class: center, middle, inverse, title-slide .title[ # Social Networks Theories and Methods ] .subtitle[ ## How to find and talk about networks ] .author[ ###
James Hollway
] --- class: center, middle .pull-1[.circleon[![](https://globalquiz.org/media/pic/400/8340.jpg)]] .pull-1[.circleon[![](https://cdn.imgbin.com/14/23/6/imgbin-people-connected-tXALjkJraJY0PjsUKek18tXrH.jpg)]] .pull-1[.circleon[![](https://i.pinimg.com/originals/8a/08/f1/8a08f1d673c8b64ddb067de8d8d879d8.jpg)]] --- # Course sessions .pull-left[ .polaroid[![:scale 45%](https://panarchic.ch/images/team/james_hollway.jpg)] **Lectures**, [James Hollway](mailto:james.hollway@graduateinstitute.ch) Wednesdays, 14-16, S7 Office hours Fridays, 10-12 1. Mix of conceptual and practical 1. Complementary to readings ] -- .pull-left[ .polaroid[![:scale 45%](https://external-content.duckduckgo.com/iu/?u=https%3A%2F%2Fjdi-switzerland.weebly.com%2Fuploads%2F9%2F8%2F0%2F2%2F98020646%2Fpublished%2Fdsc-0484.jpg%3F1603886744&f=1&nofb=1)] **Lab overflows**, [Zakaria Imessaoudene](mailto:zakaria.imessaoudene@graduateinstitute.ch) Tuesdays, 14-16, tbc 1. Deepening comprehension 1. DIY experience ] --- background-image: url(https://media.giphy.com/media/b7sNm0aR2EOqI/giphy.gif) background-size: contain --- class: center, middle # Definitions .pull-1[.circleon[![](https://globalquiz.org/media/pic/400/8340.jpg)]] .pull-1[.circleoff[![](https://cdn.imgbin.com/14/23/6/imgbin-people-connected-tXALjkJraJY0PjsUKek18tXrH.jpg)]] .pull-1[.circleoff[![](https://i.pinimg.com/originals/8a/08/f1/8a08f1d673c8b64ddb067de8d8d879d8.jpg)]] --- ## Seven Bridges of Königsberg .pull-left-1[ - Königsberg founded in 1255 by Teutonic Order - Became a wealthy port set around two islands on the Pregel River - Citizens spent Sunday afternoons walking its seven bridges - They devised a puzzle about how to cross all seven bridges only once in a single walk... ] .pull-right-2[![:scale 120%](https://upload.wikimedia.org/wikipedia/commons/5/5d/Konigsberg_bridges.png)] --- ## Lisez Euler... -- .pull-left[ - Reply to mayor of Danzig: - “. . . I do not understand why you expect a mathematician to produce [a solution...], for the solution is based on reason alone, and its discovery .red[does not depend on any math]ematical principle.” - Yet, later that year he admitted to an Italian mathematician: - “This question is so banal, but seemed to me worthy of attention in that .red[[neither] geometry, nor algebra, nor even the art of counting was sufficient] to solve it.” - He saw the problem related to what Leibniz called .red[geometria situs], or the geometry of position... ] .pull-right[![:scale 80%](https://www.scienceabc.com/wp-content/uploads/ext-www.scienceabc.com/wp-content/uploads/2017/12/Leonhard-Euler.jpg-.jpg)] --- ## Bad for Königsberg, Great for graph theory .center[![:scale 55%](https://divisbyzero.files.wordpress.com/2008/09/konigsberggraph.jpg) ] - In 1735, Euler founded .red[graph theory] by proving no solution to the problem - Proof consisted of two steps: - First, route inside each land mass irrelevant, only the connections, meaning he could *abstract* to a graph of .red[nodes] connected by .red[ties] - Second, in unique path no more than 2 nodes should have an odd .red[degree] (#ties) -- if enter, also need to leave, so intermediate nodes must be even - Since all nodes had odd degrees, no way of solving the problem --- ## What is/are “social networks”? -- .pull-left-1[![:scale 65%](https://thedigitalhealthsociety.com/wp-content/uploads/2019/05/Facebook-1.png)] .pull-right-2[ - Not **a** social network - not a type of actor, but structures ] -- .pull-left-2[ - Not a social **network** - not a type of structure, but any structure ] .pull-right-1[![](https://databigandsmalldotcom.files.wordpress.com/2016/04/hierarchymktnetwork.png)] -- .pull-left-1[![](https://images05.military.com/sites/default/files/styles/full/public/2018-08/inigo_montoya_1200x800.jpg?itok=D1PNX-Hv)] .pull-right-2[ - Not a **social** network - not any structure of a specific type of relation, but structures of any relations ] --- ## An abbreviated history of social networks <img src="ISON_L1b_Networks_files/figure-html/unnamed-chunk-1-1.png" width="1008" /> <!