Is a polarized society inevitable, where people choose to be exposed to only political news and commentary that reinforces their existing viewpoints? We examine the relationship between the numbers of supporting and challenging items in a collection of political opinion items and readers’ satisfaction, and then evaluate whether simple presentation techniques such as highlighting agreeable items or showing them first can increase satisfaction when fewer agreeable items are present. We find individual differences: some people are diversity-seeking while others are challenge-averse. For challenge-averse readers, highlighting appears to make satisfaction with sets of mostly agreeable items more extreme, but does not increase satisfaction overall, and sorting agreeable content first appears to decrease satisfaction rather than increasing it. These findings have important implications for builders of websites that aggregate content reflecting different positions.

1. Presenting Diverse Political Opinions:
How and How Much
Sean Munson, Paul Resnick
School of Information, University of Michigan

2. Is a polarized world inevitable?

5. Risks
of
polarization
&
bene1its
of
diversity

6. Competing theories
CHALLENGE AVERSION. People prefer
material that supports their opinions and avoid
challenging information. (Selective exposure &
homophily)
DIVERSITY SEEKING. People prefer
information that contains both challenging and
supporting opinions. (Stromer-Galley 2003, Pew
IALP 2004)
SUPPORT SEEKING. People prefer supporting
material but are indifferent to challenging
material (Garrett 2009)

7. Study goals
INDIVIDUAL PREFERENCES for opinion diversity.
EVALUATE SIMPLE PRESENTATION TECHNIQUES for
making challenging items more palatable.

8. Study design

9. Study design: Overall
1. Show people with known political biases a list of
links with a predicted percent of agreeable items.
2. Ask them how satisfied they are with the
opinions in the set of links

10. Study design: Subjects
Recruited via Mechanical Turk SUBJECTS
Restricted to people in US. Age: 34.3 years
median: 31 years,
standard dev: 11.8 years
Qualification task
Gender: 83 men, 87 women
• Demographic questions (gender, zip code,
age) Location: 37/50 states
• Political affiliation (7-point scales for party
and liberal-conservative)
• 3 questions about political knowledge
Quality control
During study, subjects randomly re-asked n
demographic questions. Responses from 5 Manipulation check: 30
subjects discarded for impossible or improbable Subjects: 40 (satisfaction)
replies. 38 (bias)

11. Study design: Subjects
Recruited via Mechanical Turk
Restricted to people in US.
strong
Qualification task democrat
• Demographic questions (gender, zip code,
age)
strong strong
• Political affiliation (7-point scales for party liberal conservative
and liberal-conservative)
• 3 questions about political knowledge
strong
republican
Quality control
During study, subjects randomly re-asked
demographic questions. Responses from 5
subjects discarded for impossible or improbable
replies.

12. Study design: Articles
Link source
• Items linked from 500 political blogs; blogs coded as liberal,
independent, conservative
• Daily: select 40 most-linked stories from the previous 24 hours that
had ≥2:1 liberal:conservative or conservative:liberal link ratio
• Filter out tweets, YouTube videos, Wikipedia articles, items not
matching predicted bias. Average of 23 items / bias / day remain.
Manipulation check
• 30 turkers. Shown links, asked if they agree or disagree. Each link seen
by ≥3 turkers.
• Kept links to which they reacted consistent with expectation.
Discarded others.

13. Study design: Experiment
2x3 factorial design.
• LIST LENGTH: 8 or 16 item lists
• PRESENTATION
Baseline: article title (linked) + abstract
Highlight: Agreeable items highlighted
Highlight+Order: Agreeable items highlighted and placed
first
• Varying PERCENT OF AGREEABLE ITEMS

15. Questions
ASSIGNED TO EITHER:
Satisfaction
“Suppose this was the front page of a political opinion aggregator.
How would you feel about the viewpoints represented in it?” (5
point Likert scale, Very dissatisfied to very satisfied)
Bias
“What, if any, is the political bias of this collection?” (5 point
Likert scale, Very Liberal to Very Conservative)
EVERYONE:
Why they gave the rating they did (open-ended)
Random demographics check

