Fake News, Echo Chambers, and Filter
Bubbles: Research and Policy
William H. Dutton @BiIIDutton
Quello Professor of Media and Information Policy
Quello Center, Michigan State University
@QuelloCenter
Presentation for a workshop at Bruegel on fake news and filter
bubbles, Brussels, 25 October 2017.
Panic
Fake
News
Filter
Bubbles
Echo
Chambers
Immediatefuture.co.uk
Medium.com
The Part Played by Search in Shaping Political Opinion
• Quello Center team in collaboration with the Oxford
Internet Institute (OII), University of Oxford and
Department of Communication, University of Ottawa
• Professor William H. Dutton (Quello)
• Dr. Bianca C. Reisdorf (Quello)
• Dr. Grant Blank (OII)
• Dr. Elizabeth Dubois (Ottawa)
• With the assistance of:
• Craig Robertson (PhD Student, Quello)
• Sabrina Ahmed (BA Student, Ottawa)
• Support from Google, with thanks to Jon Steinberg
Centrality of Information to
Democratic Processes
Mass Media
•News, Radio, Television, and the Fourth
Estate
The Internet and Search
•Search Engines, Algorithms, Social
Media, User Choice, and a Fifth Estate
The Role of the Internet, Search and
Social Media?
Enable citizens to make well political
decisions?
Distort the information citizens gain
access to and choose in taking
political decisions and actions?
Technological Determinism
• More Informed Rational Citizens, Voters
• Social Media Movements, Surges
• Filter Bubbles
Social Determinism
• Spiral of Silence
• Power Law
• Echo Chambers
Social and Technical Shaping of Democratic Processes
• Agenda Setting
• Cue Taking and Giving (“group think”) - Two-Step Flow
• Werther Effects
• Fifth Estate (enabled by search and social media)
Multiple Theoretical Perspectives
Cross-National Comparative Research
A User-Centered Perspective
Review of
Literature
Trace
Data
Survey of
7 nations
Britain
France
Germany
Italy
Poland
Spain
United States
14,000 Internet
Users, January
2017
Mitigating
Problems
Centrality of Search
for Info about
Politics
Diversity of
Sources/Viewpoints
Check, Confirm,
Information
Finding Info that is
New, Unexpected, or
Wrong
Seldom Block,
Unfriend, Censor
Others
Fifth Estate, Not
Mass Media
You are the ‘first
algorithm’
KEY
THEMES
The Centrality of Search
Origins in WWW as Mountain of Trash
Becoming the first place people go for information
One of the most common uses of the Internet
Politics is a limited, specialized topic of search
Frequency of Using a Search Engine
France Germany Italy Poland Spain UK US Total
Never
1.4 0.14 0.2 0.2 0.2 0.5 1.1 0.54
Less than monthly
1.4 2.02 0.8 0.1 0.6 2.5 3 1.5
Monthly
1.8 2.4 0.8 0.2 1.4 2.4 2.4 1.62
Weekly
12.1 17.5 6.9 2.7 6.5 14.2 11.3 10.1
Daily
22.5 28.9 19.3 21.8 19.8 24.3 20 22.4
Greater than once per day
60.9 49.0 72.1 75 71.4 56.2 62.3 63.85
Total N
1,972 1,972 1,979 1,992 1,989 1,961 1,995 13,859
Total percent
100 100 100 100 100 100 100 100
The Purpose of Search
3.19
3.19
3.22
3.44
3.57
3.66
3.87
3.97
0 1 2 3 4
Mean results out of 5
Find entertaining content
Politics and current events
Medical or health questions
Check accuracy of news,info
Look up news on topic, event
Look up fact(s)
Navigation to sites
Info about particular topic
0=never; 4=very often
The Reliability of Search:
A Learned Level of Trust
As reliable as other major sources, e.g., TV
Users in Poland, Italy, and Spain more trusting
Users in Germany, France, and Britain less trusting
One of first places to go for information about politics
Reliability of Different Sources
2.69
3.35
3.41
3.41
3.47
3.49
3.52
0 1 2 3 4
Mean results out of 5
Social media
Television
Newspapers
Online news
Family, friends, colleagues
Radio
Search engine results
0=not reliable at all; 4=totally reliable
Reliability of Search Engine Results
France Germany Italy Poland Spain UK US Total
1 Total unreliable
