and general
warnings, then becoming valuable supportive information,
and finally maturing into being more focused and
contextually accurate instructive information. Furthermore,
previously discussed arguments (Middleton, Middleton &
Modafferi, 2014) about volume of tweets being significantly
greater during crisis situations did not quite find its
reflection in this research study, where authors did not
indicate a noticeably higher volume of emergency-related
tweets when UK was flooded in December 2013 and its
effects were just at the initial stages of their development,
but rather high volume of more or less accurate and thereby
usable tweets was indicated when the floods were already at
the peak of their magnitude till the point of their decrease in
intensity in February 2014. This peculiar but rational
observation does question those previously mentioned
arguments about information traffic being greater at the
initial stages of a crisis event, which would actually raise
certain doubts about reliability and accuracy of the
previously discussed burst-detecting method. Rationality of
this observation is justified by the fact that it is not unusual
or surprising to see people post more information about a
crisis as they got more aware of its details.
a. Overview of methods for analysis of user-
generated information on social networks
All three research studies that were reviewed in this sub-
section focused on extraction and analysis of emergency-
related information generated by users on Twitter, which
was stated practically by all authors as having a great
potential in being used by emergency agencies, due to
people using it primarily for broadcasting urgent
information, which in its turn would help to react to
emergencies in a timelier manner. The most frequently
proposed methods for analysis of information were based on
its thematic classification and geographical distribution. In
the following table 4, there is a summary of methods that
were proposed by the authors of the reviewed research
studies for extraction and analysis of information.
TABLE IV
SUMMARY OF INFORMATION EXTRACTION/ANALYSIS METHODS PROPOSED
IN THE REVIEWED RESEARCH STUDIES
Method for analysis Presented in papers
Burst-detection Yin, Lampert, Cameron,
Robinson, & Power, 2012
Geographical clustering /
mapping
Yin, Lampert, Cameron,
Robinson, & Power, 2012;
Middleton, Middleton &
Modafferi, 2014
Thematic clustering Yin, Lampert, Cameron,
Robinson, & Power, 2012;
Parsons, Atkinson, Simperl &
Weal, 2015
Impact classification Yin, Lampert, Cameron,
Robinson, & Power, 2012
January 10, 2016
8
V. Examples of real cases of social media application during
emergencies
There was one research study (Ehnis & Bunker, 2012),
where the authors attempted to explore experiences with
emergency-oriented social media application during the
floods that occurred in Queensland, Australia in 2011. In
that case, in order to interact with their community, the local
police service used such social networks as Twitter and
Facebook in a strategically aligned combination with each
other, where the former one was used for broadcasting a
short message about an issue with the link to a Facebook
account that contained a more detailed information about
that particular issue. It’s important to note that before the
actual floods, the police service had established a proper
social media presence, the importance of which was so
emphasized in the previously discussed research studies
(Alexander, 2014; Anikeeva, Steenkamp & Arbon, 2015).
Namely, just as it was suggested by Alexander (2014),
social media presence was established by assigning staff for
continuous maintenance of accounts on social networks, in
terms of updating information and spreading the word about
their accounts being available for people to turn to for
accessing reliable information during emergencies.
Furthermore, some people might not closely follow the
information posted by the police, so in order to keep people
updated on the latest emergency developments the police
posted summaries of the most important information on a
daily basis. Interestingly, even though the issue with people
being exposed to overwhelmingly inaccurate information on
social networks was addressed as efficiently as possible in
this case by the police service, it still appeared that the two-
sided communication with those people was neglected. In
other words, people were merely “listeners” or receivers of
information rather than contributors to the actual emergency
management process, which was actually the opposite of
what the police had expected from social media. This was
primarily due to the fact that the police service had a limited
staff, while the amount of user comments was simply
overwhelming for them to handle. Besides, the police did
not have any extraction and analysis techniques that could
help them to make use of those comments. So, regardless of
the encountered challenges, the police nonetheless took the
initiative and successfully attempted to enhance their ability
to deliver important information to a larger number of
people in their community during emergencies by utilizing
social media capabilities. Besides, the police used social
media in emergency management when this approach was
not so well investigated, therefore, their experiences were
undoubtedly scientifically valuable and enlightening.
