You are on page 1of 10

Confirmation Bias as a Source of Affective

Polarization in University of Colorado Students


Quinn Kudzma Park, Mikaela Lodl, Aaron Pineda, Rachel Saggau

Abstract: In a politically tense environment, such as in the United States today, peoples
opinions and biases heavily influence their outlook on the opposition and political issues.
Confirmation bias can be a large influencer on ones opinions, making it easier to find people
that agree with you, and lessening your exposure to opposing viewpoints. Our research delves
into how much one is aware of their confirmation bias and how this plays into affective
polarization of other viewpoints. Affective polarization refers to the emotional tag that
influences the processing and acceptance of information. Our survey asked questions to obtain a
general understanding of the participants demographic as well as their opinions on US policy and
opinions of the opposite political party. We then analyzed this information to see how much a
persons confirmation bias could be playing into their political opinions. Participants opinions
on members of opposing political parties were used to analyze their affect toward the other party.
The survey was distributed to University of Colorado, Boulder students in order to gather
necessary data on this phenomena and conclude the existence of partisan polarization stemming
from confirmation bias.

Introduction:
A defining feature in humans is our ability to make complex abstract decisions without
exhausting all cognitive resources. We make hundreds of decisions a day without a conscious
effort, and even the ones that require conscious thought can be made in a matter of minutes.
That is to say our cognitive capacity is not explicitly devoted to every small decision. It would be
cognitively debilitating to have to process every choice consciously. So we rely on unconscious
cues and implicit decision making to process information more quickly. A clear example of this
can be seen when people engage in voting. There are a number of massively complex issues that
by themselves would require tremendous amounts of cognitive computation.
So how then do people vote so efficiently? Affective polarization suggests that the
emotional tag assigned to a particular issue serves as a filter by which people process and
accept information. Furthermore, it suggests that attachment to parties is socialized early in life
through family and social networks. One possible root of affective polarization may be
confirmation bias. Taber and Lodge define confirmation bias as: such that when [people are]
free to choose what information they will expose themselves to people will seek out confirming
over disconfirming arguments (Lodge, 2006). Confirmation bias, while surely not the sole
cause of affective polarization, may play a part in the emotional salience politics often has and
thereby affecting the emotional tags individuals assign to topics.
Background and Objectives:
Affective polarization is grounded in a social-identity approach to partisanship, which is
a deeply rooted and enduring psychological attachment to one of the political parties. The public
has become increasingly polarized over the last four decades leading people to engage in more
biased information search and processing, and thus tend to be less trusting of people who identify
with the political opposition (Pew Research Center, 2014). The salience of this is best
exemplified in the percentage of people who view the other party as a threat to the nations
well-being. The percentage of people who dislike the opposing party has increased as well and
a majority of the respective groups from both parties actually views the opposing parties as a
threat to the nations well-being. Pew Research Center has found that political polarization in the
United States has increased dramatically in the past two decades, stating that ideological
overlap between the two parties has diminished (Pew Research Center 2014). So, then what
human behaviors have led to increased partisan attitudes? One possible root is confirmation bias.
Studies have observed a confirmation bias effect in those associations to being primed stupid
led to a gradual decrease in performance, whereas associations to being clever did not.
(Bengtsson, 2013). This priming suggests that confirmation bias has to do with self-esteem,
defined as the evaluative dimension of self-concepts (Harter and Baumeister, 1993). In the
context of affective polarization, this may imply that an opposing partisan identification calls
into question a persons self-esteem. This gives way to confirmation bias. It is advantageous to
maintain and accept information that reconfirms those prior evaluations of self-concepts.
In order to suggest the legitimacy of the presence of confirmation bias in affective voting,
a primed and unprimed survey will be sent out to college attendees or college educated
individuals. Although, one might expect that educated individuals might be aware of their own
bias, there is evidence to suggest otherwise, and therefore, it is expected that individuals will not
be aware of their degree of bias. When primed we expect individuals to be aware of this bias or
at least actively thinking about it. As such we expect there to be a noticeable difference in the
responses from the primed and unprimed survey regarding the extremeness of answers as well as
in the accuracy of assessing those of differing political affiliation.
The University of Colorado, Boulder was an opportune place to conduct this study due to
the population attending. The population is made up of individuals who are at least high school
educated (or have the equivalent G.E.D.), and are pursuing higher education. Because our study
focuses on educated individuals and the proximity and availability of students to the researchers,
the University of Colorado Boulder was a top choice. We chose to include educated participants
to emphasize that confirmation bias is present regardless of higher education.