2.6 2.6 1.7 0.7 0.7 1.5 1.1 1.5
2
7.9 8.3 5.6 6.1 7.5 6.9 5.7 6.8
3
39.9 44.8 37 36.6 36.9 40.7 39 39.2
4
40.7 38.1 46.9 46.8 44.8 42.6 42.8 43.3
5 Total reliable
9 6.3 8.9 9.8 10.1 8.3 11.4 9.1
Total N
1,910 1,920 1,938 1,958 1,966 1,895 1,950 13,537
Total percent
100 100 100 100 100 100 100 100
The Diversity of Sources
Those interested in politics look at
multiple (4.5) sources of information
More than two (2.4) offline, and more
than two (2.1) online
Search engines: most frequent online
source
Multiple Sources of Information about Politics
1.53
1.82
2.13
2.24
2.31
2.50
2.51
N
ever
Som
etim
es
O
ften
Very
often
Mean results out of 4
Charities, religious groups
Political websites
Radio
Print news
Family & friends
Online sources
TV
Online Sources of Information about Politics
2.36
2.52
2.54
2.88
3.02
3.07
3.49
N
ever
R
arely
Som
etim
es
O
ften
Very
often
Mean results out of 5
Political website
Email
Online video platforms
Social media
Online sites of news & mags
Online news sites
Search engines
Diversity of Views Encountered Online
36 percent of sample read news they disagree
with ‘often’ or ‘very often'’
Diversity of Views Among People Communicated
with Online (Table 4.25)
• 15%: Views Different from you
• 65%: Mixed Beliefs
• 20%: Same as you
Users Check, Confirm, Information
Multiple approaches to confirming
information
Over 80 percent use search to check facts
Three fourths (74%) use search to check
information on social media
Practices Tied to Confirming a Story
2.70
2.87
2.92
3.08
3.15
3.16
N
ever
R
arely
Som
etim
es
O
ften
Very
often
Mean results out of 5
Look for opinion of trusted source
Ask friend or family to confirm
Check different news sources
Check major offline news
Confirm by searching online
Read something disagree with
Finding Information
Unexpected, New, and Wrong
76 percent occasionally or often find information they
were not looking for (serendipity) through search
48 percent ‘often’ learn something new – serendipity
Many recognize ‘wrong’ information
Occasionally change their opinion on issues
Relative Prevalence of Practices
1.27
1.68
2.13
2.35
0 .5 1 1.5 2 2.5 3
Mean results out of 5
Changed opinion on political issue
Find wrong information
Discovered important information
Learn something new
0=never; 3=often
Rarely Block or Unfriend
20 percent have unfriended or blocked someone who posted
because of differing political views or offensive content
12 percent block or unfriended a person who disagree with
political content they posted
15 percent posted content they worried would offend friends
or people they follow
Mitigating
Problems
Centrality of Search
for Info about
Politics
Diversity of
Sources/Viewpoints
Check, Confirm,
Information
Finding Info that is
New, Unexpected, or
Wrong
Seldom Block,
Unfriend, Censor
Others
Fifth Estate, Not
Mass Media
You are the ‘first
algorithm’
KEY
THEMES
• Google Studie
• Forschungsfragen
• Methodologie
• Erste Resultate
• Bedeutung
• Diskussion
Factors Shaping Individual Differences in Search
Factors Shaping Good Internet & Search Practices
in the Political Arena
Political
• Interest in Politics
• Online political participation
Internet
• Skills
• Levels of Internet Use
• Mobile, Next Generation Users
Information
• Diversity of Sources
• Learned Level of Trust
Operationalizing those ‘Vulnerable’
Interest in Politics: ‘Not at all’ v. ‘Somewhat
Interested, Interested, Very Interested’
Skill: Ability to use a search engine: ‘Bad, Fair, or
Poor’ v ‘Good or Excellent’
IF: Not at all interested in politics AND
Bad/Fair/Poor ability THEN rated VULNERABLE
Percent Vulnerable (No Interest, Low Skill)
0 0.5 1 1.5 2 2.5 3 3.5 4
US
France
Poland
Britain
Spain
Germany
Italy
Vulnerables
Vulnerable
Vulnerable
compose from
1-4% of
Internet Users
The Vulnerables?
