Understanding and Enabling Online Social Networks to Support Healthy - - PowerPoint PPT Presentation

Understanding and Enabling Online Social Networks to Support Healthy Behaviors

Noshir Contractor Jane S. & William J. White Professor of Behavioral Sciences Jane S. & William J. White Professor of Behavioral Sciences Professor of Ind Engg & Mgmt Sciences McCormick School of Engineering Professor of Ind. Engg & Mgmt Sciences, McCormick School of Engineering Professor of Communication Studies, School of Communication & Professor of Management & Organizations, Kellogg School of Management, Director, Science of Networks in Communities (SONIC) Research Laboratory nosh@northwestern.edu

SONIC

Advancing the Science of Networks in Communities

Supported by NSF IIS-0729505, Army Research Institute (W91WAW-08-C-0106), and Sony Online Entertainment

Networks in Health

• Networks of researchers, especially multi‐,

inter‐, trans‐ disciplinary

• Networks to assemble teams of

practitioners practitioners N t k f ti t

• Networks of patients
• Networks of “Science of Science” policy

makers

Patientslikeme.com Patientslikeme.com

Healthy Lifestyle Network Healthy Lifestyle Network

MTML Social Drivers: h d l h Why do People Connect to Others?

• Theories of self‐interest • Theories of contagion

Theories of self interest

• Theories of social and

resource exchange

• Theories of contagion
• Theories of balance
• Theories of homophily

resource exchange

• Theories of mutual

interest and collective

• Theories of homophily
• Theories of proximity

action

Sources: Sources: Contractor, N. S., Wasserman, S. & Faust, K. (2006). Testing multi-theoretical multilevel hypotheses about organizational networks: An analytic framework and empirical example. Academy of Management Review. Monge, P. R. & Contractor, N. S. (2003). Theories of Communication Networks. New York: Oxford University Press.

“Structural Signatures” of MTML Structural Signatures of MTML

A

A

B F C E

• +

B F C E

• +

D

D

Theories of Self interest Theories of Exchange Theories of Balance

B F A

+

• +

F E B C A

• +

C E D

G o v e rn m en t In d u stry

D

Novice Expert

Theories of Collective Action Theories of Homophily Theories of Cognition Theories of Collective Action Theories of Homophily Theories of Cognition

Statistical “MRI” for Structural Signatures Statistical MRI for Structural Signatures

• p*/ERGM: Exponential Random Graph Models

p /ERGM: Exponential Random Graph Models

• Statistical “Macro‐scope” to detect structural

motifs in observed networks motifs in observed networks

• Move from exploratory to confirmatory

k l i d d l i network analysis to understand multi‐ theoretical multilevel motivations for why we i l k create our social networks

Multidimensional Networks Multiple types of Nodes and Multiple Types of Relationships

Its all about “Relational Metadata”

• Technologies that “capture” communities’ relational meta‐data

(Pingback and trackback in interblog networks blogrolls data (Pingback and trackback in interblog networks, blogrolls, data provenance)

• Technologies to “tag” communities’ relational metadata (from Dublin

C i f lk i (‘ i d f d ’) lik Core taxonomies to folksonomies (‘wisdom of crowds’) like – Tagging pictures (Flickr) – Social bookmarking (del.icio.us, LookupThis, BlinkList) Social citations (CiteULike org) – Social citations (CiteULike.org) – Social libraries (discogs.com, LibraryThing.com) – Social shopping (SwagRoll, Kaboodle, thethingsiwant.com) – Social networks (FOAF, SIOC, SocialGraph) Social networks (FOAF, SIOC, SocialGraph)

• Technologies to “manifest” communities’ relational metadata

(Tagclouds, Recommender systems, Rating/Reputation systems, ISI’s Hi tCit N t k Vi li ti t ) HistCite, Network Visualization systems)

