Exploring opinion leadership and homophily in political discussion networks of korean twitter users

Post on 10-May-2015

183 views 0 download

Tags:

transcript

Exploring Opinion Leadership and Homophily

in Political Discussion Networks of Korean Twitter Users

December 14, 2013

Yoonmo Sang (UT-Austin), Myunggoon Choi (Sungkyunkwan U), Hanwoo Park (Yeungnam U)

SNSs and Political Discourse

Platform for political engagement

The relationship between social media use and political engagement

Mixed result (see Kushin & Yamamoto, 2010)

• Beneficial effects on political engagement (e.g., Kim & Geidner, 2008;

Valenzuela, Park, & Kee, 2009)

• Questioning SNSs’ role in facilitating political engagement (e.g., Gil de Zuniga,

Puig, & Rojas, 2009; Zhang, Johonson, Seltzer, & Bichard, 2010)

Opinion Leadership in Political Communication

Two-step flow theory (Katz & Lazarsfeld, 1955/2006)

Diffusion studies (Rogers, 2003; Vishwanath & Barnett, 2011)

Sociometric approach to opinion leadership (Monge & Contractor, 2003; Rogers,

2003; Valente & Pumpuang, 2007; Valente, 2010)

Opinion leadership in Twitter networks (Wu, Hofman, Mason, & Watts, 2011)

• Bimber (2004)

• Mutz & Mondak (2006)

• Brundidge (2010)

• Garrett et. al. (2011)

• McPherson, Smith-Lovin, & Cook

(2001)

• Adamic & Glance (2005)

• Sunstein (2007)

• Gilbert & Karahalios (2009)

Divergent Findings on homophily and selective exposure

Uniqueness of Twitter necessitates further research on this issue.

Conover et al. (2011); Gruzd (2012); Himelboim, McCreery, & Smith (2013);

Murthy (2012); Yardi & boyd (2010)

Research questions

RQ1: Who are the opinion leaders on Twitter in the discussion on President

Myung-Bak Lee? Specifically, are opinion leaders on Twitter polymorphic or

monomorphic through political events?

RQ2: Is the Twitter-based network broken down into subgroups with similar

political interests?

Method

Data collection

Data collection period: from Nov. 1, 2011 to Apr. 20, 2012

NodeXL (the open-source network analytic tool which can collect and visualize

network data)

UCINET6, Gephi for Data Analysis & Visualization

53, 165 Twitter users/ 1,144,306 lines (Three types: following, retweet, mention)

Moderate correlation between mutual and interaction networks (Gruzd et al., 2011)

Measurement

Opinion leadership : In-degree centrality

* In-degree Centrality: The number of ties received from other actors

Network density within and between clusters (calculated by UCINET6)

* Density in the study: The proportion of ties formed through “following,”

“mention,” and “RT(retweet)” by Twitter users who sent tweets about

Myung-Bak Lee

Measurement (Cont.)

Systematic steps to conduct the study

No Step Description1 Data collection - NodeXL using Twitter Search API 1.0

- All available tweets (Singleton, Retweet, and Mention) Korean president’s full name, “Myung-Bak Lee,”

- From November 1, 2011 to April 20, 2012.

Measurement (Cont.)

Systematic steps to conduct the study

No Step Description2 Identifying opinion

leaders

- In-degree centrality

- Examining political views of opinion leaders by their twitter profiles and the People Search service offered by Naver

Measurement (Cont.)

Systematic steps to conduct the study

No Step Description3 Dissecting the

networks

- Modularity algorithm via Gephi

- Limiting to those days that show above average network density

Measurement (Cont.)

Systematic steps to conduct the study

No Step Description4 Determining the

political inclination of each cluster

- Systematically selecting 10% of Twitter users with the highest in-degree centrality within each sub-cluster

- Coding their political views by two independent coders and using Krippendorff’s α (Krippendorff, 2004) to assess inter-coder reliability

Measurement (Cont.)

Systematic steps to conduct the study

No Step Description5 Correlation of

words’ frequencies between core and peripheral groups

- Semantic analysis

- Comparing word frequency with regard to the tweets generated by 10% core nodes with the highest in-degree centrality, and by 90% the peripheral nodes

- Geul-Jap-I

Measurement (Cont.)