Methods:
I. Participants
The survey included 32 participants of which 31 were within the range of 18 to 24 years
of age and 1 was within the range of 25-34 years. The participants consisted of 11 males and 21
females . All participants were sourced from the University of Colorado Boulder. Of the
participants, 12 were in the college of Engineering, 16 were in the college of Arts and Sciences,
3 were in the school of Business, and one was in Media. Furthermore, 19 identified as
Democrats, 3 identified as Republicans, and 10 identified as Other. Text responses for Other
included Independent for two individuals, and each of the following for one individual:
Unaffiliated, Socially progressive, financially conservative, None, Libertarian, Free-
Thinker, and Green Party. Two who responded other did not fill in a response to classify
themselves.

II. Survey
In order to test the degree to which individuals are aware of bias a short survey was sent
out via email and social media (i.e Facebook and Slack). There were two different surveys sent
out tha served to separate participants into two groups: one that received a prime about
confirmation bias and one that did that. Other than the inclusion of prime, the surveys were be
identical. When participants were primed with the idea that they may have bias, it was expected
that they would be more accurate in assessing the opposite political party. The prime for the
survey consisted of the following statement:

Taber and Lodge posit three mechanisms of partisan or biased processing:


1. A prior attitude effect , whereby people who feel strongly about an issue--even when
encouraged to be objective and leave their preferences aside--will evaluate supportive arguments
as stronger and more compelling than opposing arguments;
2. A disconfirmation bias, such that people will spend more time and cognitive resources
denigrating and counter arguing attitudinally incongruent than congruent arguments; and
3. A confirmation bias, such that when free to choose what information they will expose
themselves to people will seek out confirming over disconfirming arguments.

Firstly the participants were be asked to declare basic identifying characteristics such as
age, gender, political affiliation and field of study. Then a series of questions were asked that are
typically considered to be partisan non-neutral. Questions such as Would you prefer to have a
walkable community or more room? (Liberals tend to favor walkable communities while
conservatives tend to favor things to be more spaced out) and Do you think there should be
more or less foreign aid support? (Liberals tend to favor more foreign aid while conservatives
favor less) were included in the survey. Then a follow up question was asked: What percentage
of the national budget do you think is allocated to foreign aid? The actual percentage is less
than 1% but the average American drastically overestimates that. This was to aid in measuring
the number of people who were political sophisticates. Political sophisticates hold their biases
more strongly and thus an important measure to account for. The remaining questions were of of
similar nature looking to asses political attitudes. (Pew Research Report, 2014). The specific
questions can be found in the Appendix.
In consideration of the lack of funds available for this research, the methods used in
assessing the degree to which confirmation bias affects affective polarization were limited to a
survey. Due to time constraints of the researchers, who are also college students, focus groups
were not formed as initially posited. The aforementioned survey was sent out to two groups. The
responses were then compared between the two groups, to see if the group with the confirmation
bias prime had any variation in responses. Since confirmation bias is by nature difficult to
measure, we were looking for any measureable difference between the two groups. After
reviewing the data, the results were be contrasted to the opposing group (confirmation bias vs.
control). The results will then be discussed along with any possible implications.

Results:
The initial five questions were intended to portray the demographic makeup of
participants. With this data, it is possible to conclude that there was demographic consistency
between the primed and unprimed surveys. As shown in Table 1, each category for these
responses contains a similar proportion in both the primed and unprimed surveys. Therefore, our
data analysis for differences between the two is not likely due to a difference in the type of
participants between the
surveys, and instead will
reveal information on
confirmation bias.

Table 1. The percentage of


responses for both the primed
and unprimed surveys are
demonstrated with this bar
graph. Most categories have
highly similar percentages, for
example, there were 38% male
responses in the primed survey
and 32% male responses in the
unprimed survey.

