Network Methods for Behavior Change

Thomas W. Valente, PhD

Professor

Department of Population & Public Health Sciences

Keck School of Medicine

University of Southern California

tvalente@usc.edu

​

Outline

1. Network Influences on Behavior
2. Interventions & Program Implementation
3. Selecting a Network Intervention
4. Network Diagnostics
5. Different Networks/Measurement
6. Conclusions

1) Networks Influences on Behavior

• Knowing how and why people change behavior is important for:
• understanding human behavior
• getting best practices implemented
• improving patient care
• understanding organizational performance
• etc.

Ego Network with 6 Alters

A

C

Ego

B

D

E

F

4

Exposure is Associated with Adoption

A

C

Ego

B

D

E

F

Rogers & Kincaid, (1981). Communication networks: . New York: The Free Press

Coleman, et al., (1966). Medical innovation: A diffusion study. New York: Bobbs-Merril.

5

Ego Network with 6 Alters

A

C

Ego

B

D

E

FF

6

A

B

C

F

C

F

H

G

J

I

K

L

Structural Equivalence is Associated with Influence

A

C

Ego

B

D

Burt, R. (1987) Social contagion and innovation: Cohesion versus structural equivalence.

American Journal of Sociology, 92, 1287-1335.

7

Indirect Exposures Matter

A

C

Ego

B

D

E

F

H

G

J

I

K

Valente, T., (1995) Network Models of the Diffusion of Innovations.

Cresskill NJ: Hampton Press.

L

8

Expanding & Contracting the Sphere of Influence

Individuals Have Varying Thresholds

A

C

Ego

B

D

A

C

Ego

B

D

Low Threshold Adopter

High Threshold Adopter

Valente, T.W. (1996). Social network thresholds in the diffusion of innovations.

Social Networks, 18, 69-89.

10

Graph of Time of Adoption by Network Threshold for One Korean Family Planning Community

Time

Threshold

100%

0%

1963

1973

11

Online vs Offline Network Influences

A

C

Ego

B

D

E

F

Huang, et al. (2013). Peer influences: The impact of online and offline friendship networks on

adolescent smoking and alcohol use. Journal of Adolescent Health.

12

Joint Participation/Identification

A

Ego

B

C

D

F

Events

1

2

3

4

5

Fujimoto, K., Unger, J. & Valente, T. W. (2012). A network method of measuring affiliation-based peer influence: Assessing of teammates’ smoking on adolescent smoking. Child Development, 83, 442-451.

6

7

8

9

10

Events

13

Alter Attributes May Affect Influence

A

Ego

D

E

F

B

C

C

Male

Female

14

Networks

• … of behavior change
• … for behavior change

​

2) Interventions & Program Implementation

If networks are so important, how can we use them to improve programs?

2015

Social Network Analysis for Program Implementation (SNAPI)

Exploration (Needs Assessment)

Network Ethnography

• Is there a network to work with?
• What is the network position of those defining the problem?
• Are there disconnected subgroups in the community?
• Are there isolates who need to be connected?

Who Provides Input for Problem Definition & Program Design?

Program

Community as Network

• Makes explicit that problem definition and priority settings will vary depending on who provides input.
• Community based organizations are always confident they can hear the voice of the community, but we are all blind to the parts of the network we can’t see.
• In this example, people somewhat central in the network are involved but still other segments are left out.

Social Network Analysis for Program Implementation (SNAPI)

Network Interventions

“Network interventions are purposeful efforts to use social networks or social network data to generate social influence, accelerate behavior change, improve performance, and/or achieve desirable outcomes among individuals, communities, organizations, or populations.”

Principle 1: Program Goals Matter

• In some cases we want to increase cohesion in others increase fragmentation
• Or increase/decrease centralization …
• E.g., slowing spread of STDs may require fragmenting a sexual contact network or accelerating adoption condoms.
• Network Interventions Are not Agnostic to Content.

​

Principle 2: Behavioral Theory

• The type of change desired will be guided by theory
• Understanding motivations for and barriers against behavior change is critical.
• A well-articulated theory of the behavior is often critical for successful interventions.

Principle 3: Learn As Well As Induce

• The interventionist should use network methodology to learn from the community as much as try to influence it.
• Programs which meet the needs of their audiences are better received than those designed asymmetrically (i.e., without community input).

