We investigate the alignment of international attention of news media organizations within 193 countries with the expressed international interests of the public within those same countries from March 7, 2016 to April 14, 2017. We collect fourteen months of longitudinal data of online news from Unfiltered News and web search volume data from Google Trends and build a multiplex network of media attention and public attention in order to study its structural and dynamic properties. Structurally, the media attention and the public attention are both similar and different depending on the resolution of the analysis. For example, we find that 63.2% of the country-specific media and the public pay attention to different countries, but local attention flow patterns, which are measured by network motifs, are very similar. We also show that there are strong regional similarities with both media and public attention that is only disrupted by significantly major worldwide incidents (e.g., Brexit). Using Granger causality, we show that there are a substantial number of countries where media attention and public attention are dissimilar by topical interest. Our findings show that the media and public attention toward specific countries are often at odds, indicating that the public within these countries may be ignoring their country-specific news outlets and seeking other online sources to address their media needs and desires.
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
What We Read, What We Search: Media Attention and Public Attention Among 193 Countries
1. What We Read, What We Search:
Media Attention and Public Attention
Among 193 Countries
Haewoon Kwak Jisun An Joni Salminen
Soon-Gyo Jung Bernard J. Jansen
Qatar Computing Research Institute, Hamad Bin Khalifa University
2. What We Read, What We Search:
Media Attention and Public Attention
Among 193 Countries
2
3. What We Read, What We Search:
Structures and Dynamics of the
Multiplex Network of Media Attention and
Public Attention Among 193 Countries
Original title was…
3
4. What We Read, What We Search:
Structures and Dynamics of the
Multiplex Network of Media Attention and
Public Attention Among 193 Countries
Original title was…
4
5. What We Read, What We Search:
Media Attention and Public Attention
Among 193 Countries
5
6. What We Read, What We Search:
Media Attention and Public Attention
Among 193 Countries
1 Media attention?
2 Public attention?
6
7. Media attention
• What news media pay attention to (report)
The NYTimes pays attention to North Korea
NYTimes
North
Korea
1
US
North
Korea
US news media pay attention to North Korea
!7
8. Public attention
• What public pays attention to
• There are several ways to measure public attention
• Survey/Poll
• Social media
• (Google) search logs
2
!8
11. Critical role of news media
• Even in the era of social media, our understanding of the
world is still dominantly shaped by news media.
http://www.journalism.org/2016/07/07/pathways-to-news/
!11
12. Bias in news reporting
• Many kinds of bias in news reporting have been studied.
• Gatekeeping bias (also called selection bias):
• What we see, read, and hear about incidents is a result
of the gatekeeping by the journalists and news media.
• Foreign news coverage reflects the social, economic,
and political relationships among countries.
!12
14. Media outlets are not isolated
but interact with the audience
• These days journalists are trying to find a balance between
what people ought to know and what people want to
know
• News industry becomes more and more competitive
• People can easily access alternative online sources
• User engagement is visible
✓However, the spatiotemporal alignment of the news
coverage and public interests has been relatively unexplored
!14
15. Our goal
• We compare and contrast the media attention and public
attention among countries using longitudinal datasets.
• RQ1: How are media attention and public attention
structurally aligned?
• RQ2: What is the causal relationship between media
attention and public attention?
• RQ3: Are there topical aspects that affect the interaction
between media attention and public attention?
