R189 Ito, T. & Goldstein, K. (2015). Tohoku stories: Identifying happy themes of disaster relief Journal of International Society of Life Information Science, 33(1), 70-75.

Article (PDF Available) · March 2015with 43 Reads
How have different types of individuals and organizations conducted disaster relief and support activities that make people happy following the March 11, 2011 disasters? This paper utilizes a mixed methods survey of 1,659 respondents involved in community support activities to uncover the types of activities that make people happy. Themes were extracted using BigML from semi-open interviews with local residents, volunteers, and concerned individuals in the disaster regions. The results highlight the types of community support projects that generate positive experiences.
70 JournaJ ollnlernalional 90tJIely 01 lInlormalion 9tJIence (19lI9) ((で吉司
Intl. Soc. Life Info. Sci. L33No.1March 2015 11
The391h身僻'Posiumon life Inlormation 9c. '1 CD 11 )'J 11
March 14-152015Toho UniversiOmori Medical Cente
TokyoJapan lLr""" .)J
Tohoku Stories: Identifying Happy Themes of Disaster Relief
Takehiko IT01
and Keith GOLDSTEIN2
1mα ko University σbkyoJapan)
Zle Hebrew University 0/ Jerusalem (JerusalemIsrael)
Abstract: How have different types of individuals and organizations conducted disaster relief and
support activities that make people happy following the March 11 2011 disasters? This paper utilizes a
mixed methods survey of 1
659 respondents involved in community support activities to uncover the
types of activities that make people happy. Themes were extracted using BigML omsemi -open
interviews with local residentsvolunteersand concemed individuals in the disaster regions. The results
highlight the types of community support projects that generate positive experiences.
Keywords: disaster reliefinformationhappinessnarrativephilanthropytext mining
1. Introduction
This study identifies happy themes of disaster
relief following the Great East Japan Earthquake.
Beginning March 1
1th2011 the region of Northeast
Japan (Tohoku) suffered a chain reactiontriple disaster:
earthquaketsunamiand radiation leak. In the aftermath
of the disaster920000 volunteers assisted in the first
year alone (Yamomoto 2013). The number of non-profit
organizations has more than doubled to over 40000.
Numerous CSR (corporate social responsibility) projects
and intemational aid organizations assist in these
ty support and reconstruction activities. This
research examined the perceptions of people in Tohoku
about these activities. We asked an open ended question:
What are the activities of organizations or individuals
that assist the people in this community? We aimed to
both evaluate and leam about people's experiences with
such activities. Our goal was not only academic. Rather
we wished to assist organizations and people in the
disaster region: to publicize the good things that were
taking placeto enable them to further their workand to
motivate others to engage in future endeavors.
The Storytelling Project aimed to provide
recognition to the reconstruction support activities of
groups and individuals. In doing sowe also leamed a
great deal about themes and subthemes of disaster relief
activities. The themes were derived deductivelyusing
exploratory statistical methods. The theoretical model
identified provides the academic community with greater
understanding of how disaster relief activities are
suctured. More importantlyit provides adminisators
of disaster relief with examples of disaster support
activities that generate happy experiencesprojects that
can be supported and mimicked.
Takehiko ITO. shimoebi@gmail. com www.i to ehiko.com
Wako UniversityKanai 2160MachidaTokyo 1958585JAPAr呼.
2. Background
The Storytelling Project was initially developed by
GlobalGiving with the goal of creating areputation
system for philantmopy"whereby communication
between donorsorganizations and clients could be
facilitated by field based interviews (Maxson and
Kuraishi 2012). The initial storytelling project was
conducted in Kenya and U ganda and succeeded in
gathering approximately 58000 stories. Based on this
experienceMaxson and Kuraishi suggest technological
innovations for economical feedback collection. Owing
both to the feedback omthe Aica storytelling project
and the exmely tech-savvy nature of the environment
the Japan Storytelling Project incorporated a number of
technological innovations. Several types of mediums
were used for data collection: self-reported and
interview; phone surveycomputer surveyand paper
form; local scribeJapanese scribeand foreigner scribe.