A Taxonomy of Network Interventions

Strategy

Tactic

Tactic

Tactic

Operational-

ization

Operational-

ization

Operational-

ization

Operational-

ization

Operational-

ization

Operational-

ization

Operational-

ization

Operational-

ization

Operational-

ization

1. Opinion Leaders

• The most typical network intervention
• Easy to measure
• Intuitively appealing
• Proven effectiveness
• Over 20 studies using network data to identify OLs and hundreds of others using other OL identification techniques

10 Methods Used to Identify Peer Opinion Leaders

Diffusion Network Simulation w/ 3 Initial Adopter Conditions (Valente & Davis, 1999)

Cochrane Review of OL Studies (Flodgren, et al., 2011)

• 18 trials
• 5 trials OL vs. No Intervention, +0.09;
• 2 trials OL vs. 1 Interventions, +0.14;
• 4 trials OL vs. 2+ Interventions, +0.10; and
• 10 trials OL+ vs. + Interventions, +0.10.
• Overall, the median adjusted risk difference (RD) was +0.12 representing 12% absolute increase in compliance.

In-degree Centrality Used

• Easy to calculate
• Easy to explain
• Do not symmetrize data
• Can compare in-degree scores with other centrality measures
• Compare degree scores with Key Player analysis

Other Centrality Measures

• In-degree preferable:
• Robust to missing data
• Supported by prior work
• Supported by simulation
• Can expand/change number or proportion of leaders selected (e.g., if not enough OLs participate then you can expand the list of those invited)

Other Centrality Measures

• Theoretical diffusion processes may suggest other centrality measures, closeness, betweenness, etc.
• Can use other measures as tie-breakers (i.e., 2 nodes of same in-degree choose one with higher closeness)

FREEMAN'S DEGREE CENTRALITY MEASURES

----------------------------------------------------------------------

Diagonal valid? NO

Model: ASYMMETRIC

Input dataset: C:\MISC\DIFFNET\OL\com18

1 2 3 4

OutDegree InDegree NrmOutDeg NrmInDeg

------------ ------------ ------------ ------------

19 26 5.000 2.000 13.889 5.556

20 27 5.000 7.000 13.889 19.444

3 11 5.000 5.000 13.889 13.889

4 12 5.000 6.000 13.889 16.667

5 13 5.000 6.000 13.889 16.667

6 14 5.000 7.000 13.889 19.444

25 31 5.000 7.000 13.889 19.444

8 16 5.000 6.000 13.889 16.667

9 17 5.000 8.000 13.889 22.222

10 18 5.000 1.000 13.889 2.778

11 19 5.000 3.000 13.889 8.333

12 2 5.000 2.000 13.889 5.556

13 20 5.000 1.000 13.889 2.778

14 21 5.000 11.000 13.889 30.556

15 22 5.000 4.000 13.889 11.111

34 6 5.000 5.000 13.889 13.889

17 24 5.000 6.000 13.889 16.667

. . .

Additional Thoughts about Leaders

1. Many ways to define and identify
2. Most frequently used network intervention
3. Many studies of leaders and diffusion have been conducted so that relation is somewhat understood
4. In-degree is robust to missing data
5. Key-player depends greatly on the number of leaders selected

Implementation Issues

• Do you just turn leaders loose? How to structure interactions:
• Schedule 1-1 between leaders & members
• Have leaders give formal presentations
• Have leaders call a meeting
• Allow leaders to decide how to promote change

Passive Active OL

Involvement Involvement

1.C. Identify Bridging Nodes

• Leaders are important within groups
• Bridges may be critical to diffusion between groups
• Bridges may be less “burdened” than leaders so easier to work with.
• Bridges may be more amenable to change.

​

41

3 Ways to Identify Bridges

• Sociometric bridges: Nodes whose links, on average, are most critical in the network.
• Valente, T. W., & Fujimoto, K. (2010). Bridging: locating critical connectors in a network. Social Networks, 32(3), 212-220.
• Using betweenness:
• Everett, M. G., & Valente, T. W. (2016). Bridging, brokerage and betweenness. Social Networks, 44, 202-208.
• Burt’s Structural Holes (inverse Constraint)

​

​

1.D. Identify Isolates or Peripherals

• Individuals on the periphery of the network or isolates receive information late or not at all.
• In some settings, isolates may be at increased risk, e.g., mental health concerns.