!15
16. Data collection - media attention
• Unfiltered News run by Jigsaw
http://unfiltered.news/ !16
17. Why Unfiltered.news?
• Top 100 topics per country retrieve from Google News
• Covers 100+ countries
• Translate contents in major languages
!17
18. Data collection - media attention
• Top 100 popular topics for each country for every day
from March 7, 2016 to April 14, 2017
• Co-mentions for each topic (context)
• 1,322,730 records of media attention among countries
• 152,557 unique co-mentions
!18
19. Data collection - public attention
• Google Trends
Search Term
Origin
Location
Time
!19
20. Data collection - public attention
• For each country, we collect search volumes about other
countries from March 7, 2016 to April 14, 2017
e.g. Time-series of search volume about US and KR in Korea
!20
21. Build the multiplex network
NM: Media attention network
(from Unfiltered News)
NP: Public attention network
(from Google Trends)
+
!21
22. Building daily NM and NP on t
Media
attention
Public
attention
193 countries Node 193 countries
News media in ci
cover cj on t
Link
from ci to cj
Public in ci search cj
on t
!22
23. Build the weighted network
NM: Superimposing daily NM over all t
NP: Superimposing daily NP over all t
!23
24. Country A Country B
Topic 1
Topic 2
Topic 3
Country B
Topic 5
…
Topic 100
7 March 2016
Country B
Topic 2
Topic 3
Topic 4
Topic 5
…
Topic 100
10 March 2016
Topic 1
Topic 2
Topic 3
Topic 4
Topic 5
…
Country B
12 April 2016
weight=3
!24
25. Backbone extraction
• Applying disparity filter proposed by [41] and extracting
significant links (called a “backbone”)
!25
27. RQ1: How are media
attention and public attention
structurally aligned?
!27
28. Afghanistan
Iran
Pakistan
United States of America
Albania
Republic of Macedonia
Algeria
Egypt
France
Iraq
Israel
Libya
Morocco
Russia
Saudi Arabia
SyriaTunisia
Turkey
Yemen
Angola
Brazil
China
Mozambique
Nigeria
Portugal
Anguilla
The Bahamas
Antigua and Barbuda
Barbados
Guyana
Jamaica
North Korea
Trinidad and Tobago
United Kingdom
Argentina
Bolivia
Chile
Colombia
Cuba
Ecuador
Mexico
Peru
Spain
Uruguay
Venezuela
Armenia
Azerbaijan
Georgia
Germany
Australia
Bosnia and Herzegovina
Fiji
New Zealand
Austria
Greece
Italy
Switzerland
Kazakhstan
Turkmenistan
Ukraine
Bahrain
Jordan
Kuwait
Oman
Qatar
United Arab Emirates
Bangladesh
Burma
India
Haiti
Saint Kitts and Nevis
Saint Lucia
Belarus
Belgium
Cameroon
Netherlands
Belize
Guatemala
Taiwan
Benin
Gabon
Gambia
Niger
Panama
Bermuda
Canada
Bhutan
Japan
Croatia
Montenegro
Serbia
Botswana
Namibia
South Africa
Zimbabwe
British Virgin Islands
Brunei
Philippines
South Korea
Bulgaria
Romania
Burkina Faso
Mali
Thailand
Cambodia
Indonesia
Malaysia
Vietnam
Central African Republic
Chad
Senegal
Cayman Islands
Togo
Hong Kong
Costa Rica
Nicaragua
Slovenia
Cyprus
Czech Republic
Slovakia
Dominica
Dominican Republic
Puerto Rico
Sudan
El Salvador
Honduras
Eritrea
Ethiopia
Estonia
Latvia
Kenya
Somalia
Papua New Guinea
Samoa
Tonga
Finland
Norway
Sweden
French Guiana
Martinique
French Polynesia
Sierra Leone
Ghana
Gibraltar
Greenland
Iceland
Guadeloupe
Guam
Guinea
Suriname
Macau
Hungary
Sri Lanka
Singapore
Ireland
Isle of Man
Kyrgyzstan
Tajikistan
Uzbekistan
Rwanda
Tanzania
Uganda
Kiribati
Lithuania
Lebanon
Liberia
Liechtenstein
Poland
Luxembourg
Madagascar
Mauritius
Réunion
Malawi
Zambia
Maldives
Malta
Mauritania
Moldova
Monaco
Montserrat
Nepal
New Caledonia
Paraguay
Burundi
Mayotte
San Marino
Seychelles
Solomon Islands
Swaziland
United States Virgin Islands
B(NM): Media attention
Afghanistan
India
Iran
Pakistan
Turkey
United States of America
Albania
France
Germany
Greece
Italy
United KingdomAlgeria
Egypt
Japan
Morocco
Saudi Arabia
Spain
Syria