SONIC

Advancing the Science of Networks in Communities

The Hubble telescope: $2.5 billion

SONIC

Advancing the Science of Networks in Communities

Source: David Lazer

CERN particle accelerator: $ b ll / $1 billion/year

SONIC

Advancing the Science of Networks in Communities

Source: David Lazer

The Web: priceless*

* Apologies to MasterCard

Source: David Lazer

SONIC

Advancing the Science of Networks in Communities

SONIC

Advancing the Science of Networks in Communities

Using Digital Traces to Test MTML Using Digital Traces to Test MTML

• Massively‐multiplayer online games (MMOGs) have

Massively multiplayer online games (MMOGs) have

• ver 45 million users worldwide and over $3 billion

in revenue in 2008

• What does social behavior in online worlds tell us

about the “real” world and vice versa?

– Online games exhibit features that map onto real world processes:

• Social networks economics groups communication conflict
• Social networks, economics, groups, communication, conflict,

expertise, leadership, crime, innovation, epidemics, etc.

– Online games already capture the signatures of these b h i i h d b j i i b l d behaviors in huge databases, just waiting to be analyzed.

Rise of WoW

Source: http://www.mmogchart.com/

SONIC

Advancing the Science of Networks in Communities

Expertise/Information Retrieval Time One

SONIC

Advancing the Science of Networks in Communities

Expertise/Information Retrieval Time Two

Expertise/Information Retrieval Time Three

Unraveling the “St t l Si t ” “Structural Signatures”

n Incentive for creating a WoW link with g someone = ‐1 55 (cost of creating a link) [Self‐interest] = 1.55 (cost of creating a link) [Self interest] + 0.55 (benefit of reciprocating) [Exchange] ( f f f f f ) + 0.89 (benefit for being a friend of a friend) [Balance]

+ 0.04 (benefit of connecting to an expert)

[Cognition] [Cognition]

All coefficients significant at 0.05 level

SONIC

Advancing the Science of Networks in Communities

Proximity & Homophily Proximity & Homophily

• Long tradition of examining how physical proximity

affects communication (Bossard, 1932; Stewart, 1942; Festinger, Schachter, & Back, 1950; Gullahorn, 1952)

• Technology‐mediated proximity research began with the

Technology mediated proximity research began with the availability of the telephone (Mayer, 1977)

• With the computer, researchers began studying how

ffli di t ff t li i t ti (C i

• ffline distance affects online interaction (Cummings,

Lee, & Kraut, 2006; Eveland & Bikson, 1986; Hampton & Wellman, 2001; Kraut, Egido, & Galegher, 1988)

• New research suggests that online interactions organized

more by the cost to communicate than by physical geography (Falk & Abler, 1980). g g p y ( )

Relevant MTML Theories Relevant MTML Theories

• Theory of Proximity

– People tend to form social relations with those who are geographically close to them.

Th f H hil

• Theory of Homophily

– People tend to form social relations with those who share some prominent characteristics some prominent characteristics

• Theory of Self‐interest

– People are selective about making links online. People are selective about making links online.

• Multi‐Theoretical Multilevel (MTML) Models

– Contractor et al, 2006; Monge & Contractor, 2003 , ; g ,

Four Types of Relations in EQ2 Four Types of Relations in EQ2

• Partnership: two players play together in combat activities;

p p y p y g ;

• Instant messaging: two players exchange messages through

Sony universal chat system

• Player trade: players meet “face‐to‐face” in EQ2 and one gives

items to another;

• In‐game mail: one player sends a message and/or items to

In game mail: one player sends a message and/or items to

• thers by in‐game mail

Synchronous Asynchronous Synchronous Asynchronous Interpersonal interaction Partnership, Instant messaging Transactional interaction Player trade In‐game mail Transactional interaction Player trade In game mail

Relation 1: Partnership Relation 1: Partnership

Relation 2: Instant Messaging Relation 2: Instant Messaging

Relation 3: Player Trade Relation 3: Player Trade

Relation 4: In‐game Mail Relation 4: In game Mail

Black: male Red: female

Partnership Instant messaging Trade In-game mail

Hypotheses ‐ Proximity

• H1: (Spatial proximity) Individuals who are proximate

Hypotheses Proximity

H1: (Spatial proximity) Individuals who are proximate in geographical distance are more likely to engage in

• nline interaction than those who are not proximate.
• H2: (Interaction types) Individuals who are proximate

are more likely to engage in online interpersonal interactions (i.e. partnership and instant messaging) than transactional interactions (i.e. trade or mail).