There was another case of social media application that
was studied in one research study (Dashti, Palen, Heris,
Anderson, Anderson & Anderson, 2014). In that case, the
study was based in the context when Colorado was flooded
in 2013 after a series of severe storms, which resulted in
infrastructure (e.g. roads, bridges, sewage lines etc.) failing
to maintain its integrity and operational state, causing
effects of those floods to be long-lasting effects for several
communities in the region. So, since the state-funded
emergency agency that was responsible for provision of
relief efforts needed to get a detailed assessment about
where and what kind of efforts would have to be deployed
before their actual deployment, they used user-generated
information on social networks, such as Twitter and
YouTube, as sources of necessary intel, considering that on-
site systematic investigation was rendered impossible to
perform by emergency organizations without external
assistance of a community. The first step for collection of
relevant information was to find tweets that contained
keywords matching the ones that were defined through a
pre-study of contextually explorative information (e.g.
journals, social network entries etc.) to be representative of
the particular crisis event, which was done by using
Twitter’s publically available API. Afterwards, the collected
tweets were further analysed and grouped by the type of
information they provided, whether it was about a specific
location (geo-tag, place name) or it was visually more
descriptive information such as photos or videos. The
geospatial data was processed and mapped over the map of
the affected region, presenting the geographical distribution
and extent of emergency-inflicted damage that had to be
dealt with. The analysis was aimed to provide almost
immediate but at the same time useful insight information
that could help the emergency agency to provide optimized
relief efforts as soon as possible. However, the
productiveness of social media application in this case was
due to it not being used as a standalone solution. Namely, as
it was importantly stated by the authors, user-generated
information could not produce a holistic informational
coverage, due to the majority of tweets containing neither
visually descriptive information nor location-specific
information, all of which yelled the need for social media to
be used in mutually complementary combination with other
techniques (e.g. satellite imagery), just as it was suggested
in one of the previously discussed papers (Gao, Barbier &
Goolsby, 2011).
The chemical attack against civilians in Damascus in
Syria in 2013 attracted international attention mainly thanks
to its coverage through social media. So, the authors of one
research study (Rosman, Eisenkraft, Milk, Shiyovich,
Ophir, Shrot & Kassirer, 2014) attempted to investigate that
case and, by keyword-wise identifying relevant videos that
were uploaded to YouTube and then analysing the types of
injuries and symptoms of people recorded in those videos,
to prove that user-generated material on social networks
could be used to make medically accurate conclusions. This
might be useful when there is no other way to gain direct
access to victims and potentially assist local health-workers
by sharing knowledge with them about possible ways of
treatment, especially when talking about a war-torn country
the presence in which for on-site investigation could be
potentially life-threatening. So, in this study, the video
material was analysed, where classification of injuries and
symptoms was performed only by close observation of that
January 10, 2016
9
material. However, as it was importantly mentioned by
authors, the authenticity of posted information could not be
verified. So, on the one hand, the authors managed to
indicate some interesting symptoms that could explain what
caused the mass poisoning in the region, but accuracy of
those indications were largely based on the assumptions,
which is not acceptable for provision of accurate medical
advices. On the other hand, rather than fully relying on
those potentially misleading advices they could perhaps be
treated as potential clues for direction of investigation.
However, unless certain mechanisms are implemented for
verification of reliability and authenticity of information
posted on social networks, it is crucial to be very careful
when using video material posted on YouTube for making
conclusions about diagnosis and ways of treatment,
especially when there is no way to directly examine victims
or at least when information does not originate from reliable
sources. Therefore, rather than fully relying on such
material, it could be used in combination with traditional
techniques employed by a particular medical organization.
DISCUSSION AND CONCLUSION
Social media has been rapidly gaining popularity, becoming
popular communicational medium for many people all over
the world. People tend to post information about different
about various events, including the ones related to
emergencies. Therefore, such high volumes of information
can be used by emergency agencies when planning their
relief efforts, having a great potential in becoming a
productive component in emergency management.
However, the area of emergency-oriented social media
application is still at an embryonic stage of its
understanding and as it appeared from the analysis of
numerous research studies, it is still largely under-explored
and needs more research studies to scientifically explore its
various aspects. So, in this report, even though it did not aim
to produce new knowledge, it still attempted to explore
aspects of social media usage emergency management
during the last 4 years by analysing some of the currently
available relevant research studies. Namely, the purpose of
this report was to get an insight about currently known
advantages/challenges with usage of social media in
emergency management, approaches that were proposed to
be used for extraction and representation of relevant data as
well as this study sought to explore how social media was
used in real emergency situations.
I. What advantages and challenges arise from usage
of social media in emergency management?