Tunisia
Angola
Brazil
Portugal
South Africa
AnguillaBarbados
British Virgin Islands
Canada
Dominica
Dominican Republic
Guyana
Jamaica
Puerto Rico
Saint Kitts and Nevis
Saint Lucia
Trinidad and Tobago
Antigua and Barbuda
Argentina Chile
Colombia
Mexico
Paraguay
Uruguay
Armenia
Azerbaijan
Georgia
Russia
Ukraine
Australia
New Zealand
Austria
Croatia
Switzerland
China
Bahrain
Bangladesh
Kuwait
Philippines
Qatar
United Arab Emirates
Belarus
Lithuania
Poland
Belgium
Netherlands
Belize
Guatemala
Honduras
Benin
Ghana
Nigeria
Togo
Bermuda
Bhutan
Nepal Thailand
Bolivia
Peru
Bosnia and Herzegovina
Montenegro
Serbia
Slovenia
Botswana
Zambia
Zimbabwe
Guadeloupe
Hong Kong
Singapore
United States Virgin Islands
Brunei
Indonesia
Malaysia
Bulgaria
Burkina Faso
Gabon
Mali
Burma
Burundi
Kenya
Rwanda
Tanzania
Uganda
Cambodia
Vietnam
Cameroon
Cayman Islands
Cuba
Central African Republic
Senegal
Chad
Libya
Sudan
Macau
Venezuela
Costa Rica
Nicaragua
Panama
Ecuador
Cyprus
Czech Republic
Slovakia
Haiti
Israel
Lebanon
El Salvador
Eritrea
Ethiopia
Estonia
Finland
Latvia
Norway
Sweden
Somalia
Fiji
French Guiana
Martinique
Suriname
French Polynesia
New Caledonia
Gambia
Guinea
Gibraltar
Greenland
Iceland
Grenada
Guam
South Korea
Hungary
Iraq
Ireland
Isle of Man
Jordan
Oman
Kazakhstan
Uzbekistan
Kiribati
Papua New Guinea
Samoa
Taiwan
Kyrgyzstan
Liberia
Liechtenstein
Luxembourg
Madagascar
Malawi
Maldives
Sri Lanka
Malta
Mauritania
Mauritius
Mayotte
Moldova
Romania
Monaco
Mongolia
Montserrat
Mozambique
Namibia
Niger
Republic of Macedonia
Réunion
San Marino
Yemen
Seychelles
Sierra Leone
Solomon Islands
Swaziland
Tajikistan
The Bahamas
Tonga
Turkmenistan
Turks and Caicos Islands
B(NP): Public attention
!28
30. If one country gets high media
attention from other countries,
does it also get high public
attention?
!30
31. Centrality of a country in B(NM)
and B(NP) positively correlate
One country gets high media
(public) attention from other
countries, it also gets high public
(media) attention.
!31
32. Do media and public in one
country pay the most
attention to the same country?
!32
33. Top@k neighbors: Top k countries who
get the attention from one country
Country i Country j
Country k
Country l
!33
34. Top@k neighbors: Top k countries who
get the attention from one country
Country i Country j
Country k
Country l
Top@1 neighbor
!34
35. Top@k neighbors: Top k countries who
get the attention from one country
Country i Country j
Country k
Country l
Top@2 neighbors
!35
36. Overlap of top@k neighbors of
each country between NM and NP
!36
37. Overlap of top@k neighbors of
each country between NM and NP
In 63.2% of the countries, the
media and the public pay the
most attention (k=1) to different
countries.
Media’s heavy attention to the
United States is noticeable.
Temporally stable; but
worldwide convergence can
happen.
!37
38. The media and the public pay
attention to other countries in different
ways by focusing on different
countries with different strengths
This experiment is omitted in the presentation.
Please check our paper or slide 67-69
39. Attention flow among three
countries: Network motifs
• (Usually 3- or 4-) sized subgraphs that repeat in a given
network
!39
40. Proportions of each motif (compared to a
null model) are different according to the
types of networks
Superfamilies of Evolved and Designed Networks
Ron Milo, Shalev Itzkovitz, Nadav Kashtan, Reuven Levitt, Shai Shen-Orr, Inbal Ayzenshtat,Michal Sheffer, Uri Alon
Science, 303(5663), 2004
!40
42. Similar motif profiles in
B(NM) and B(NP)
• FFL
FFL / Double feedback loop
Transitive hierarchy exists in
media and public attention
e.g. [Algeria → France,
Algeria → the U.S.,
France → the U.S.]