Each form of data collection has its own deficiencies and
advantages. Respondents tended to provide very brief
stories that often reflected a desirability bias. on the
other handinterviews 0en emphasized current events.
The use of self-reports versus interview-based data is a
hotly contested subject (polkinghome 2005). While
behaviorists generally rejected introspective data
post-modemists thrived on it. Behaviorists claimed that
empirical truth was based on measurable phenomenon.
Post-modemists contended that there is no singular truth
and hence while perspectives may contradict one another
they are justifiable data. These stories represe
individuals' perceptions of the truth. They discuss
individuals' experiences as volunteers and recipients of
disaster relieThis research moderates between this feud
of subjective and 0ective thsdelineating a mixed
methods samplingunder the supposition that by using a
variety of data collection methodswe can arrive at a
JournaJ ollnlernalionaJ Society 01 lile Inlormalion Science
J. Intl. Soc.
Sci. 'L33No.1March 2015 11 (t:[J.J 11 71
The 39.助喧rmposiumon life Inlormalion Science IIC )'J 11
March 14-152015Toho UniversiOmoriMedical Cente
おか 0Japan
closer approximation of empirical truth.
When conducting research in a disaster region
there is an exceptional level of sensitivity concems. In
Japan individuals have a conception ofwhat is known as
Giri (responsibility for returning a favor). Henceif the
respondents were asked to provide a story of how
someone helped themthey would be left with the
feeling that they were weakdestite and should now
dwell on the fact that they owe a
vor. When conducting
researchanthropologists just as medical practitioners are
bound by the law of non-maleficence. Henceit was
incumbent upon the interviewers to ensure that
respondents would feel good about the stories they
shared. This deficiency in the research question was
identified early in a pilot study with aid workers and
academIcs. It was corrected by avoiding direct
questioning of how others helped them and focusing on
the activities that they are personally or presently
involved in. As a resultthe results discuss primarily
happy stories about disaster relief The task of
conducting interviews in a disaster region is especially
dicultsince the interviewer needs to gain the ust of
the peoplewho often feel that the research is only being
done for the sake of a universitya publicationor a
broadcastand not necessarily to help them. Often
scientific studies require structured interviews with
standardized questionsbut researchers have often
recognized the necessity to create informal conversations
for dacollection σkes 2011). Research on tsunami
devastated areas in particular requires using a
questioning route" by which semI-s uctured interviews
can lead to a thematic analysis (Fauci e
t. al 2012). This
questioning route focused first on the personal or present
storyand en followed up by asking them to identian
additional story of community support activities they
may have seen or heard about.
This research follows in the tradition of the
Grounded Theory (Glaser and Strauss 1967). No
pre-given hypotheses were used in the research. This
hypothesis development method dirs omditional
hypothesis testingin that no. given presumptions are
made at exist outside of the data. An exploratory
questioning amework enabled the respondents to take
part in hypothesis generation. This was accomplished by
using open-ended interview questionssuch as howare
you involved in support activities in your community?"
Howeverthe scribe often provided close-ended
questionssuch as how many people come to eCity
Hall meetings?" In order to facilitate the respondents'
understanding of the purpose of the questioning,白e
initial a
pproachoftenfocused 0nevent or
orgiza侃甜tiontha幻.t the scribe isviお凶s討叫iting忘:
about thetypes0fvolunteer activities 0ftl刷lIs
organization? Why isthisfes坑副tival important?" Etc.
Following this initial storythe respondent was
encouraged to provide an additional story that was not
related to the same organization.
3. Purpose
The purpose of the present study is to reveal
themes of community support activities and to uncover
specific pes of activities that make people happy by
analyzing the words used in narratives collected omthe
1apan Storytelling Project.