Systematic steps to conduct the study

No Step Description6 Interactions between

the liberal and conservative cluster

- Densities between and within clusters (calculated by UCINET)

Results

Opinion Leaders in the Context of Twitter

The demographic characteristics (including occupation and political view) of

influential Twitter users from their Twitter profiles and through search engines

Those who showed the highest indegree centrality in the discussion network at

least two times

Liberal Twitter users’ considerable influence on the Twitter network (Hsu & Park,

2011; Hsu & Park, 2012)

No. Twitter IDNo. of Twitterians with the highest indegree centrality

Followings Followers Tweets OccupationPolitical

view

1 User1 6 (12.24%) 17,391 17,596 17,484 Journalist Liberal

2 User2 6 (12.24%) 6,005 89,696 16,269 Researcher Liberal

3 User3 4 (8.16%) 381 267,531 15,012 Novelist Liberal

4 User4 4 (8.16%) 45,256 48,586 9,859 Media outlet Liberal

5 User5 4 (8.16%) 5,048 54,821 27,188 Journalist Liberal

6 User6 2 (4.08%) 220,879 202,123 51,645 Unknown Not clear

7 User7 2 (4.08%) 68,725 139,033 65,106 Journalist Liberal

8 User8 2 (4.08%) 24,830 28,476 11,370 Politician Liberal

9 User9 2 (4.08%) 25,156 43,290 15,776 Media outlet Liberal

Table 1. Opinion leaders in the discussion on Myung-Bak Lee

Mapping Interactions Within and Between Subgroups

Date(month/day/year)

Liberal clusters Conservative clusters

Core 10% nodes-peripheral 90% nodes

Core 10% nodes-peripheral 90% nodes

11/26/11 0.874*** 0.824***11/30/11 0.927*** 0.773***12/14/11 0.881*** 0.848***12/18/11 0.909*** 0.679***12/19/11 0.499*** 0.844***12/24/11 0.871*** 0.522***12/25/11 0.912*** 0.866***12/26/11 0.866*** 0.675***01/01/12 0.899*** 0.903***01/14/12 0.857*** 0.836***

Table 2. Pearson correlation coefficient

Mapping Interactions Within and Between Subgroups (Cont.)

Date(month/day/year)

Liberal clusters Conservative clusters

Core 10% nodes-peripheral 90% nodes

Core 10% nodes-peripheral 90% nodes

01/15/12 0.786*** 0.747***01/21/12 0.624*** 0.786***01/22/12 0.823*** 0.502***01/27/12 0.851*** 0.842***01/29/12 0.879*** 0.872***02/09/12 0.909*** 0.759***03/10/12 0.581*** 0.701***03/18/12 0.882*** 0.827***03/24/12 0.913*** 0.840***03/30/12 0.930*** 0.934***

Table 2. Pearson correlation coefficient

Date

(Mon/

Day

/Year)

ModularityNo. of

Clusters

Cluster1

(%)

Cluster2

(%)

Cluster3

(%)

Cluster4

(%)

Cluster5

(%)

Cluster6

(%)

Cluster7

(%)

Liberal

(%)

Conservative

(%)