For the question concerning housing size and vicinity to businesses, the primed and
unprimed responses saw a majority preference for a smaller house and a majority expectation
that an individual with an opposing political affiliation would have a preference for a larger
house, as demonstrated in Table 2. The responses were consistent with the percentages
associated with political affiliation in both the primed and unprimed surveys, i.e. Democrats tend
to prefer smaller houses and Republicans tend to prefer larger houses. The percentages are
consistent with the reported political affiliation suggesting the reliance on partisan cues. The
reliance on these cues is intrinsically important in the determination of confirmation bias. That is
to say, it is an important control that must be accounted for in order to asses the validity of bias
in decision making. These results demonstrate partisan reliance, wherein an individuals political
affiliation aids in decision making.

Table 2. This table shows responses to


the question Would you prefer to live
in a community where with the
option of larger houses farther away
from schools and businesses, or smaller
houses closer to schools and
businesses. The first two options
represent the participants preference,
while the third and fourth options
represent the participants expectation
for the opposing partys preference.

Responses for primed and unprimed surveys both has high proportions (around half of
participants) choosing 1% for the percent of the national budget allocated to foreign aid. As
mentioned above, this would demonstrate a high proportion of political sophisticates, and thus
biases are expected to be held more strongly in this study by those individuals. Among
Americans less than 25% correctly assign the allocation of the National budget that is
contributed to foreign aid. Thus, there is an abnormal number of political sophisticates
participating in this survey, which limits the application of these findings. Furthermore, for the
question on whether the opposing partys agenda posed a threat to the nations well being, the
primed survey recorded no responses for either definitely yes or definitely no; this is shown in
Table 3. This suggests that because the primed group was made aware of their potential
confirmation bias, they were better able
to assess the opposing political party.

Table 3. For this question, responses from


the unprimed survey were more spread out
than the primed survey.
Discussion:
The research into confirmation bias has provided valuable insight into the political
ideologies of students at the University of Colorado Boulder. The main identifying factors that
were used in the study were age, political affiliation, gender and field of study. Hopefully this
will provide a basis from which to determine the growth or decline of confirmation bias within
the college student populus throughout the following years.
Due to having an insufficient amount of time and resources we found it best to adapt our
method of research from focus groups to a survey. A careful design and accommodation was
made in the survey to ensure a measureable and controlled effect to which confirmation bias may
play a roll. It is worthy to note that confirmation bias is more difficult to test for in a survey
rather than a discussion setting, and thus affects the confidence of our results. During the
development phase it became apparent that conducting the original goal of four control focus
groups, and four focus groups informed of what confirmation bias is, was not going to be
achievable within our timeframe with the budget and resources at hand. The plan for the focus
group was then adapted to fit into one survey that covered all points the focus group would. This
required unique strategies to ensure the participant was not actively or subconsciously cued to be
aware of their bias when answering questions about opposing political views, while still relaying
how much the participant did to confront this bias.
The questions that were specifically geared towards participants opinions on
controversial political topics and consideration of opposing viewpoints, helped reveal how
much the participants combatted their confirmation bias. Using a survey restricts the possibility
of getting more specific data on the explicit differences in opinions of the opposing political
parties between those participants that knew about confirmation bias, and had been thinking
about it, and those that were not told about confirmation bias, and not directed to think about it
when responding.
Within the survey, a significant difference in the amount of participants that identified as
female rather than male was observed. The female participants made up roughly two thirds of the
studies population; males made up the other one third of the population. This could lead to
inaccurate results if there were outliers within the male participant population. In the future a
study which addressed this difference and tried to close the gap so that the number of male and
female participants was equal would be beneficial to the studys accuracy..
Keeping all of this in mind, using the survey allowed us to reach a broader population
and provided some qualitative data, and even more quantitative data. By reviewing the acquired
data we determined that the demographic remained consistent among both groups that
participated in the survey. The majority declared that they were democratic and chose similar
answers to the survey questions. It was also found that the group that received the prime was
more likely to answer negatively when asked about their respective opposing party, which
demonstrates an decrease in partisan reliance when aware of the existence of confirmation bias.
In the future researchers might consider conducting multiple focus groups along with the survey,
as it will allow them to reach a more accurate conclusion.
While there are limitations in the application of results, being that only a narrow
demographic was accounted for the results provide some political insight into the decision
making of partisan individuals. Typically speaking extremely partisan individuals hold their
biases more strongly than moderates. So, the observed effects are not representative of a broader
population without further research.
Given the acknowledged limitations there were statistically significant differences in
some but not all of the questions between the primed and unprimed survey. Both surveys
indicated a reliance on party identification to make decisions or were at the least ideologically
consistent with the identified party. This fits the trend for political sophisticates and suggests
some degree of attitude stability. These are important things to control for when assessing the
role of confirmation bias. It is hard to measure in and outside setting which is why the controls
and additional associations are important to rule out.
The individuals that received the prime, more often accurately assessed the opposing
political parties beliefs than those who did not. This suggests the mere prime about bias, at least
among political sophisticates, helps individuals avoid bias. Given the small survey size and the
demographic limitations caution should be taken in broadening these conclusions to a more
general population.