The ‘vulnerables’ tend to be somewhat:
• Older, less educated, no children in home, retired,
unemployed, female
• Lower income, some disability
However, ‘vulnerables’ scattered across all
categories
• Difficult to target by demographics
• Socially identify, alert to problems, and nudge to develop
interests and skills
Nudging* Internet Users:
Encourage an interest in politics; make it interesting
Don’t undermine trust by demonizing the Internet and search
Recommend consulting multiple sources, on- and off-line
Identify good online practices
• Use the Internet and search
• Value diversity of views and social networks
• Check questionnable news & facts (use search)
*Nudge theory associated with Richard Thaler’s 2017 Nobel Prize .
Key Issues Moving Forward
Theoretical Perspectives
• Be more critical of deterministic perspectives
• Don’t underestimate users & social shaping of search
Interest in Politics
• Are those least interested & involved more vulnerable?
Skills in Search and Internet
• Are those offline or least skilled at greater risk?
Digital Media Literacy or Best to Nudge All Users?
• Avoid inappropriate regulation of content: don’t panic!
• Nudge individuals – users – to reduce risks (Richard Thaler)?
The Report Plus
Dutton, W.H., Reisdorf, B.C., Dubois, E., and Blank, G. (2017), Search
and Politics: The Uses and Impacts of Search in Britain, France,
Germany, Italy, Poland, Spain, and the United States, Quello Center
Working Paper available on SSRN: http://ssrn.com/abstract=2960697
Dutton, W.H. (2017), ‘Fake News, Echo Chambers, and Filter
Bubbles: Underresearched and Overhyped’: https://theconversation.com/fake-
news-echo-chambers-and-filter-bubbles-underresearched-and-overhyped-76688
Dutton, W. H. (2017), ‘Bubblebusters’, NESTA. http://readie.eu/bubblebusters-
countering-fake-news-filter-bubbles-and-echo-chambers/

Search&politics bruegel-25 oct2017

  • 1.
    Fake News, EchoChambers, and Filter Bubbles: Research and Policy William H. Dutton @BiIIDutton Quello Professor of Media and Information Policy Quello Center, Michigan State University @QuelloCenter Presentation for a workshop at Bruegel on fake news and filter bubbles, Brussels, 25 October 2017.
  • 2.
  • 3.
    The Part Playedby Search in Shaping Political Opinion • Quello Center team in collaboration with the Oxford Internet Institute (OII), University of Oxford and Department of Communication, University of Ottawa • Professor William H. Dutton (Quello) • Dr. Bianca C. Reisdorf (Quello) • Dr. Grant Blank (OII) • Dr. Elizabeth Dubois (Ottawa) • With the assistance of: • Craig Robertson (PhD Student, Quello) • Sabrina Ahmed (BA Student, Ottawa) • Support from Google, with thanks to Jon Steinberg
  • 4.
    Centrality of Informationto Democratic Processes Mass Media •News, Radio, Television, and the Fourth Estate The Internet and Search •Search Engines, Algorithms, Social Media, User Choice, and a Fifth Estate
  • 5.
    The Role ofthe Internet, Search and Social Media? Enable citizens to make well political decisions? Distort the information citizens gain access to and choose in taking political decisions and actions?
  • 6.
    Technological Determinism • MoreInformed Rational Citizens, Voters • Social Media Movements, Surges • Filter Bubbles Social Determinism • Spiral of Silence • Power Law • Echo Chambers Social and Technical Shaping of Democratic Processes • Agenda Setting • Cue Taking and Giving (“group think”) - Two-Step Flow • Werther Effects • Fifth Estate (enabled by search and social media) Multiple Theoretical Perspectives
  • 7.
    Cross-National Comparative Research AUser-Centered Perspective Review of Literature Trace Data Survey of 7 nations Britain France Germany Italy Poland Spain United States 14,000 Internet Users, January 2017
  • 8.