11/26/11 0.172 6 36.67 18.62 17.58 10.68 8.98 1.04 - 82.89 10.68

11/30/11 0.223 4 32.28 25.48 23.75 13.69 - - - 69.72 25.48

12/14/11 0.222 4 33.85 29.95 26.3 3.12 - - - 56.25 33.85

12/18/11 0.179 6 26.91 18.6 14 12.47 11.82 10.94 - 67.83 26.91

12/19/11 0.209 4 38.03 22.14 17.6 17.41 - - - 73.04 22.14

12/24/11 0.156 6 35.62 21.14 12.34 11.48 11.37 3.54 - 91.95 3.54

12/25/11 0.184 4 43.77 21.22 19.98 11.8 - - - 84.97 11.8

12/26/11 0.216 5 38.55 19.74 18.91 16.96 3.15 - - 75.25 18.91

01/01/12 0.153 4 31.7 24.56 22.71 13.99 - - - 78.97 13.99

01/14/12 0.170 5 32.97 26.49 26 5.98 5.08 - - 90.54 5.98

01/15/12 0.145 6 32.6 25.64 13.59 8.73 7.07 5.64 - 84.54 8.73

01/21/12 0.170 6 20.08 16.94 16.45 15.87 14.79 10.87 - 78.55 16.45

01/22/12 0.200 5 22.43 21.11 20.36 19.49 11.4 - - 73.68 21.11

01/27/12 0.222 4 35.1 23.62 18.94 18.51 - - - 72.55 23.62

01/29/12 0.254 6 28.59 21.17 15.88 12.61 11.55 1.44 - 70.07 21.17

02/09/12 0.264 4 34.96 30.71 22.04 5.03 - - - 52.75 34.96

03/10/12 0.169 6 33.66 18.14 16.88 14.08 10.56 2.35 - 82.76 10.56

03/18/12 0.139 7 23.73 19.96 18.57 11.92 9.93 7.35 4.27 83.81 11.92

03/24/12 0.189 5 48.64 19.1 18.34 6.49 5.55 - - 91.63 6.49

03/30/12 0.229 4 40.22 31.84 24.85 1.8 - - - 65.07 31.84

Table 3. Result of Modularity Analysis

Date

(Mon/

Day

/Year)

ModularityNo. of

Clusters

Cluster1

(%)

Cluster2

(%)

Cluster3

(%)

Cluster4

(%)

Cluster5

(%)

Cluster6

(%)

Cluster7

(%)

Liberal

(%)

Conservative

(%)

11/26/11 0.172 6 36.67 18.62 17.58 10.68 8.98 1.04 - 82.89 10.68

11/30/11 0.223 4 32.28 25.48 23.75 13.69 - - - 69.72 25.48

12/14/11 0.222 4 33.85 29.95 26.3 3.12 - - - 56.25 33.85

12/18/11 0.179 6 26.91 18.6 14 12.47 11.82 10.94 - 67.83 26.91

12/19/11 0.209 4 38.03 22.14 17.6 17.41 - - - 73.04 22.14

12/24/11 0.156 6 35.62 21.14 12.34 11.48 11.37 3.54 - 91.95 3.54

12/25/11 0.184 4 43.77 21.22 19.98 11.8 - - - 84.97 11.8

12/26/11 0.216 5 38.55 19.74 18.91 16.96 3.15 - - 75.25 18.91

01/01/12 0.153 4 31.7 24.56 22.71 13.99 - - - 78.97 13.99

01/14/12 0.170 5 32.97 26.49 26 5.98 5.08 - - 90.54 5.98

01/15/12 0.145 6 32.6 25.64 13.59 8.73 7.07 5.64 - 84.54 8.73

01/21/12 0.170 6 20.08 16.94 16.45 15.87 14.79 10.87 - 78.55 16.45

01/22/12 0.200 5 22.43 21.11 20.36 19.49 11.4 - - 73.68 21.11

01/27/12 0.222 4 35.1 23.62 18.94 18.51 - - - 72.55 23.62

01/29/12 0.254 6 28.59 21.17 15.88 12.61 11.55 1.44 - 70.07 21.17

02/09/12 0.264 4 34.96 30.71 22.04 5.03 - - - 52.75 34.96

03/10/12 0.169 6 33.66 18.14 16.88 14.08 10.56 2.35 - 82.76 10.56

03/18/12 0.139 7 23.73 19.96 18.57 11.92 9.93 7.35 4.27 83.81 11.92

03/24/12 0.189 5 48.64 19.1 18.34 6.49 5.55 - - 91.63 6.49

03/30/12 0.229 4 40.22 31.84 24.85 1.8 - - - 65.07 31.84

During the period of analysis, liberal Twitter users

(M = 76.34, SD = 10.88) were clearly more active

in discussing Myung-Bak Lee than conservative

ones (M = 18.01, SD = 9.48).

Table 3. Result of Modularity Analysis

Figure 1. Temporal changes in network structure over time

Figure 2. Longitudinal density for liberal and conservative clusters and between clusters

Mapping Interactions Within and Between Subgroups (Cont.)

ANOVA result:

Significant difference among three groups (F(2, 57) = 24.8327, p < .001).

- Conservatives: The most densely interconnected (M = 0.0508, SD = 0.0329)

- Liberals: Less densely interconnected (M = 0.0360, SD = 0.0065)

- Conservatives and liberals: Low level of interconnectedness (M = 0.0080, SD = 0.0026)

Limitations and future research

Conclusion

Discussion and conclusion

Thank you for listening…

Myunggoon ChoiDepartment of Interaction Science

Sungkyunkwan University

E-mail: myunggoon.choi@gmail.com