Whom to Coordinate With and How in Online Social Communities during Crisis Response Coordinaiton

!!

With!Whom!to!Coordinate,!Why!&!How!in!Social!Media!Communi8es!during!Crisis!Response!!

ISCRAM'2014,-May-19!Hemant-Purohit--(PhD%candidate,%Computational%Social%Sci.)%

Shreyansh%Bhatt%(CS),%Andrew%Hampton%(Psychology),%Valerie%Shalin%(Psychology),%%

Amit%Sheth,%(CS),%John%Flach%(Psychology)%

Outline!1.  Motivation!and!

Preliminaries!

2.  Identifying!Community-In>luencers:!With!Whom!to!Engage!on!Priority!Basis!!!

3.  Enabling!Faceted-Engagement-with!Community:!User!Categorization!by!Expertise!Summarization!

4.  Application-Interface-

5.  Conclusion-and!Future!Work!

Purohit!et!al.!2014:!With!Whom!to!Coordinate,!Why!&!How!in!Social!Media!Communities!during!Crisis!Response! 2!


Preliminaries!






Coordina8on!!A"Macro(level"Social"Phenomenon"

!  Coordination!!  Managing-dependencies!between!activities!(Malone!and!

Crowston,!1994)!

!  Examples!!  Managing!response!priorities!for!speciRic!needs!during!crisis!!  Prioritizing!whom!to!engage!with!about!concerns!for!a!brand!


Mo8va8on:!WHY!Coordina8on?!Consequence"of"Uncoordinated"Engagement"

!  2nd!disaster!to!be!managed:!formal!response!community!!  UnderSsupply!of!required!demands!!!  OverSsupply!of!not!required!supplies!

•  Hurricane!Sandy!example,!“Thanks,!but!no!thanks”,!NPR,!Jan!12!2013.!Story!links:!•  http://www.npr.org/

2013/01/09/168946170/thanksSbutSnoSthanksSwhenSpostSdisasterSdonationsSoverwhelm!


Second'costliest-hurricane-in-United-States-history-estimated-damage-$75-billion-

90'115-mph-winds-in-#Sandy-

State-of-Emergency-in-New-York--

285-people-killed-on-the-track-of-#Sandy--

750,000-without-power-(NY)--

Immense-devastation-and-Human-suffering-

Mo8va8on:!Why!Social!Media?!!Informal!Community!Responds!to!the!Disaster!Scale!!


Source:!http://www.hufRingtonpost.com/2012/10/30/hurricaneSsandySpowerSoutageSmapSinfographic_n_2044411.html!!

*20M!Tweets!in!1st!week!of!#Sandy!!!*Velocity:!200!tweets/sec.!!*Well,!not!everything!was!useful,!but!some!signals!!How!to!leverage?!

Characteris8cs!of!Informal!AdGhoc!Social!Media!Communi8es!

!  Presence!of!both!real,!onSground!informants!(Starbird,%Muzny,%and%Palen,%ISCRAM%2012)!and!virtual!volunteers!(Reuter,%Heger,%and%Pipek,%ISCRAM%2013)%

!  Potential!to!provide!both!timely!and!valuable!situational!awareness!(Vieweg,%Hughes,%Starbird,%and%Palen,%CHI%2010)%

!  !A-Big-challenge-to-Address:-!  Information!overload!problem!(Hiltz%and%Plotnick,%ISCRAM%2013)!


Where!and!What!to!give!aHen8on?!Whom!to!give!aHen8on?!!


Voluminous!!Data!and!Users!!!

Poten8al!Approaches!to!Address!Informa8on!Overload!

!  Content&driven:!Filter!content!of!the!messages!with!‘important!features’!!  DeRine!‘importance’!

!  Content%type:!damage,!donation,!etc.!(Imran,%Elbassuoni,%Castillo,%Diaz,%Meier,%ISCRAM%2013)%

!  Demand/request!and!Supply/offer!of!Help!!(Purohit,%Castillo,%Diaz,%Sheth%and%Meier,%FIRST%MONDAY%2014)!

!  People.content.network&driven:!Filter!source!of!the!information:!whose!voice!is!heard?!!  Better!for!passive!(content!reception)!and!active!(veriRication!and!

channeling)!engagement!


Proposal:!Coordina8on!Assistant!Framework!