​

​

44

1.E. Identify Low Threshold adopters

• Individuals with low thresholds innovate early.
• They require fewer initial adopters.
• They may be experienced at persuading others.
• But will depend on some prior, related behavior

​

45

Graph of Time of Adoption by Network Threshold for One Korean Family Planning Community

Time

Threshold

100%

0%

1963

1973

Steps to Individual Interventions

1. Collect/obtain network data
2. Identify opinion leaders (or others)
3. Recruit them as champions
4. Convert them (if need be)
5. Assist them in their behavior change promotions

47

2. Segmentation

• Intervention is directed toward or includes a whole group of people
• Segmentation interventions identify and expect a whole group to adopt the innovation at the same time

2.A. Groups

• Sets of people/nodes that are densely connected
• Groups can reinforce (or inhibit) the behavior change process
• Behavior change may be appropriate for groups
• Finding groups

Defining Groups

• Components
• Cliques/Kplexes/Cycles, etc.
• Girvan-Newman algorithm
• Provides mutually exclusive groups
• Provides measure of group fit
• Other community detection algorithms

Many Networks are Modular

Implementation Issues

• Do groups need to be the same size?
• In school-based programs, usually they do
• In organizations they can vary somewhat but then group size becomes an issue
• Does the socio-demographic composition of the group matter?
• Most cases groups will be homogenous
• Some cases may need to impose homogeneity on the group (e.g., sex education in primary schools)

​

2.B. Positions Rather Than Groups

• Positions may be more relevant than groups
• Hierarchical position may be relevant (e.g., supervisors versus line staff)
• Positions may also be defined on in-degree

Defining Positions

• Formal charts or hierarchies
• Indegree may distinguish positions
• Blockmodel on attributes (image matrix/diagram)
• Can also use positional analysis such as CONCOR or automorphic equivalence to identify positions

​

Implementation Issues

• Positions may vary on the extent they are explicit or implicit
• Positions may vary considerably in size
• Positions may be hard to organize

3. Induction

• Strategies 1 and 2 use the network to identify individuals/groups
• Strategy 3 tactics use the network structure
• These tactics activate or use the dyads in the network

​

​

57

3.A. Word of Mouth

• Many interventions hope/expect to excite people to talk about their products and “spread the word.”
• Have the behavior “go viral.”
• Many online retailers build this in allowing you to post purchase information to Facebook, Twitter, other social media.

3.B. Snowball Sampling: �Respondent Driven Sampling

• Epidemiologists have employed contact tracing for years – often data are not published or publicly available
• Several studies have used snowball methods to recruit a sample
• Several studies have used snowball methods to recruit an intervention group
• http://www.respondentdrivensampling.org/

Snowball/RDS -Implementation Issues

• Ties are homophilous
• Need coupons as incentives for recruiters
• $20 works fine for most applications but it is probably going up
• Need to ID coupons
• Challenge to keep ID numbers straight
• Might be able to use automated debit cards

Do groups/positions need leaders?

• Can let groups determine their leaders or leadership structure when:
• Behavior change issue is controversial
• Behavior change process is controversial
• Imposing a leadership structure may be preferred when:
• Behavior change process is accepted
• Goals are well-defined

3.D. Match Leaders to Groups

• Rather than have leaders unattached, assign them to people who think they are leaders
• Leadership is local
• Emphasizes homophily between leaders and members
• Builds on naturally occurring networks
• Leaders can be more effective if assigned to those who nominate them

​

Opinion Leaders

Individuals Receive the Most Nominations

Networked Condition

Optimal leader/learner matching

Tobacco Use Prevention Among Adolescents in Culturally Diverse California�

Intervention Overview

• Test of a Culturally Tailored Tobacco Prevention Curriculum
• Two curricula created and implemented in 16 middle schools
• Compared against 8 control schools
• CHIPS – standard social influences program
• FLAVOR – culturally tailored
• Data collected in 6^th, 7^th and 8^th grades

​

Comparison of 3 Conditions

Study Design

Regression Results on Post Program Appeal �(Lower Scores Better)

(N=1961; k=84; Beta Coefficients)

Regression Results on Post Program Attitudes (Lower Scores Better, Beta Coefficients)

71

Classroom Level Analysis

(N=k=84; Beta Coefficients)

1-Year Change in Smoking by Curricula &

Implementation Condition

Results Summary

Network condition

• was most appealing
• reduced pro-tobacco attitudes
• reduced susceptibility

​

Network effect was dependent on curriculum

TND Network

• How would a network condition compare to an existing evidence-based program?
• TND is a tobacco and drug use prevention curriculum tested in multiple setting.
• Created TND Network designed to be TND plus interactivity and network method for leader and group definitions.