16. Results

17. Results: Diversity preferences

18. Coded as diversity-seeking if…
WANT MORE CHALLENGE
“e articles in this list showed some of both sides on some issues,
but on other issues like health care was rather one sided. If that and
a few other articles had been given two sides I would be completely
satisfied. I like to read both sides even though I am mostly
conservative.”
WOULDN’T WANT LESS CHALLENGE
“ere is an even distribution of right and le wing articles. I think
it is best to cover both sides of the issue.”
“I like that there are views from both Democrats and Republicans
and seems to be a great mix of both sides of the fence.”
Cohen’s
kappa:
0.89

19. Diversity preferences

20. Model for satisfaction

21. Support-seeking or challenge-averse?

22. Support-seeking or challenge-averse?

23. Support-seeking or challenge-averse?
30 SUBJECTS 10 SUBJECTS

24. Results: Presentation
BASELINE ★ HIGHLIGHT ★ HIGHLIGHT + ORDER
Ar#cle
#tle
Ar#cle
#tle
Ar#cle
#tle
Story
abstract
would
go
here.
And
so
Story
abstract
would
go
here.
And
so
Story
abstract
would
go
here.
And
so
on.
It
would
say
stuff.
And
have
opinion.
on.
It
would
say
stuff.
And
have
opinion.
on.
It
would
say
stuff.
And
have
opinion.
Ar#cle
#tle
Ar#cle
#tle
Ar#cle
#tle
Story
abstract
would
go
here.
And
so
Story
abstract
would
go
here.
And
so
Story
abstract
would
go
here.
And
so
on.
It
would
say
stuff.
And
have
opinion.
on.
It
would
say
stuff.
And
have
opinion.
on.
It
would
say
stuff.
And
have
opinion.
Ar#cle
#tle
Ar#cle
#tle
Ar#cle
#tle
Story
abstract
would
go
here.
And
so
Story
abstract
would
go
here.
And
so
Story
abstract
would
go
here.
And
so
on.
It
would
say
stuff.
And
have
opinion.
on.
It
would
say
stuff.
And
have
opinion.
on.
It
would
say
stuff.
And
have
opinion.
Ar#cle
#tle
Ar#cle
#tle
Ar#cle
#tle
Story
abstract
would
go
here.
And
so
Story
abstract
would
go
here.
And
so
Story
abstract
would
go
here.
And
so
on.
It
would
say
stuff.
And
have
opinion.
on.
It
would
say
stuff.
And
have
opinion.
on.
It
would
say
stuff.
And
have
opinion.

25. Highlighting

26. Highlighting + Agreeable First

27. Results: Presentation

28. Conclusions & Future Work

29. Good news and bad
DIFFERING INDIVIDUAL PREFERENCES for opinion diversity.
Challenge aversion is not an inherent human characteristic, but
neither is diversity seeking.
SIMPLE PRESENTATION TECHNIQUES, such as highlighting and
ordering agreeable items, do not make diverse opinions more
appealing to challenge-averse individuals.

30. Future work: Preferences
Distribution of preferences
Do individual preferences change over time?
Do individual preferences differ by topic?

31. Future work: Presentation
Continue to explore ways to make diversity
appealing to challenge-averse people.
Field trials for ordering
Social influence techniques
Ideas derived from NewsCube (Park et al, CHI 2009): show
agreeable item first or prominently, then indicate that there
are differing views.

32. Future work: Murky middle
Include people who are less extreme or who don’t fall
into neat categories.
strong
democrat
strong strong
liberal conservative
strong
republican

33. thanks!
Sean Munson samunson@umich.edu
Paul Resnick presnick@umich.edu
balance.projects.si.umich.edu
is project was supported by NSF Award IIS-0916099.