Hypotheses ‐ Homophily Hypotheses Homophily

• H3: (Gender) Individuals of the same gender are

H3: (Gender) Individuals of the same gender are more likely to engage in interaction than those of

• pposite genders.
• H4: (Age) Individuals who have smaller age

differences are more likely to engage in interaction than those who have bigger differences.

• H5: (Experience) Players who have similar years of

l k l game experience are more likely to engage in interaction than those who have bigger differences.

Hypotheses ‐ Network Hypotheses Network

• H6: (Selectivity) Individuals are not likely to engage in

H6: (Selectivity) Individuals are not likely to engage in interaction randomly in a virtual world.

• H7: (Popularity) Individuals with many interactions are

( p y) y more likely to engage in interaction than those have a few interactions.

• H8: (Transitivity) Two individuals who both interact

with the third parties are more likely to engage in h h d h interaction than those do not have common parties between them.

Data Description Data Description

• 3140 players from Aug 25 to Aug 31 2006 in

3140 players from Aug 25 to Aug 31 2006, in Antonia Bayle

– 2998 US 142 CA ; 2447 male 693 female – 2998 US, 142 CA ; 2447 male, 693 female

• Demographic information

– Gender, age, and account age (years played Sony ) games) – Zip code, state, and country

Descriptive Statistics Descriptive Statistics

• 3140 players from Aug 25 to Aug 31 2006, in Antonia Bayle

p y g g , y

– 2998 US, 142 CA ; 2447 male, 693 female

Min. 1st Qu. Median Mean 3rd Qu. Max.

Age 5 668 25 410 31 350 32 340 37 550 71 040 Age 5.668 25.410 31.350 32.340 37.550 71.040 AcctAge 0.003 1.014 1.803 2.451 3.275 9.400 Latitude 18.27 33.88 38.89 38.28 42.11 71.29 Longitude 158 00 112 00 88 38 94 19 80 58 50 71 Longitude ‐158.00 ‐112.00 ‐88.38 ‐94.19 ‐80.58 50.71

Network Nodes Edges Density Degree (mean) Degree (max) Diameter Centralization (degree)

Partner 1924 1789 0.097% 1.860 14 24 0.63% Instant messaging 548 517 0.34% 1.887 10 27 1.49% d Trade 2456 3812 0.13% 3.104 24 19 0.85% Mail 2090 3120 0.14% 2.986 84 19 3.83%

Construct Physical Proximity Construct Physical Proximity

• Zip code ‐> latitude/longitude time zone

Zip code > latitude/longitude, time zone

– U.S. ZIPList5 and Canada Geocode from ZipInfo com ZipInfo.com

• Latitude/longitude ‐> geographical distance

Spherical law of cosines: – Spherical law of cosines:

acos(sin(lat1)×sin(lat2)+cos(lat1)×cos(lat2)×cos(long2-long1))×6371 Km

Network Models Network Models

• Model 0 – Network effect

Model 0 Network effect

• Model 1 – Homophily

G d t hi (1 f M M F F 0 f M F – Gender matching (1 for M‐M or F‐F; 0 for M‐F match) Age difference (absolute value in years) – Age difference (absolute value in years) – Account age difference (absolute value in years)

d l l

• Model 2 – Spatial proximity

– Physical distance (absolute value in kilometers)

p*/Exponential Random Graph Models p /Exponential Random Graph Models

• Analysis network data with Interdependencies

Analysis network data with Interdependencies – endogenous correlation among the relations

• ERGMs are a class of stochastic models:
• ERGMs are a class of stochastic models:

( ) i f k i i h

1 ( ) exp( ( )) ( )

T

P Y y g y k θ θ = =

– g(y) is a vector of network statistics such as network structural measures and node attributes θ i t f ffi i t – θ is a vector of coefficients.