EventSoriented!data!collection!

IdentiRication!of!whom!to!engage!with!

Categorization!of!user!types!for!faceted!engagement!!

Outcome:!Visual!Interface!to!provide!

engagement!!

Goal:!PeopleGContentGNetwork!view!!!!


Twitris!v3:!!http://twitris.knoesis.org/yolandastorm2013/networkTest/!!!


Preliminaries!






Whom!to!Engage!With!!  Known&reliable&sources:!Identify!onSground!informants!(Starbird%et%al.,%ISCRAM%2012)%!  Scope!for!improvement:!What!can!we!do!if!we!do!not!have!onS

ground!responders?!

!

!  Alternative&reliable&sources:!Identify!inRluencers!among!virtual!responders!whose!voice!is!heard!the!most!!!  For!better!user!categorization,!identify!inRluencers!per!need!type!

(e.g.,!medical,!shelter,!volunteering,!etc.)!!!


Approach:!Implicit!(Interac8on)!LinkGbased!

!  Classify!content!by!need!types!(e.g.,!medical,!shelter,!volunteering,!etc.)!

!  Create!user!interaction!network:!!  Node:!Users!!  Edge!(directed):-from-A-to-B,-if-A-‘interacts’(retweets/replies/mentions)-B-

!  Total!edge!weight:!weighted!aggregation!of!proportions!of!speciRic!interaction!type!

!  Apply!PageRank!on!user!interaction!network!!  How!to!Incorporate!features!in!PageRank!Analysis!

!  By!modifying!edgeSweights!!  By!Personalized!PageRank:!assigning!initial!probability!to!nodes!!  By!Hypothetical!Node!creation!

Purohit!et!al.!2014:!With!Whom!to!Coordinate,!Why!&!How!in!Social!Media!Communities!during!Crisis!Response! 14!(Purohit!et!al.,!ICWSM,!2012a)!

Relevant!Insights!of!this!Approach!to!Support!its!Use!in!our!Context!

!  On!various!brandSpage!event!oriented!community!datasets!of!similar!evolving!nature:!!  Implicit!Network!analysis!a!good!proxy!for!explicit!network!analysis!in!sparsely!connected!communities!!  Positive!correlation!between!rankings!!  Comparable!accuracy!to!Rind!inRluentials!(~80%)!

!  User!activity!and!its!inRluence!does!not!correlate!!



Preliminaries!






Problem!of!User!Engagement!with!Iden8fied!Influencers!


Mo8va8on:!(Generic)!Problem!of!User!Engagement!with!Influencers!


Huf>ingtonPost--•  No-relevance-cues-about-who-they-are-and-why-they-are-in>luencers-to-engage-with-

•  Problem-of-user-engagement-with-the-content-

Mo8va8on:!Tagline!Summariza8on!!


•  Now-I-can-see-why-…-

Senior!Media!Reporter!for!The!HufRington!Post!

Better!engagement!with!content!!

User!Data!Mining!

Analysis!&!Aggregation!

Tagline-

(Purohit!et!al.,!Social!Informatics,!2012b)!!

What!Data!to!Exploit?!

!  Meformer-!  Self!Descriptive!!  e.g.,!ProRile!Bio,!Content!of!the!user!posts!

!  Informer:-works-for-well'known-users-in-the-community-!  Others!describe!the!person:!available!in!external!knowledge!bases!!  e.g.,!Wikipedia,!IMDB,!Freebase!etc.!

20!Purohit!et!al.!2014:!With!Whom!to!Coordinate,!Why!&!How!in!Social!Media!Communities!during!Crisis!Response!

Extrac8ng!Occupa8on!PaHerns!for!Taglines:!!Occupa8on!Knowledge!bases!

!  Occupation!titles!collection!!!  from!Wikipedia!and!!!  US!Dept.!of!Labor!Statistics!Reports!

!  Given!a!user’s!proRile!bio!!  Spot!the!occupation!titles!in!the!NSgrams!of!the!bio!!  Create!candidate!tagline!by!joining!NSgrams!in!the!order!of!their!

appearance!in!proRile!bio!(Intuition:!User’s!own!described!order!of!what!he!is!expert!for!)!!

!  Create!Rinal!candidate!set!by!those!falling!in!the!UI!space!constraint!threshold.!