4. Methods
The data collection began in August 2013. The
stories were collected primarily by means of a form
containing 1215 questions. This form is based on the
stoηtelling forms used by GlobalGiving in A
son 2012). The stories were collected by both
bilingual and Japanese-only scribes. Bilingual scribes
slated interviewsreviewing details
of the anslation with the respondent. Japanese-only
scribes later anslated their stories with the bilingual
scribes. Stories were also occasionally recorded and
relevant sections anscribed. The storytelling interviews
attempted to focus on disaster reliefbut the respondent
was given liberty to provide any interesting story that
might be relevant. In generalthe respondent was simply
asked to tell a story about any pe of community
support activity. This can be something they have done
receivedor just heard about. Everyone has a storyand
every story is important.
For the preseSdy1659 stories were analyzed.
The most relevant question for the current research is
how does tllis story make you feel?" 411 respondents
answered this question (24.8%). 90 of them categorized
their story as happy" (2 1. 9%). 58% of the stories took
place in Miyagi Prefecture10% in Fukushima8% in
Iwateand the remaining 24% in various prefecres in
and around Tohoku. Owing to the large scale of support
from around the co四位yand evac瑚.tion to other
prefecresTohoku support extends beyond just the
:ffected areas.τ 'he high representation of Miyagi
occurred because it has the highest population of tsunami
:ffected p飽食cωresthe central role that Sendai
Ishinomakiand Kesennuma played as centers for the
researchand the large number of organizations that are
doing disaster relief activities there. Respondents were
recruited by snowball samplingas well as networking
with local non-profit organizations and individuals in the
disaster region. Emails and letters requesting
participation in the storytelling project were sent to 29
orgizations affiliated with GlobalGiving in Japan and
hundreds of non-affiliated NPOs. The Storytelling Team
(aμ escribes) also gave lec
and workshops at
72 JournaJ of International Society of life Information Science ρ'SlIS)
1. Intl.
Sci 1.3 3No.1 Marclz 2015 11 (c... μ│
mposiumon life Information Science 1
March 14-152015Toho UniversOmori Medical Cente
おか 0Japan
schools and universitiesafter which participants were
interviewed or completed forms. Most importal1 tly
scribes volunteered with numerous different
organizations around Tohokudoing evrything om
debris removal to child care. Participating in volunteer
activities enabled the scribe to earn contacts and trust
with local inforrnants. Not only was the stoη elling
project about collecting data ominformantsit was also
about glvmg something beyond academic recognition
back to those informants and communities. 1n summary
the Storytelling Team put a lot of muscle and sweat into
collecting the se interviews.
The data was analyzed using BigML (Dol1 aldson
and Donaldson 2012). BigML is a classification system
designed for big data sets. This model builds a tree
sirnilar to cluster analysiswhereby the correlations
tween words in the stories create distinct subsets that
will minimize the squared e
or (Donaldson et. al 2013)
Splits occur where one word is connected to two or more
words that provide optimal predicting value. The
algorithm improves the model by adjusting the
coecients using a stochastic gradient descent
(Carpenter 2008)whereby the predictive value of the
model is recomputed based on the partial results of
logistic regressionsbut only converges when the
maximum likelihood of the model approaches 1. Further
1. ActMtj担誌 for
95upprt 円lE! a門年
of inLemet
8; SiDrles om
quantitative analysis was performed by categorizing the
stories according to the themes and subthemesthen
examining the odds ratio of whether words that
correlated with each subtheme were included in stories
that made people feel happy.
5. Results
Fig. 1 shows the 1sults of a BigML analysis ofthe
entirdata se t. 11 themes of community support were
identified. Each theme was categorized ad hoc into three
types: 1) Stories (for) that discuss an activity that targets
a specific population or (at) that focus on a specific
location; 2) Stories (about) that relate directly to the
disaster or (企 om) that involve volunteers who came to
help; And 3) stories (by means of) that discuss the
technical aspects of providing support. Further analysis
ofacωal stories that represent each category revealed 15
subthemes. Since the exploratory analysis was
completed based on the words in each storythere is a
certain level of overlap in themes and subthemes but not
in the actual words used to define th m. Fig. 2 shows the
themes and subthemes for the analysis. These subthemes
are important as they are later used to explore
correlations with happiness.