– Frank & Strauss, 1986; Pattison & Wasserman, 1999; Robins, Pattison, & Wasserman, 1999; Wasserman & Pattison, 1996; Hunter, 2007; Robins, Snijders, Wang, Handcock, & Pattison, 2007

Estimation Methods Estimation Methods

• p*/ERGM model variables

p /ERGM model variables

– Non‐directed network: dichotomized partner, IM, trade, and mail relations trade, and mail relations – Network structures: edges, gwdegree, and gwesp – Dyadic attributes: geographical distance and time y g g p zone difference between players – Actor attributes: gender, age, and account age

• Estimation tools

– Statnet version 2.1, R‐2.8.0 , – Social Sciences Computing Cluster (SSCC) at NU

Results 0: Network Model Results 0: Network Model

Supported: Supported: Individuals are not likely to engage in interaction randomly. Partially supported: High degree Individuals are more likely to engage in partner and IM relations engage in partner and IM relations but not trade and mail. d f Supported: If two individuals have shared partners, they are more likely to engage in partner d l i and IM relations.

Results 1: Homophily Model Results 1: Homophily Model

Not supported: Individuals of the same gender are NOT Supported: Individuals with likely to engage in interaction. Supported: Individuals with similar age and experience are more likely to engage in interaction.

Results 2: Proximity Model Results 2: Proximity Model

Supported: Distance reduces the likelihood of interaction

Hypotheses Tested Hypotheses Tested

More Detailed Study More Detailed Study

Hypotheses Partnership IM Trade Mail Hypotheses Partnership IM Trade Mail

H1 Sparsity Yes Yes Yes Yes H2 Popularity Yes Yes No No H3 Transitivity Yes Yes N/A N/A H4 Geographic proximity Yes Yes Yes Yes H4a Short distance Yes Yes Yes Yes H4b Interaction types High Low Medium Medium H5 Temporal proximity Yes Yes Yes Yes H6 Synchronization High Low Medium Medium H6 Synchronization High Low Medium Medium H7 Gender homophily No No No No H8 Age homophily Yes Yes Yes Yes H9 Experience homophily Yes Yes Yes Yes

Results and Discussion Results and Discussion

• Selectivity and transitivity exists in all online relations.

Selectivity and transitivity exists in all online relations.

• Homophily of age and game experience is supported

in all fours relations.

• Distance matters but has different impacts on

different relations.

• Gender homophily is not supported for all relations

and players are more likely to interact with the

• pposite gender.

Find New Friends Nearby Find New Friends Nearby

F ll l th b l t f l ’

Km)

For all players, the box plot of players’ distances to their partners shows that players tend to end up with partners closer to them when they have more experience in the game.

n thousand

y p g

Distance (in D Account Age (in year)

Summary Summary

• Distance still plays an important role in

Distance still plays an important role in people’s virtual relations

• Theories of proximity and homophily are valid
• Theories of proximity and homophily are valid

in a virtual world.

Design Examples: M i & E bli N t k i

b h b

Mapping & Enabling Networks in …

Tobacco Research: TobIG Demo i l h l Computational Nanotechnology: nanoHUB C b i f C S Cyberinfrastructure: CI‐Scope Demo O f ili O IKNOW Oncofertility: Onco‐IKNOW T l i l S i C ll CTSA Translational Science: Cornell CTSA

SONIC

Advancing the Science of Networks in Communities

Acknowledgements Acknowledgements