    Mitigating Problems Centrality of Search forInfo about Politics Diversity of Sources/Viewpoints Check, Confirm, Information Finding Info that is New, Unexpected, or Wrong Seldom Block, Unfriend, Censor Others Fifth Estate, Not Mass Media You are the ‘first algorithm’ KEY THEMES
  • 9.
    The Centrality ofSearch Origins in WWW as Mountain of Trash Becoming the first place people go for information One of the most common uses of the Internet Politics is a limited, specialized topic of search
  • 10.
    Frequency of Usinga Search Engine France Germany Italy Poland Spain UK US Total Never 1.4 0.14 0.2 0.2 0.2 0.5 1.1 0.54 Less than monthly 1.4 2.02 0.8 0.1 0.6 2.5 3 1.5 Monthly 1.8 2.4 0.8 0.2 1.4 2.4 2.4 1.62 Weekly 12.1 17.5 6.9 2.7 6.5 14.2 11.3 10.1 Daily 22.5 28.9 19.3 21.8 19.8 24.3 20 22.4 Greater than once per day 60.9 49.0 72.1 75 71.4 56.2 62.3 63.85 Total N 1,972 1,972 1,979 1,992 1,989 1,961 1,995 13,859 Total percent 100 100 100 100 100 100 100 100
  • 11.
    The Purpose ofSearch 3.19 3.19 3.22 3.44 3.57 3.66 3.87 3.97 0 1 2 3 4 Mean results out of 5 Find entertaining content Politics and current events Medical or health questions Check accuracy of news,info Look up news on topic, event Look up fact(s) Navigation to sites Info about particular topic 0=never; 4=very often
  • 12.
    The Reliability ofSearch: A Learned Level of Trust As reliable as other major sources, e.g., TV Users in Poland, Italy, and Spain more trusting Users in Germany, France, and Britain less trusting One of first places to go for information about politics
  • 13.
    Reliability of DifferentSources 2.69 3.35 3.41 3.41 3.47 3.49 3.52 0 1 2 3 4 Mean results out of 5 Social media Television Newspapers Online news Family, friends, colleagues Radio Search engine results 0=not reliable at all; 4=totally reliable
  • 14.
    Reliability of SearchEngine Results France Germany Italy Poland Spain UK US Total 1 Total unreliable 2.6 2.6 1.7 0.7 0.7 1.5 1.1 1.5 2 7.9 8.3 5.6 6.1 7.5 6.9 5.7 6.8 3 39.9 44.8 37 36.6 36.9 40.7 39 39.2 4 40.7 38.1 46.9 46.8 44.8 42.6 42.8 43.3 5 Total reliable 9 6.3 8.9 9.8 10.1 8.3 11.4 9.1 Total N 1,910 1,920 1,938 1,958 1,966 1,895 1,950 13,537 Total percent 100 100 100 100 100 100 100 100
  • 15.
    The Diversity ofSources Those interested in politics look at multiple (4.5) sources of information More than two (2.4) offline, and more than two (2.1) online Search engines: most frequent online source
  • 16.
    Multiple Sources ofInformation about Politics 1.53 1.82 2.13 2.24 2.31 2.50 2.51 N ever Som etim es O ften Very often Mean results out of 4 Charities, religious groups Political websites Radio Print news Family & friends Online sources TV
  • 17.
    Online Sources ofInformation about Politics 2.36 2.52 2.54 2.88 3.02 3.07 3.49 N ever R arely Som etim es O ften Very often Mean results out of 5 Political website Email Online video platforms Social media Online sites of news & mags Online news sites Search engines
  • 18.
    Diversity of ViewsEncountered Online 36 percent of sample read news they disagree with ‘often’ or ‘very often'’ Diversity of Views Among People Communicated with Online (Table 4.25) • 15%: Views Different from you • 65%: Mixed Beliefs • 20%: Same as you
  • 19.
    Users Check, Confirm,Information Multiple approaches to confirming information Over 80 percent use search to check facts Three fourths (74%) use search to check information on social media
  • 20.
    Practices Tied toConfirming a Story 2.70 2.87 2.92 3.08 3.15 3.16 N ever R arely Som etim es O ften Very often Mean results out of 5 Look for opinion of trusted source Ask friend or family to confirm Check different news sources Check major offline news Confirm by searching online Read something disagree with
  • 21.