From the three research studies that were analysed with the
purpose of identifying benefits and challenges with usage of
social media in emergency management, it appeared that the
reviewed studies published in the period from 2011 to 2015
had a common argument about one of the fundamental
advantages of social media. Namely, they pointed out its
ability to be used as a source of highly versatile and
valuable user-generated information that can be used to get
a sufficient and nearly immediately acquired amount of
details that can be used to respond as well as to recover
from crisis events with contextual accuracy. At the same
time, as with any Big Data, there is a major challenge with
making use of such high volumes of diverse data. This
challenge was explicitly emphasized not only by the three
research studies analysed in the respective section of this
report, but practically by all research studies chosen for the
review. All of the studies were explicit in their statement
that in order for crowd-sourced information available on
social networks to be useful it would be absolutely essential
to find the ways to not only extract relevant information,
while disregarding inaccurate or deliberately falsified one,
but also to represent it in a comprehensive and easily
readable manner so that emergency agencies could use it
without any outstanding hindrances. Another disadvantage
that was commonly brought up by paper discussed in the
respective section was that dependence of social media
network availability and technology, such as mobile
devices, being depend on electricity (for charging) could be
highly problematic especially during large-scale natural
disasters with high probability of causing severe damage to
such infrastructure. That is why, as it was frequently pointed
out by many of the reviewed research studies, one should
not be blinded by potential advantages of social media in
emergency management, but rather should realistically
understand its highly critical shortcomings and preferably
employ social media in strategically aligned and mutually
contributing combination with other emergency
management techniques, instead of fully relying on its
assumed effectiveness.
II. What types of approaches can be used for
extraction/representation of relevant information
from user-generated content on social networks?
As it appeared from the three research studies that were
reviewed in the respective section of the Literature Review,
one of the most frequently proposed methods for extraction
and consequent representation of relevant user-generated
content retrieved from social networks was based on the
concept of keyword-based identification of relevant user
entries with their consequent thematic clustering, where
information would be grouped by event- or subject specific
topics. The significance of topics would be defined by the
frequency of them being mentioned by people on social
networks. Such clustering could help to bring out the topics
that are actually important to be considered when planning
emergency relief efforts, because an emergency agency
would be able to account for what is really important during
decision making processes. The second most frequently
proposed method for making use of user-generated content
during emergencies was based on the concept of
geographical clustering, where relevant information would
be represented by its geographical distribution by using
specific attributes of users’ entries (e.g. location names, geo-
tags) as geospatial data. The geographical mapping of user-
January 10, 2016
10
generated information onto a real map of an affected region
could give a more visually descriptive understanding of the
extent of a crisis situation and thereby provide necessary
details for a better planning of relief efforts. Unfortunately,
such geographical clustering is not always possible to
perform, because frequently users’ entries on social
networks do not contain location-specific attributes needed
for their geographical representation.
III. How is social media used in emergency
management in real emergency situations?
The diversity of user generated-information being capable
of providing a great informational coverage of a particular
emergency situation with an unprecedented possibility to get
a highly valuable and insightful understanding of its internal
details was emphasized in all three case studies as a main
advantage of social media, which makes it potential in the
area of emergency management very important to be
properly understood. Furthermore, even though the authors
of the three case studies had a common opinion about an
advantage of social media application in emergency
management, they also had a common opinion about the
main challenges that arise from using such approach.
Namely, the previously discussed challenges with extraction
of relevant and accurate/reliable information were
encountered practically in all three of the studied case
studies. However, those challenges were not as critical as
the other research studies have pointed out, perhaps
primarily due to the fact that in the discussed cases social
media was mainly used in combination with other
emergency management techniques, while other research
studies discussed those challenges in the context of social
media being used as a self-sufficient emergency
management technique and a source of information. In
general, from the reviewed research studies it appeared that
there is still no standardized approach on how to extract and
represent user-generated information, which was even
further supported by the discussed case studies, where those
who dealt with disaster relief efforts had come up with their
own ways how to make use of crowdsourced information.
So, they either used custom-designed
extraction/representation techniques, such as geographical
and thematic clustering of information, or in one case, the
relief efforts were based on the simple power of
observation, which could however be very susceptible to
misleading/biased interpretation of information. On the
other hand, the diversity of techniques is understandable
because in different cases there might be a need for a
different type of data and thereby. In some cases, there is a
critical need to have geographical representation of
information, while in other cases (see chemical attack in
Syria) there is a need for a more graphical information.
IV. Limited and future work
Due to the very limited scope of this assignment, it was not
feasible nor necessary to review a larger number of research
studies about emergency-oriented social media application,
which resulted in this report having a somewhat
focused/limited theoretical understanding of this area of
research, but it would certainly be enlightening to get a
more complete understanding. This understanding is
particularly important, considering that it is still at an
embryonic stage, where potential of social media is yet to be
explored and thoroughly understood for its more productive
and seamless utilization as an integral component of a multi-
technique emergency management strategy or as a fully
capable and reliable stand-alone solution to such
management. In general, this report could be used as a first
step towards building a theoretical framework, upon which
an entirely different research study could be based for
further exploration of this important research area.