-- -- --> <!-- Social Network Analysis (SNA) interdisciplinary with basis in sociology, anthropology, mathematics, social psychology, --> <!-- public policy, and increasingly political science and international relations. --> <!-- .footnote[See Freeman (2004)] --> --- class: middle .pull-left[ .center[ ![](https://upload.wikimedia.org/wikipedia/commons/thumb/6/60/Graph_betweenness.svg/220px-Graph_betweenness.svg.png) ]] .pull-right[ Social networks concerns the abstraction, theorising, and analysis of relations into structures. Relations can be key to understanding (changes in) - macro-structures or micro-structures - the distribution of attributes within those structures - or even what those structures consist of ] --- class: center, middle ![](https://www.seekpng.com/png/detail/783-7835409_contemporary-social-network-analysis-was-formalized-analyze-the.png) --- ## Terminology - .red[Graph] or .red[network], `\(G = (V, E)\)` - .red[Vertices] or .red[nodes] from a .red[node set] or .red[mode], `\(V = \{a,b,c,...\}\)` - .red[Edges] or .red[ties] (also links, lines, connections, arcs), `\(E = \{\{a,b\}, \{b,c\}, ...\}\)` -- .pull-left-1[ - (un)directed - (un)weighted - (un)labelled - (un)signed - simple/complex - multiplex - multimodal - multilevel - cross-sectional/longitudinal/event ] .pull-right-2[ ![](https://www.researchgate.net/publication/347300725/figure/fig1/AS:969208926044162@1608088823984/Different-types-of-graphs-and-their-corresponding-adjacency-matrix-representations-The.ppm) ] --- .pull-left-3[ - .red[graphs] pretty and easy to interpret - quickly difficult to discern though and results may vary... ] .center[![:scale 55%](ISON_L1_Networks_files/figure-html/GraphMat.png)] .pull-left[ - .red[edgelists] 2(+) (ordered) columns of numbered/labelled pairs - easy in Excel, incl. attributes, and memory efficient - more complicated data, statistics, and analysis difficult though ] .pull-right[ - .red[matrices]' rows are senders and cols recipients - memory inefficient for sparse networks and somewhat incomprehensible - encodes all relational info and quick, flexible analysis ] ??? - dimensions - density - outdegrees - indegrees - isolates - reciprocity - transitivity --- ## Let's try a few... .pull-1[ <img src="ISON_L1b_Networks_files/figure-html/unnamed-chunk-2-1.png" width="504" /> ] .pull-1[ ``` ## [,1] [,2] [,3] [,4] ## [1,] 0 0 1 0 ## [2,] 1 0 0 0 ## [3,] 1 0 1 1 ## [4,] 0 1 1 1 ``` ] .pull-1[ ``` ## from to ## 1 1 2 ## 2 1 4 ## 3 2 3 ## 4 2 1 ## 5 3 4 ## 6 4 1 ``` ] --- class: middle ## Describing networks .pull-left[ ![](ISON_L1b_Networks_files/figure-html/plotting-1.png) ] -- .pull-1[ - _what could this network be of?_ - type/format of network - dimensions/density/diameter ] -- .pull-1[ - _describe the structure..._ - centralisation/core - components/communities ] -- .pull-1[ - _describe the position of nodes..._ - isolates/centrality - brokerage and other roles ] -- .pull-1[ - _describe the situation of ties..._ - reciprocal or asymmetric - transitive/cyclical embedded ] ??? - same group of actors (some composition change allowed) - same relational variable (states, not events) - some, but not too much change... The data used are from the Children of Immigrants Study, (c) MZES Mannheim, Manfred Kalter. - social network ties are costly? (low outdegree) - individuals form and maintain reciprocal ties? - transitivity leads to clustering - status hierarchy shapes friendship networks (ties to