" Acti',r IUes t
Radlti o- nCom:ert1 li
l忌苗与 ter
Fig. 1 BigML analysis of the entire data set with 11 themes and subthemes of community support
JournaJ of Inlernalional 'o cietyofl'nformalion Science (ISUでま 罰}
J. Intl. Soc. Life lnfo. Sci1.33No.1 March 2015 11 (t1.. μl
The391hmposiumon life Informalion Science Ilr "J11
March 14-152015Toho UniversOmori Medical Cente
あわ玖 Japan 1哩三;;;VJ
Kokoro (h 聞は)
For At About From By Means Of
CorrTenRadiation The Volun- Donat-
Disaster I teers
Themes and Subthemes of Stories about Community Support Activities
12 11l tin-i.E 一戸し.
J J/J;pl
Fig. 3 Odds Ratio of Stories That Make People Feel Happy Featuring Each Subtheme
i fh III-51
Fig. 4 Odds Ratio of Stories That Make People Feel Happy Featuring Each Word
74 JournaJ 0110'I'nationaJSolJIely 01 lile Inlormation SlJIence (ISLIi.Fヲ宮市
1. Intl. Soc. L供向(0. Sci. VoL33No.1March 2015 11 (t1.μl
The 39.
mposiumon life Inlormation Science II( ))J 11
rch 14-152015Toho UniversiOmori Medical Cente
おか 0Japan ll e' p:: I)J
Subthemes were developed based on a reading of
the common words that were equently used in the story
(Table 1). Certain words that were especially common
such as volunteeringassociated with almost all of the
subthemes and were excluded. The goal of the word
analysis was to create mutually exclusive subcategories
such that they could be compared. The words were
chosen by examining the equencies of word use and
coding them as binominal variables according to the
The 'equency of each word was computed based
on the total story list and the stories that made people
feel happy. Nextthe odds ratio was computed based on
the equency of a word being included in the happy
stories data set divided by its equency in the total data
set (Fig. 3). The results showat tools were most often
featured in happy stories. Tools included wordssuch as
intemetradionewspaperscomputersand media. Art
was also highly valued in happy stories. Art included
wordssuch as paintingcraftsmusicconcertsand
gardening. On the other handregulation and
collaboration make people the least happy. Regulation
included wordssuch as govemmenttaxand insurance.
Collaboration included wordssuch as interactmeeting
and integration. Certain subthemes that were not
expected to have happy features didsuch as evacuation.
Howeverother subthemes that were expected to be part
of happy stories were notsuch as sales. The results
highlight how a lot ofhappy stories involve people being
involved in community support activities that promote
hands-on activities and media.
An identical anais was completed on just the
words in order to look at specicexamples of happiness
related activities. The reverse analysis was impossible to
conductas there were a vast amount of words wi1no
mention of happiness. The results in Fig. 4 show that
treatment had the highest proportion of making people
feel happy. Despite the negative imagery associated with
mental health in Japan,住 ealent does manifest positive
6. Discussion
The Japan Storytelling Project is part of an
ongoing global e
urt to provide recognitionsupport
and information about community support activities. The
Storytelling Project revealed the importance of human
'ength to recovery and the role of volunteer
interventions. Happy themes of disaster relief are
associated with participants' activities in eatment and
紅t. Continuing research on this data set should compare
themes of community support between Japan and other
couniesextracting examples of activities that create
happy experiences. Based on these preliminary results
we emphasize aid workers to focus on artistic activities
and empowering people with toolssuch as computers
and language.