References

1. (2014). Political Polarization in the American Public. Pew Research Center.


2. Bengtsson, S.L., & Penny , W.D. (2013). Self-associations influence task performance
through Bayesian inference . Frontiers in Human Neuroscience,
3. Bodi N, Keri S, Nagy H, Moustafa A, Myers CE, Daw N, Dibo G, Taka ts A, Bereczki
D, Gluck MA (2009) Reward-learning and the novelty-seeking personality: a between-
and within-subjects study of the effects of dopamine agonists on young Parkinsons
patients. Brain 132:23852395.
4. Byrne, K. A., Davis, T., & Worthy, D. A. (2016). Dopaminergic Genetic Polymorphisms
Predict Rule-Based Category Learning. Journal of Cognitive Neuroscience, 28(7), 959
970. http://doi.org/10.1162/jocn_a_00942
5. Calabresi P, Gubellini P, Centonze D, Picconi B, Bernardi G, Chergui K, Svenningsson
P, Fienberg AA, Greengard P (2000) Dopamine and cAMP-regulated phosphoprotein 32
kDa controls both striatal long-term depression and long-term potentiation, opposing
forms of synaptic plasticity. J Neurosci 20:8443 8451
6. Doll, B.B., Hutchison, K.E., & Frank, M.J. (2011). Dopaminergic Genes Predict
Individual Differences in Susceptibility to Confirmation Bias. Journal of Neuroscience , .
7. Harter, S., and Baumeister, R. F. (1993). Self-Esteem: The Puzzle of Low Self- Regard.
New York,NY: Plenum Press, 87116.
8. Menza M. The personality associated with Parkinsons disease. Curr Psychiatry Rep
2000; 2: 4216.

Appendix
1. Please indicate your age range.
a. 18-24
b. 25-35
c. 35 and older
2. How do you define your gender?
a. Male
b. Female
c. Other
3. What academic college or school are you in?
a. Business
b. Arts and Science
c. Education
d. Engineering/Applied Science
e. Media
f. Music
g. Other
4. Are you interested in politics?
a. Definitely yes
b. Probably yes
c. Might or might not
d. Probably no
e. Definitely no
5. What is your political affiliation?
a. Democrat
b. Republican
c. Other (fill in blank)
6. Would you prefer to live where
a. The houses are larger and farther apart, but schools, stores, and restaurants are
several miles away
b. The houses are smaller and closer to each other, but schools, stores, and
restaurants are within walking distance.
7. Would an individual with an opposing political affiliation prefer to live where
a. The houses are larger and farther apart, but schools, stores, and restaurants are
several miles away
b. The houses are smaller and closer to each other, but schools, stores, and
restaurants are within walking distance.
8. Do you think the United States should increase or decrease foreign aid?
a. Increase
b. Decrease
c. Remain the same
9. Would an individual with an opposing political affiliation think the US should increase or
decrease foreign aid?
a. Increase
b. Decrease
c. Remain the same
10. Do you think the opposing political partys agenda is a threat to the nations well-being?
a. Definitely yes
b. Probably yes
c. Might or might not
d. Probably no
e. Definitely no
11. Would you marry/date someone of the opposite political party?
a. Definitely yes
b. Probably yes
c. Might or might not
d. Probably no
e. Definitely no

You might also like