popular actors) - gender/ethnic homophily? --- class: center, middle # Design .pull-1[.circleoff[![](https://globalquiz.org/media/pic/400/8340.jpg)]] .pull-1[.circleon[![](https://cdn.imgbin.com/14/23/6/imgbin-people-connected-tXALjkJraJY0PjsUKek18tXrH.jpg)]] .pull-1[.circleoff[![](https://i.pinimg.com/originals/8a/08/f1/8a08f1d673c8b64ddb067de8d8d879d8.jpg)]] --- ## Levels of analysis .pull-left[ ![](ISON_L1b_Networks_files/figure-html/plotting-1.png) ] -- .pull-right[ - .red[Network level] `\((O=1)\)` - e.g. is this network centralized? - .red[Dyad level] `\((O = n(n-1)/2)\)` - e.g. are similar nodes connected? - .red[Node level] `\((O = n)\)` - e.g. are some nodes more popular than others? ] --- class: middle ## Types of analysis .center[ |IV..DV |Node |Ties |Network | |:-------|:------------------------------------------|:---------------------------------------------------------|:---------------------------------------------| |Node |e.g. does centrality drive success? |e.g. are friendships formed through attraction/homophily? |e.g. are networks clustered around diversity? | |Ties |e.g. do unequal exchanges drive success? |e.g. are friendships formed through propinquity? |e.g. are networks clustered around balance? | |Network |e.g. do brokerage positions drive success? |e.g. are friendships formed through transitivity? |e.g. are networks coevolving into clustering? | ] .footnote[See also Agneessens (2021)] <!-- - Network as .red[independent variable] --> <!-- - i.e. (how) do ties explain attributes? --> <!-- - e.g. is the network structure responsible for who gets what? --> <!-- - Network as .red[dependent variable] --> <!-- - i.e. (how) do attributes explain ties? --> <!-- - e.g. is who has what responsible for the network structure? --> <!-- - Network as .red[interdependent variable] --> <!-- - i.e. (how) do attributes and ties coevolve? --> <!-- - e.g. are there positive or negative feedback loops? --> --- ## Dynamic analysis ![](ISON_L1b_Networks_files/figure-html/plotboth-1.png) - How do we get from `\(t_1\)` to `\(t_2\)`? - e.g. is it popularity, reciprocity, transitivity, or homophily? --- ## Ego networks Usually .red[name generator], .red[name interpreter], and sometimes .red[name interrelater]. .center[![:scale 65%](https://ars.els-cdn.com/content/image/3-s2.0-B9780123822291000114-f11-02-9780123822291.jpg)] .pull-left[ + can analyse larger networks (sampling) + compatible with traditional methods (and designs) + can study intersecting social circles (focal nodes) ] .pull-right[ - demands traditional assumptions (exchangability) - missing structural data (betweenness) - recall often biased toward longer-term interactions .small[(Freeman, Romney, & Freeman 1987)] - recall inaccurate when reporting perceptions of relationships between third parties .small[(McEvily 2014)] ] ??? _Name generator_: ego (survey respondent) names important peers on one or more dimensions, e.g. ask for frequency of contact? valence? _Name interpreter_: ego provides information about these peers (alters' attributes) and the ties that bind them (tie attributes) _Name interrelater_: ego provides information about the potential ties among peers, e.g. how frequently do these two exchange information? For more see: ![](https://images-na.ssl-images-amazon.com/images/I/51dI6OU+pKL._SX331_BO1,204,203,200_.jpg) ![](http://assets.cambridge.org/97811071/31439/cover/9781107131439.jpg) --- .pull-left[ ## Cognitive social structure - Also known as a .red[CSS] - Asks respondents to report on the structure of relations of others in the network from their point of view - Then the similarities and discrepancies between the network as reported by individuals in the network is analysed - Or the impacts of perceived social structure on outcome is considered, e.g. perceived influence/ popularity .footnote[See Krackhardt 