!42
43. Similar motif profiles in
B(NM) and B(NP)
• FFL
Fully connected triad
Geographically closely
located countries
pay attention to each other
e.g. [China, Hong Kong,
Macau]
!43
44. Attention flow among
regions
• Recent studies on media attention report:
• Strong regionalism [21]
• ‘Global village’ trend [22]
• International news agencies play an essential role
!44
45. Community structures in B(NM)
and B(NP) by InfoMap [39]
•
‘Global village’ trend in media
attention:
80 countries in one community
Clearer geographical splits
!45
46. Community structures by
InfoMap [39]
The core of civilizations
proposed by Huntington [15] is
still observed.
!46
The trend of the global village
and other variations require new
models and explanations.
47. RQ2: What is the causal
relationship between media
attention and public attention?
!47
48. Granger causality
• One time series X Granger-causes the other time series Y
when past values of X can improve the explanation of the
current value of Y compared to when past values of Y are
used alone.
https://en.wikipedia.org/wiki/Granger_causality
!48
50. What influences the Granger-
causal relationships?
• Distance between two countries
• Out-degree in B(NM)
• GDP of the source country
• Internet penetration in the source country
• GDP of the destination country
!50
51. What influences the Granger-
causal relationships?
• Distance between two countries
• Out-degree in B(NM)
• GDP of the source country
• Internet penetration in the source country
• GDP of the destination country
Consistent with previous work
on newsworthiness of events
(news values)
!51
52. RQ3: Are there topical aspects that
affect the interaction between media
attention and public attention?
!52
54. Why one country gets
media attention
5 in G8 countries are covered with
politics, but France, Italy, and Japan are
covered with travel.
!54
55. Why one country gets
media attention
Negative stereotyping of the third world
by media attention in the early 2000s is
disappearing?
!55
56. How one country is covered by
countries in a specific region
• The United States, the United Kingdom, and Russia are
always covered with ‘politics’ by any region
• Countries get covered with ‘travel’ more by the same
region than other regions
• Several consistent connections between a country and a
topic across the regions, such as Brazil and sports, and
most of the Middle Eastern countries and politics.
!56
57. Top 2 topics that lead to the most
Granger-causal relationships
• Sports and travel
• High engagement of the public in the entertainment-
oriented news (often called soft news [37]) and supplies
of such news by the media
!57
58. Summary (1): Structural alignment of
media attention and public attention
• The importance of a country is positively correlated.
• To whom one country pays significant attention is different.
• The distribution of attention across neighbors is different;
public attention being distributed more equally than the
media attention.
• The relationship among three countries is similar.
• Media attention shows the trend of a global village, while
public attention shows clearer geographical splits.
!58
59. Summary (2): Interaction between
media attention and public attention
• Media attention and the public attention within the same
region strongly associate with each other.
• Along with distance, country properties influence the
Granger causality.
• There are some variations in the interplay between the
media attention and the public attention according to the
topics.
60. My Two Riyals …
• Unfiltered News and Google Trends are good data
sources to track media and public attention.
• The media and public attention are often at odds
• The media may be losing the power of agenda setting
in the online era, and the public may be seeking other
online sources to address their media needs.
!60
63. Multiplex network
• A multilayer network is a network made up by multiple
layers, each of which represents a given operation mode,
social circle, or temporal instance.
• In a multiplex network, each type of interaction between
the nodes is described by a single layer network.
Text from http://cosnet.bifi.es/network-theory/multiplex-networks/
Image from https://github.com/gajduk/social-networks-analysis-wan-bms
3
!63
67. Distribution of attention
strengths
Image from https://community.lithium.com/t5/Science-of-Social-Blog/The-Economics-of-90-9-1-The-Gini-Coefficient-with-Cross/ba-p/5466
Gini Index
!67
68. Gini coeff. of the attention strengths
for each node in NM and NP
Uniform
Distribution
Skewed
Distribution
!68
69. Gini coeff. of the attention strengths
for each node in NM and NP
The public attention
goes more equally to
other countries than
media attention does.
Uniform
Distribution
Skewed
Distribution
Patterns of paying
attention are quite diverse
!69
70. Well-known names of each
motif
Feed forward loop (FFL)Fan-out Cascade
Fan-in
Fully connected triad
Double feedback loop
!70
71. Limitations
• We model media attention and public attention from a
single service each.
• We set the scope of the analysis as the foreign news
coverage and the corresponding web search.
!71