    Finding Information Unexpected, New,and Wrong 76 percent occasionally or often find information they were not looking for (serendipity) through search 48 percent ‘often’ learn something new – serendipity Many recognize ‘wrong’ information Occasionally change their opinion on issues
  • 22.
    Relative Prevalence ofPractices 1.27 1.68 2.13 2.35 0 .5 1 1.5 2 2.5 3 Mean results out of 5 Changed opinion on political issue Find wrong information Discovered important information Learn something new 0=never; 3=often
  • 23.
    Rarely Block orUnfriend 20 percent have unfriended or blocked someone who posted because of differing political views or offensive content 12 percent block or unfriended a person who disagree with political content they posted 15 percent posted content they worried would offend friends or people they follow
  • 24.
    Mitigating Problems Centrality of Search forInfo about Politics Diversity of Sources/Viewpoints Check, Confirm, Information Finding Info that is New, Unexpected, or Wrong Seldom Block, Unfriend, Censor Others Fifth Estate, Not Mass Media You are the ‘first algorithm’ KEY THEMES
  • 25.
    • Google Studie •Forschungsfragen • Methodologie • Erste Resultate • Bedeutung • Diskussion Factors Shaping Individual Differences in Search
  • 26.
    Factors Shaping GoodInternet & Search Practices in the Political Arena Political • Interest in Politics • Online political participation Internet • Skills • Levels of Internet Use • Mobile, Next Generation Users Information • Diversity of Sources • Learned Level of Trust
  • 27.
    Operationalizing those ‘Vulnerable’ Interestin Politics: ‘Not at all’ v. ‘Somewhat Interested, Interested, Very Interested’ Skill: Ability to use a search engine: ‘Bad, Fair, or Poor’ v ‘Good or Excellent’ IF: Not at all interested in politics AND Bad/Fair/Poor ability THEN rated VULNERABLE
  • 28.
    Percent Vulnerable (NoInterest, Low Skill) 0 0.5 1 1.5 2 2.5 3 3.5 4 US France Poland Britain Spain Germany Italy Vulnerables Vulnerable Vulnerable compose from 1-4% of Internet Users
  • 29.
    The Vulnerables? The ‘vulnerables’tend to be somewhat: • Older, less educated, no children in home, retired, unemployed, female • Lower income, some disability However, ‘vulnerables’ scattered across all categories • Difficult to target by demographics • Socially identify, alert to problems, and nudge to develop interests and skills
  • 30.
    Nudging* Internet Users: Encouragean interest in politics; make it interesting Don’t undermine trust by demonizing the Internet and search Recommend consulting multiple sources, on- and off-line Identify good online practices • Use the Internet and search • Value diversity of views and social networks • Check questionnable news & facts (use search) *Nudge theory associated with Richard Thaler’s 2017 Nobel Prize .
  • 31.
    Key Issues MovingForward Theoretical Perspectives • Be more critical of deterministic perspectives • Don’t underestimate users & social shaping of search Interest in Politics • Are those least interested & involved more vulnerable? Skills in Search and Internet • Are those offline or least skilled at greater risk? Digital Media Literacy or Best to Nudge All Users? • Avoid inappropriate regulation of content: don’t panic! • Nudge individuals – users – to reduce risks (Richard Thaler)?
  • 32.
    The Report Plus Dutton,W.H., Reisdorf, B.C., Dubois, E., and Blank, G. (2017), Search and Politics: The Uses and Impacts of Search in Britain, France, Germany, Italy, Poland, Spain, and the United States, Quello Center Working Paper available on SSRN: http://ssrn.com/abstract=2960697 Dutton, W.H. (2017), ‘Fake News, Echo Chambers, and Filter Bubbles: Underresearched and Overhyped’: https://theconversation.com/fake- news-echo-chambers-and-filter-bubbles-underresearched-and-overhyped-76688 Dutton, W. H. (2017), ‘Bubblebusters’, NESTA. http://readie.eu/bubblebusters- countering-fake-news-filter-bubbles-and-echo-chambers/