1987; Knoke 1998; Knoke et al. 2019] ] .pull-right[ ![](https://www.researchgate.net/publication/282047850/figure/fig1/AS:614321356824596@1523477033705/Cognitive-social-structure-components.png) ] --- ## Whole networks Respondents presented with a .red[roster] (complete list of individuals in population of interest; Marsden 1990) .center[![:scale 36%](http://communication.iresearchnet.com/wp-content/uploads/2019/11/Social-Networks-fig-1.jpg)] .pull-left[ + reduces recall bias stemming from unreliably recalling interaction partners’ names + reduces perception bias cos reports cross-checked, or not even requested (though see Boda et al) ] .pull-right[ - only possible for small networks or networks where relational data publicly recorded - larger rosters also become unreliable (Pustejovsky & Spillane 2009) ] --- ## Snowball sampling .pull-left[ .center[![:scale 80%](https://www.researchgate.net/profile/Stephen-Borgatti/publication/227085871/figure/fig1/AS:302299076415488@1449085124312/HRS-study-macro-network-largest-connected-component-by-ethnicity-Total-n-193-q.png)] Needle sharing among IV drug users in Hartford, CT ] .pull-right[ some .red[seed nodes] recommend others, who recommend more, usw; a non-probability sampling method that concentrates on most accessible part of the population - cheap, simple, and cost-efficient - easy way into hard-to-reach (small, covert) groups But - oversamples most public and well connected (more often seeds and lie on more recruitment paths) and may miss isolated/weakly connected individuals/groups - biases towards particular network structures (like high degree) and not representative Respondent-driven sampling (RDS) aims to mitigate some of these concerns by weighting the sample to compensate for non-random recruitment patterns ] --- ## Boundaries .center[![:scale 30%](ISON_L1b_Networks_files/figure-html/plotting-1.png)] -- .pull-1[ - .red[Relational approach] (i.e. connected): - e.g. all “relations” connected socially to main/ seed individuals ] .pull-1[ - .red[Event-based approach] (i.e. attendance): - e.g. all “regulars” that go to the beach each day for 3 days ] .pull-1[ - .red[Positional approach] (i.e. characteristics): - e.g. all “employees” employed by an organization ] ??? - Ego, e.g. Bott - CSS, e.g. Krackhardt - Snowball, e.g. Kaplan - Complete, e.g. Snijders - Cross sectional networks (MAN/MNA) - Longitudinal (panels) - Continuous (events) --- ## Sources .pull-left[ #### Self-reports - Surveys .small[(e.g. Hogan et al 2016)] - using e.g. [Network Canvas](https://networkcanvas.com) - name generators vs rosters - Interviews .small[(e.g. Bellotti 2014)] #### Constructions - Web-scraping - Chrome or Firefox extensions - Rcrawler, rvest and vosonSML packages - Python (e.g. to access Twitter API) - Manual/automatic text coding - Considerable literature-based datasets - Can extract relationships (e.g. similarities) in or across documents ] .pull-right[ #### Observation - Participant observation .small[(e.g. Wyatt et al 2011)] - RFID badges .small[(e.g. Elmer et al 2018)] #### Archives - Individual - Diary research .small[(e.g. Fu 2008)] - Historical records .small[(e.g. Padgett & Ansell 1993)] - Social media .small[(e.g. Golder et al 2007)] - Communication logs .small[(e.g. Goldberg et al 2016)] - Organizational - Publication or patent records .small[(e.g. Lazega et al 2008, Goetze 2010)] - Agreement data between countries .small[(e.g. Hollway and Koskinen 2016)] ] ??? - Network Canvas - beta/open source - egoweb 2.0 (David Kennedy at Rand) - free/open source - RedCap - one week free trial - Limesurvey - free-`$$`; installing on Institute server requires IT - Qualtrics - limited account free; advanced features `$$` - Can include a JavaScript autocomplete function (Timon Elmer) - ONA Surveys - limited export free, range of export formats by subscription - Voxco (Jason Diasno) - `$$` - Socioworks - `$$` - Kronenwett & Adolphs - `$$$` - DatStat Illume Next (Michelle Loxley) --- background-color: #96BCC6 .pull-left[ ## Ethical concerns - Consider study ethics from initial research design - IRB reviews - Report in publications - Study participation needs - voluntary, informed consent - minimal intrusiveness - appropriate protections of privacy - One of the main advantages of networks is also a source of greatest ethical concern... - _what is it?