This research was funded by GlobalGiving and
coordinated by IsraAid and SP. We are tha
1to Prof.
Kathy Matsui at Seisen University for her encouraging
CarpenterB.: Lazy sparse stochastic gradient descent
for regularized multinomiallogistic regression. Alias-i
Inc.Tech. Rep1202008.
DonaldsonJ., and DonaldsonM. J.: Package 'Big ML '.
The Comprehensive RArchive Network2012.
DonaldsonJ. J. AshenfelterA., MartinF.,治問 oerdJ.
OrtegaJ. A., Parkerc., & Ara0M.: U. S. Patent
Application 13/667: 5422012.
FauciA. J., BoncianiM.and GuerraR.: Quality oflife
lnerability and resilience: A qualitative study of the
tsunami impact on the affected population of Sri Lanka.
Annali dell'Istituto Slψeriore di sanita48(2): 177188
GlaserB.and SussA.: The discovery 01 grounded
theoη'. London:Weiden:ti eld & Nicolson1967.
JormA. F., NakaneY., ChristensenH.YoshiokaK.,
GrithsK. M.,姐 dWataY.: Public beliefs about
treatment and outcome of mental disorders: a
comparison of Ausalia and Japan. BMC medicine
3(1): 122005.
MaxsonM.and KuraishiM.: Making use ofreputation
systems in philanropy. M. Hassan and M. Tovey
(EdsThe repuωtion society: How online opinions
are reshaping the offline worldMA: The MIT Press
Pp. 97-1072012.
MaxsonM.: The Real book lor St01evaluation. Global
Giving Foundation2012.
ParkesE.: Wait! I' m not a joumalist": Conducting
qualitative :ti eld research in post-disaster situation.
Graduate Journal 01 Asia-Pacific Studies7(2): 3045
PolkinghomeD. E.: Language and meaning: Data
collection in qualitative research. Journal 01
Counseling Psychology52(2): 1372005.
YamamotoK.: Volunteer activities in time of disaster in
Japan's highly information-oriented society. Journal 01
Earth Science and Engineering3: 1902022013.
JournaJol'lernationaJSociety 01 lile Inlormalion Sciwce(ISU
Intl. Soc. LInfo. Sci. L33No.1March 2015 11 (t!.. μ11 75
The 39.
rposium on life Inlormation Science II( 311
March 14-152015Toho University Omori Medical Cente
TokyoJapan llr-JJ
(Tohoku Stories: Idenfying Happy Themes of Disaster Reliet)
いとうたけひこ 1、キース・ゴールドスティン 2
(Takehiko ITO and Keith GOLDSTEIN)
要旨:地震、津波、そして原発事故と 2011 311 日に東日本大震災で被災した人々への災害
のテーマがあるはずである。本研究では 1659 人の調査に基づいて質的・量的分析をおこない、
2011 311 日の東日本大震災においては様々
メディアとして、グローパノレギピンクゃと JISP が協力
方法 ストーリーテリングの物語 1659 のうち、 411
人が感情について回答しており、うち 90 人が「幸
データの解析ソフト BigMLを用いて分析を行った。
付』という 11 のテーマが抽出され (Fig.l) 、それと
ルス][報告][規則]という 15 のサブテーマが示された
健」を構成する単語の例を Table 1に示した。
15 のサブテーマと幸福の感覚との関連の深さを
単語の頻度とのオッズ比により比較した (Fig.3)
伊藤武彦.1958585 東京都町岡市金井町2160 番地
shimoebi@gmail. com www.i totakehiko.com
ット JrラジオJr新聞Jrコンビュータ Jrメディ
ア」があった。 2番目に幸福感と関連の深いサブカ
ンサート Jrガーデニング」などの単語がその例で
る。 2番目に幸福感と縁遠いのは[協働]であり「交
また、幸福感と関連の高い単語を見ると (Fig.4)
高い順に「処置・治療Jr描画Jrコンビュータ Jr