Presentation on theme: "SoMe Lab Social Media UW #aag2013 Occupied Geographies Relational and Otherwise Josef Eckert, Department of Geography."— Presentation transcript:
SoMe Lab Social Media UW #aag2013 Occupied Geographies Relational and Otherwise Josef Eckert, Department of Geography Jeff Hemsley, Information School University of Washington April 11, 2013
SoMe Lab Social Media UW #aag2013 Occupy Wall Street Included both digital and urban spaces Localized, networked processes New social media tactics
SoMe Lab Social Media UW #aag2013 What role does place play within network structures of Twitter? Are actors both in place and on Twitter interacting with one another?
SoMe Lab Social Media UW #aag2013 Motivation Twitter and Social Network Analysis seem to be trending right now #overlyhonestmethods
SoMe Lab Social Media UW #aag2013 Motivation Urban processes are lived experiences (Lefebvre) These experiences are digitally mediated (Crampton, Leszczynski) The digital is inextricably part of urban life (Kitchen, Dodge, Zook, Graham)
SoMe Lab Social Media UW #aag2013 Motivation Twitter as a tactic for organization and protest (Gerbaudo) Decentralized, networked organizations of protest (Castells) A geographic focus on networks and the role they play in contentious politics (Leitner et. al, Nicholls) Moving beyond the geotag as a unit of place (Crampton et. al)
SoMe Lab Social Media UW #aag2013 A first cut at exploration…. Testing a “common sense” assumption: Are protesters that represent themselves as living in places with protest locations more likely to interact with others that also represent themselves as living in that place?
SoMe Lab Social Media UW #aag2013 Data Gathering 10/19/2011 – current day Gathered using streaming (now REST) API 300k – 1m tweets per day, 215 “keyterms” Have to slice the data
SoMe Lab Social Media UW #aag2013 That’s SoMe Toolkit!
SoMe Lab Social Media UW #aag2013 Data Preparation Six hashtags representative of protest locations: #occupyslc, #occupyportland, #occupyseattle, #occupyhouston, #occupydenver, and #occupyorlando (and #ows for fun) Reduced dataset to only those users with both in- and links (those interacting bi-directionally) Temporally bounded: 7-day (10/19/2011 – 10/25/2011) 30-day (10/19/2011 – 11/19/2011) #ows: 1-day (10/19/2011) – my computer is melting!
SoMe Lab Social Media UW #aag2013 Data Preparation: Users “in Place” Avoiding geotagging, attempting to use user-defined location Obtained a list of user-defined places for users participating in a hashtag – checking for alternative city matches (“SLC”) Used Regular Expression matching to determine if a user was “in place” for a given hashtag (e.g. “Salt Lake | Salt Lake City | SLC”)
Orlando, 30 days
Denver, 30 days
OWS, *1* day
SoMe Lab Social Media UW #aag2013 QAP Testing Matrices QAP uses random Monte Carlo iterations rather than inference metrics Tests against three null hypotheses: x1: Users with mutual ties do one another in a way that significantly differs from a random distribution x2: Users in a mutual place do not … x3: Users with more followers are Step 1: QAP Testing (Quadradic Assignment Problem) JeffJoe Jeff Joe Shawn
SoMe Lab Social Media UW #aag2013 EY ij = β 0 + β 1 X 1ij + β 2 X 2ij + β 3 X 3ij IV: Matrix, # Intercept DV: Mutual Tie Matrix DV: Users “in Place” Matrix DV: Follower Count Matrix Step 2: Fit QAP coefficents to OLS Regression
SoMe Lab Social Media UW #aag2013 X 1 (mutual tie)X 2 ("In Place")X 3 (Receiver's Followers)Adjusted R 2 Orlando 7-day Orlando 30-day Houston 7-day Houston 30-day Salt Lake City 7-day Salt Lake City 30-day Seattle 7-dayXXXX Seattle 30-day Denver 7-day Denver 30-day Portland 7-day Portland 30-day OWS 1-day Insignificance is more interesting….
SoMe Lab Social Media UW #aag2013 There’s still much to do The model fit could be better Analysis across multiple temporal slices Application to the other 154 locational hashtags Continued sensitivity testing to confirm that “place matters” in social media network construction. But how?
SoMe Lab Social Media UW #aag2013 future directions in visualization portland network, portland super-clique vignette Future Directions Portland, 7 days Cliques & Topic Modeling
SoMe Lab Social Media UW #aag2013 This research was made possible by: NSF Award # INSPIRE: Tools, Models, and Innovation Platforms for Research on Social Media Thank you! Questions and Suggestions?