_ ] .pull-right[![](https://talkroute.com/wp-content/uploads/2016/02/ethical-business-talkroute-e1456238028542.jpg)] --- .pull-left[.center[![:scale 75%](https://www.researchgate.net/profile/Dorothy-Espelage/publication/314204848/figure/fig1/AS:643801722912768@1530505700607/Visualization-of-missing-social-network-data-Nodes-represent-individuals-and-directed.png)]] .pull-right[ ## Missing data - Unfortunately, even a small fraction of missing observations could significantly bias network structure - A single non-response in a large survey is `\(1\)` missing observation - A single non-response in networks is `\(n-1\)` missing observations - And what do you do with ties with missing targets? - Moreover, often not missing at random... .footnote[de la Haye 2017] ] --- ## Demanding data .pull-left[ - Networks is demanding of data - But true of all attempts at providing persuasive evidence for causality - Where data comes from is crucial because: - how meaningful your descriptive or inferential conclusions depends on tie-data being meaningful - So, with an eye to your final report, think about where you might get as meaningful, complete, whole-network data, ideally over time (as granular as possible) and with plenty of additional data on other ties and other attributes... - No big deal, right? ] .pull-right[![](https://memegenerator.net/img/instances/75304621.jpg)] --- class: center, middle # Data .pull-1[.circleoff[![](https://globalquiz.org/media/pic/400/8340.jpg)]] .pull-1[.circleoff[![](https://cdn.imgbin.com/14/23/6/imgbin-people-connected-tXALjkJraJY0PjsUKek18tXrH.jpg)]] .pull-1[.circleon[![](https://i.pinimg.com/originals/8a/08/f1/8a08f1d673c8b64ddb067de8d8d879d8.jpg)]] --- ## Most important lesson .pull-left[![](https://cdn.imgbin.com/14/23/6/imgbin-people-connected-tXALjkJraJY0PjsUKek18tXrH.jpg)] .pull-right[ - Most important takeaway: Always ask what a tie **means** - What does _“already knowing”_ mean? - How can we interpret different weights/ structures? ] --- background-color: #F6D0B9 ## Tie content .pull-left-1[ _Actions_ - talks to - sells to - gives aid/advice to - sleeps with _Cognitive-affective_ - likes - knows - despises - recognises _Role-based_ - kinship: brother of, daughter of - social: friend of, competitor of - organisational: boss of, teacher of ] .pull-right-2[![](https://ideas.ted.com/wp-content/uploads/sites/3/2018/06/featured_art_istock1.jpg)] -- .pull-right-2[ _Remember_: Ties can be directed/undirected, weighted/unweighted, signed/unsigned, labelled/unlabelled, simple/complex, uniplex/multiplex, unimodal/multimodal, etc..] --- background-color: black class: inverse .pull-left[ ## Pipes .center[![](https://116kxc3yni1m23llck48qpul-wpengine.netdna-ssl.com/wp-content/uploads/2019/08/PVC-Plumbing-Pipes-Best-Materials-for-Water-Pipes.jpg)] - network ties often seen as “plumbing” through which “stuff” flows - stuff can be ideas, “capital”, etc - micro questions about position, inequality, etc - macro questions about network intervention, resilience, etc ] -- .pull-right[ ## Prisms .center[![:scale 52%](https://wallpaperaccess.com/full/2316694.jpg)] - network ties also serve as informational “cues” - cues can be to identity, preferences, etc - micro questions about how local networks affect perception - macro questions about emergent cultures from