21!Purohit!et!al.!2014:!With!Whom!to!Coordinate,!Why!&!How!in!Social!Media!Communities!during!Crisis!Response!(Purohit!et!al.,!Social!Informatics,!2012b)!!

Facets!for!Engagement:!Classify!Exper8se!Tagline!Summaries!

!  ClassiRication!of!users!(summaries)!into!occupation!types!to!provide!classiRied!facets!for!engagement!!  Classes:!Journalism,!Medical,!Humanitarian,!etc.!!  ClassiRication!using!algorithm!desired;!A!ClassSlookup!Method:!

per!class!bagSofSwords!model!based!on!lexicons!from!Wikipedia!

!  Useful!for!enabling!coordinators!to!do!customized/!‘personalized’!engagement!as!per!class!type!!!  e.g.,!response!coordinator!may!want!to!communicate!with!

‘humanitarian’!professionals!differently!than!‘journalists’!!



Preliminaries!







Twitris!v3!demo:!!http://twitris.knoesis.org/yolandastorm2013/networkTest/!!!

Preliminary!Analysis:!data!collec8on!and!filtering!

!  Data!collection:!Philippines!Typhoon,!Nov.!11,!2013!(24hr)!


Preliminary!Analysis:!example!outcomes!from!random!sets!


Conclusion!!  Coordination!assistance!framework!to!enable!engagement!

between!formal!and!informal!communities!

!  Automatic!recommendation!for!whom!to!engage!with!!  Access!to!better!quality!information!content!!  Potential!agents!for!information!veriRication!and!coordination!!

!  Rumor!control!messages,!!!  Required!supplies!info!for!potential!donors!to!avoid!‘2nd!disaster’)!

!  Automatic!user!categorization!to!provide!faceted!engagement!!

!


Future!Work!!  Identifying!groups!to!engage!with!

!  By!modeling!low!collective!diverging!behavior!!  Preliminaries:%Purohit%&%Ruan%et%al.,%ICWSM%2014%

!  Identifying!who!is!likely!to!engage:!measuring!audience!engagement!level!for!potential!agent!use!

!  Background!knowledge!infusion!to!improve!categorization!of!candidates!for!engagement!


Acknowledgement!!  U.S.!NSF:!for%the%SoCS%grant%#IISW1111182%to%support%the%work%on%social%media%

based%coordination%during%emergencies%

!  SoCS!team:!coWauthors,%Twitris%team%&%Prof.%Srinivasan%Parthasarathy’s%OSU%group%

!  Active!Collaborators:!Drs.%Patrick%Meier%(QCRI),%Carlos%Castillo/ChaTo%(QCRI),%Fernando%Diaz%(Microsoft%Research)%

!  Colleagues:!Digital%Volunteers%from%various%organizationsW%Ushahidi,%StandBy%Task%Force,%info4Disasters,%Crisis%Mappers%network,%Humanity%Road,%etc.%%


Appendix!


Goal:!Data!to!Meaning!Transforma8on!to!assist!Coordina8on!


Coordination!of!Actions!

Citizen!Sensing!related!to!Crisis!on!Social!Media!!

Interpretation!of!sensed!data!via!context!categories!(e.g.,!need!types)!

Annotated!data!explicitly!specifying!meaning!(e.g.,!intent/behavior,!type,!whatSwhereSwhenSwhoSwhy!metadata)!


IdentifySMatchSEngage!(IME)!!Computing!Framework!

Crisis!Response!Coordina8on!Framework!


Edge-Wt (A , B) = TF*IDF TF = (w1.TotalAtoBRT + w2. TotalAtoBRP + w3. TotalAtoBMN) IDF = LOG ( (TotalNodes-1) / (w1.TotalART + w2.TotalARP + w3.TotalAMN) )

Union!Implicit!Network:!How!to!Model!Influence!Propaga8on!

!  Edge!weights!are!combined!according!to!TFIDF!type!normalization!to!propagate!weighted!inRluence!between!nodes!A!and!B!!!  Also!penalizing!A!users!who!just!keep!interacting!for!everything!!

A! B!

D!

C!

WORD! DOC!

DOC!

DOC!

E!

F!


Whom to Coordinate With and How in Online Social Communities during Crisis Response Coordinaiton

Technology