1. Social Networks during crises: a
Study
Pablo A. Acuña
10/11/2011
References on these slides: http://bit.ly/twtstudy
2. Objective
Analyze the behavior of emergency governmental
agencies on social networks in crisis situations
Some questions that arise:
- Are social networks used in crises? If so,
- When are they used? (e.g. Pre-crisis, during crisis or post-crisis)
- What kind of information is published?
- Is there a communication with other users? If so,
- With who? General public and/or other agencies?
- Is it a one-way or two-way communication?
- Other…
References: http://bit.ly/twtstudy DEI Meeting – 10/nov/2011
3. Possible data
• Quantitative:
– # of publications
– # of retweets, reply/mentions, shares, likes, etc.
• Qualitative:
– Type of content published
– Meta-data of posts (date and time, location, etc.)
– Multimedia elements
– Communication and interaction with other accounts
References: http://bit.ly/twtstudy DEI Meeting – 10/nov/2011
4. Methodology
• Grounded theory:
An inductive qualitative research method which starts from a set
of data and aims to develop a theory. It consists on an iterative
process of data collection and analysis in order to extract a theory
[Glasser & Strauss, 1967; Corbin & Strauss, 2008].
– Stages:
• Open coding
• Development of concepts
• Grouping concepts into categories
• Formation of a theory
• Content analysis:
“Any technique for making inferences by objectively and
sistematically identifying specified characteristics of messages.”
[Holsti, 1969]
References: http://bit.ly/twtstudy DEI Meeting – 10/nov/2011
5. Methodology
• Coding
Coding is a process to analyze text content that “involves interacting
with data, making comparisons between data, and so on, and in doing
so, deriving concepts to stand for those data, then developing those
concepts in terms of their properties and dimensions” *Corbin et al.,
2008].
• Common procedure:
– Based on a set of collected data, coding can be a priori or emergent.
It usually requires different analysts to code the data (or a subset),
compare and consolidate the categories, reliability test on coded
data and repetitions of the procedure in order to achieve a reliable
analysis.
• Similar approaches:
– Action-object pairing with coding techniques [Zhang & Jansen, 2008]
– Noticing, collecting and thinking [Seidel, 1998]
References: http://bit.ly/twtstudy DEI Meeting – 10/nov/2011
6. Related studies Man-made/technological crises
Studies & Events Quant. Qual. Description
[Vieweg et al., 2010] Red River Flooding, 2009, USA Qualitative data coding from
X both events. Use of E-Data
Oklahoma Fires, 2009, USA Viewer for iterations
[Starbird & Palen, Red River Flooding, 2009, USA Coding of individual tweets
2010] X X and tweet streams and
Oklahoma Fires, 2009, USA analysis of RTs, URLs, etc.
[Hughes & Palen, Gustav Hurricane, 2008 Quantitative analysis of # of
2009] tweets, tweets per user,
Ike Hurricane, 2008 replies , URLs and new users.
X
DNC, 2008, USA
RNC, 2008, USA
[Starbird & Palen, Haiti Earthquake, 2010 A priori coding for analyzing a
2011]
X particular category (#TtT)
[Caragea et al., 2011] Haiti Earthquake, 2010 Classification by keywords and
X machine learning algorithm
[Bajpai & Jaiswal, Thailand protests, 2010 Content coding based on
2011]
X categories from literature
[Vieweg et al., 2008] Virginia Tech Shootings, 2007 X Empirical based on messages
[Gaffney, 2010] Iran Elections, 2009 X Quantitative study of twitter
References: http://bit.ly/twtstudy DEI Meeting – 10/nov/2011
7. Related studies (2)
Studies & Events Quant. Qual. Description
[Mendoza et al., Chile earthquake, 2010 Study of Retweet usage, and
2010]
X X empirical analysis of rumors
[Borge-Holthoefer et #15m movement, Spain Evolution and popularity of
al., 2011]
X twitter during #15m
[Qu et al., 2011] China earthquake, 2010 X X Distributed coding
General studies of social networks during crises situations:
*Tapia et al., 2011+ : “Seeking the trustworthy tweet” (short Interviews with organizations
name) X to discover mismatchs and
future directions
[Latonero & Shklovski, 2010+: “Emergency management and Interview with social media
social media evangelism”
X “innovator” and findings
Not related to crises:
- “Twitter use by the U.S. Congress” *Golbeck et al., 2009]: Qualitative analysis using
coding to categorize congressmen behavior on twitter
- “Twitter power: Tweets as Electronic word of mouth” *Jansen et al., 2009]:
Qualitative analysis of twitter publications with respect to brands, using a coding
technique and an action-object approach
References: http://bit.ly/twtstudy DEI Meeting – 10/nov/2011
8. Ideas & proposal
• A multi-step approach consisting on:
– Selecting a set of governmental agencies involved in crisis
management from different countries and with diverse
competences (with accounts on social networking sites)
• EMA, OEP, police, national guard, etc.
– Data collection using social networks APIs from past
events
– Iterative open-coding process to define initial categories
• Various coders?
• A priori coding based on literature (on/off-topic, crises stages, …)
– Validation and consolidation of categories
• If applies, definition of subcategories and relation of concepts
– Perform reliability tests
– Repeat the process, use a present or close event to
validate coding.
References: http://bit.ly/twtstudy DEI Meeting – 10/nov/2011
9. Possible events to study
• The crisis events to analyze may have the following
characteristics:
– Recent events from the past two years or less
– Events from different locations and reach/scope
– Events with different types based on the EM-DAT classification
(http://www.emdat.be)
References: http://bit.ly/twtstudy DEI Meeting – 10/nov/2011
10. Practical issues
• Technological
– Public availability of social network publications
depends on the site preferences and each account’s settings
– Time of publications and availability:
for instance, Twitter search API only allows retrieving recent
data with a maximum of 3200 tweets
• Resources
– The coding technique requires multiple coders to code content
iteratively, comparison of categories and reliability assessment
References: http://bit.ly/twtstudy DEI Meeting – 10/nov/2011
11. Questions?
Comments?
References: http://bit.ly/twtstudy DEI Meeting – 10/nov/2011
Hinweis der Redaktion
At theend, I wanttoknowthebehavior and usage of socialnetworks in crisis situationsfromgovernmental agencies in ordertoknowiftheinformationpublishedisusefulfortheconstruction of aninstitutional/collectivememory.