certain network topologies ] --- ## Networks and culture .pull-left-1[![:scale 75%](https://pup-assets.imgix.net/onix/images/9780691137155.jpg)] .pull-right-2[ - structure long linked with culture (e.g. Berger and Luckmann, Bourdieu, Douglas) - 1970s “breakthrough” established social networks as method of structural analysis distinct from cultural analysis and Parsonian normative theory (Blau, White et al 1976) - today tentative reconciliation in different directions: - ties affect culture (Erickson, Krackhardt, Kilduff, Carley, DiMaggio) - culture affects ties (symbolic interactionalism/ cultural sociology of Emirbayer and Goodwin, Lizardo, Daisy) - a duality understanding (White 1992, Breiger, Fuhse, Mische) ] .footnote[See Mische (2011); Fuhse (2021)] --- class: middle ## Network theory Model | Social Capital | Social Homogeneity --------|---------|--------- | (explaining success) | (explaining choice) **Network flow model** | Capitalization | Contagion (ties as pipes) | (e.g. weak ties) | (e.g. diffusion) **Network coordination model** | Cooperation | Convergence (ties as bonds) | (e.g. exchange) | (e.g. equivalence) .footnote[Borgatti and Halgin (2011)] --- ## Multilevel/multimodal meaning .pull-left[ - Social-ecological networks: Bodin et al 2006 - Social-semantic networks: Roth et al 2010 - Social-organizational networks: Hollway et al 2017 - Social-political networks: Knoke et al 2021 .center[![](http://assets.cambridge.org/97811088/33509/cover/9781108833509.jpg)] ] .pull-right[ > “Political actors are creative and resourceful [...] the fields, arenas, or social spaces in which political contestation takes place are never unidimensional but contain multiple types of actors and relations. Political actors turn to or create new categories of cooperation or contestation in their efforts to build resources or flank those with whom they disagree.” > ~ Knoke et al (2021): 1. ] --- ## Social Networks **Assumption**: social life is associative, and relations are meaningful **Premise**: how social entities are connected matters **Argument**: more interdependent and contextual than traditional quantitative or qualitative work - By taking context and dependencies into account, even making them central to the explanation, networks goes beyond traditional statistics - By using graph theoretic concepts and formal measures, even scaling them to large settings, networks goes beyond traditional case studies **Promise**: to help understand social, relational life --- .pull-left[ ### Don’t “do networks” to - join .red[hot research area] (though it is) - present .red[fancy pictures] (seldom enough) - present .red[fancy analysis] (depends on data quality) - .red[explain more variance] (not simply an add-on) - .red[explain everything] (empirical settings are messy) - use .red[big data] (RQ and theory relationship required) - use .red[networks expertise] (clear motivation required) ] -- .pull-right[ ### Do “do networks” to - .red[describe social structures], both local and global - .red[identify individuals] in specal positions - correlate .red[positions with individual outcomes] - correlate .red[structures with global outcomes] - explain .red[how individuals affect social structure] - explain .red[how social structure affects individuals] - understand .red[how micro and macro outcomes relate] - understand .red[how different networks relate] - understand .red[how different mechanisms change or sustain social systems] ] --- class: center, middle # Summary .pull-1[.circleon[![](https://globalquiz.org/media/pic/400/8340.jpg)]] .pull-1[.circleon[![](https://cdn.imgbin.com/14/23/6/imgbin-people-connected-tXALjkJraJY0PjsUKek18tXrH.jpg)]] .pull-1[.circleon[![](https://i.pinimg.com/originals/8a/08/f1/8a08f1d673c8b64ddb067de8d8d879d8.jpg)]] -- .pull-left[.pull-down[What questions do you have for me?]]