REFERENCES
Alexander, D. E. (2014). Social media in disaster risk reduction and crisis
management. Science and engineering ethics, 20(3), 717-733.
Anikeeva, O., Steenkamp, M., & Arbon, P. (2015). The future of social
media use during emergencies in Australia: Insights from the 2014 Australian and New Zealand Disaster and Emergency Management
conference social media workshop. Australian Journal of Emergency Management, 30(1).
Cavoukian, A., & Jonas, J. (2012). Privacy by design in the age of big data.
Information and Privacy Commissioner of Ontario, Canada.
Dashti, S., Palen, L., Heris, M. P., Anderson, K. M., Anderson, S., & Anderson, S. (2014). Supporting disaster reconnaissance with social media
data: a design-oriented case study of the 2013 colorado floods. Proc. of ISCRAM.
Denscombe, M. (2014). The good research guide: for small-scale social
research projects. McGraw-Hill Education (UK).
Ehnis, C., & Bunker, D. (2012, January). Social media in disaster response: Queensland Police Service-public engagement during the 2011 floods. In
ACIS 2012: Location, location, location: Proceedings of the 23rd
Australasian Conference on Information Systems 2012 (pp. 1-10). ACIS.
Gao, H., Barbier, G., & Goolsby, R. (2011). Harnessing the crowdsourcing power of social media for disaster relief. IEEE Intelligent Systems, (3), 10-
14.
Imran, M., Elbassuoni, S., Castillo, C., Diaz, F., & Meier, P. (2013, May). Practical extraction of disaster-relevant information from social media. In
Proceedings of the 22nd international conference on World Wide Web companion (pp. 1021-1024). International World Wide Web Conferences
Steering Committee.
Kavanaugh, A. L., Fox, E. A., Sheetz, S. D., Yang, S., Li, L. T., Shoemaker, D. J., ... & Xie, L. (2012). Social media use by government:
From the routine to the critical. Government Information Quarterly, 29(4),
480-491.
Kawtrakul, A., Mulasastra, I., Chanlekha, H., Rajbhandari, S., Prathumchai, K., Nagai, M., & Khunthong, V. (2012). Improving Disaster
Responsiveness using a Mix of Social Media and e-Government. In KEOD (pp. 423-426).
Lohr, S. (2012). The age of big data. New York Times, 11.
January 10, 2016
11
McClendon, S., & Robinson, A. C. (2013). Leveraging geospatially-
oriented social media communications in disaster response. International Journal of Information Systems for Crisis Response and Management
(IJISCRAM), 5(1), 22-40.
Middleton, S. E., Middleton, L., & Modafferi, S. (2014). Real-time crisis mapping of natural disasters using social media. Intelligent Systems, IEEE,
29(2), 9-17.
Merchant, R. M., Elmer, S., & Lurie, N. (2011). Integrating social media into emergency-preparedness efforts. New England Journal of Medicine,
365(4), 289-291.
Parsons, S., Atkinson, P. M., Simperl, E., & Weal, M. (2015, May). Thematically Analysing Social Network Content During Disasters Through
the Lens of the Disaster Management Lifecycle. In Proceedings of the 24th
International Conference on World Wide Web Companion (pp. 1221-1226). International World Wide Web Conferences Steering Committee.
Rosman, Y., Eisenkraft, A., Milk, N., Shiyovich, A., Ophir, N., Shrot, S., ...
& Kassirer, M. (2014). Lessons learned from the Syrian sarin attack:
evaluation of a clinical syndrome through social media. Annals of internal
medicine, 160(9), 644-648.
Mayer-Schönberger, V., & Cukier, K. (2013). Big data: A revolution that will transform how we live, work, and think. Houghton Mifflin Harcourt.
Tene, O., & Polonetsky, J. (2012). Privacy in the age of big data: a time for
big decisions. Stanford Law Review Online, 64, 63.
Yates, D., & Paquette, S. (2011). Emergency knowledge management and social media technologies: A case study of the 2010 Haitian earthquake.
International Journal of Information Management, 31(1), 6-13.
Yin, J., Lampert, A., Cameron, M., Robinson, B., & Power, R. (2012).
Using social media to enhance emergency situation awareness. IEEE Intelligent Systems, (6), 52-59.