Conclusions

• Network methods were effective at changing short term outcomes
• First turn-key network-based interventions
• Network implementation methods are sensitive to curriculum.
• Who you receive the intervention with (the social context) affects its impact.

Implementation Issues

• Have assignments and information readily available, we had 1 week or less to collect network data and return the leaders and groups
• Concerns about confidentiality

​

4. Alteration (Manipulation)

• Delete/add nodes
• Delete/add links
• Rewire entire network

4.A. Delete/Add Nodes

• Calculate a network metric
• Cohesion, for example
• Delete a node
• Re-calculate the metric
• Calculate difference
• Nodes with largest difference are most important
• Referred to as Vitality in the literature
• Induced centrality (Everett & Borgatti)

4.B. Delete/Add Links

• Calculate a network metric
• Cohesion, for example
• Delete a link
• Re-calculate the metric
• Calculate difference
• Links with largest difference are most important

Rewire Calculations

• Calculate original metric
• Change link
• Delete existing link
• Add non-existence link
• Calculate new metric
• Put difference in new matrix

Change Matrix Scores��Positive Numbers=Cohesion Increased When Added� Negative Numbers=Cohesion Decreased When Deleted

Implementation Issues

• Do we know which network structural metric to maximize? (I/Borgatti used the inverse of the APL)
• Is it a zero-sum (i.e., does one need to keep the # of links constant)?
• How to control for naturally occurring network dynamics (people enter/leave the network, change affiliations, etc.)?

​

Implementation Issues (2)

• Simply re-wiring links will probably not work
• There are reasons networks are the way they are (e.g., people make bonds with those they like).
• Network dynamics – people come and go and this will affect overall network properties

​

4.C.i Rewire Networks

• Can re-wire existing networks to have certain global network properties
• Make an empirical network a small world network (more clustering)
• Make an empirical network more centralized; etc.
• Re-wire to match adopters and non-adopters

4.C.ii Rewire Including Behaviors

• SNA community (rightfully) criticized other fields for ignoring relations.
• Now SNA community needs to consider the attributes of the nodes.
• Nodes are people too!

Disliking was reduced when seating assignments were re-arranged

van den Berg, Y. H., Segers, E., & Cillessen, A. H. (2012). Changing peer perceptions

and victimization through classroom arrangements: A field experiment.

Journal of abnormal child psychology, 40(3), 403-412.

​

A Taxonomy of Network Interventions

Graphical Displays of Intervention Choices

Graphical Displays of Intervention Choices

?

3) Selecting a Network Intervention

• Availability and type of data
• Types of networks
• Existing network structure
• Behavioral characteristics
• Existing prevalence
• Perceived characteristics such as cultural compatibility; cost; trialability; etc.

​

Influence Mechanisms Aligned with Interv. Choices

Social Network Analysis for Program Implementation (SNAPI)

4) Network Diagnostics

Network Diagnostics Tool

5) Network Type

• Different networks serve different functions
• Advice (expertise) versus Discussion (trust)
• Different networks may require different intervention strategies
• E. g., ID OLs in advice networks yet structure groups using discussion networks.

97

Leadership & Influence have 2 Dimensions

Discussion Trust

Advice Expertise

Expertise vs Trust

• When the barrier is technical: One doesn’t know how to do it, expertise is important
• When the barrier is cultural: One doesn’t know what to do, trust is important

Advantages of Network Methods (1)

• Capitalize on existing interpersonal relationships
• Use community input
• Establishes a learning organization /community
• Build social capital
• Are Empowering

Advantages of Network Methods (2)

• Can be replicated
• Fidelity can be measured
• Expands array of intervention options
• Creates data!

Comments?

• Comments, suggestions, criticisms welcome:

tvalente@usc.edu