Original ArticlesSelective exposure partly relies on faulty affective forecasts
Introduction
A well-functioning democracy requires citizens to consume a diversity of views in the “marketplace of ideas” (Milton, 1644/1890). Yet, extensive research demonstrates that individuals prefer information that confirms, rather than disconfirms, their prior beliefs. This phenomenon – known as “selective exposure” or the “congeniality bias” (Arceneaux et al., 2012, Frey, 1986, Hart et al., 2009, Iyengar and Hahn, 2009) – has grown particularly pernicious as citizens gain tools to easily filter what they see, hear, and read (Sunstein, 2009). Indeed, many scholars have expressed growing concern with political polarization and the effect it has on democratic institutions (Iyengar and Westwood, 2015, Westfall et al., 2015).
Across five experiments, we demonstrate that selective exposure is partly driven by faulty affective forecasts. Specifically, political partisans systematically overestimate the extremity of negative affect that will result from exposure to opposing views. We observe such biased forecasts across communication medium (verbal vs. written), communication author (U.S. president vs. U.S. senator vs. voter), and the political spectrum (liberal vs. conservative). Additionally, we test whether underestimation of agreement with the content of opposing partisan communication drives this forecasting error, and whether correcting mistakenly extreme affective forecasts can reduce selective exposure.
Although research on selective exposure has a long history, recent polarization of American political discourse and increased control by partisans over the information they consume have reignited interest in this phenomenon. From a strictly normative standpoint, information serves a purpose: awareness of both sides of an issue should be more beneficial to decision-making and social functioning than remaining blissfully ignorant of all facts supporting the opposite view (for discussion, see Golman, Hagmann, & Loewenstein, 2017).
However, despite the benefits of holding accurate beliefs, the phenomenon of selective exposure to agreeing information has been well-documented in social psychology (Frey, 1986), political science (Iyengar and Hahn, 2009, Sears and Freedman, 1967), and communications (Stroud, 2008). For example, one of the earliest studies on selective exposure demonstrated that mothers were more likely to listen to arguments that supported their beliefs regarding hereditary and environmental factors in childrearing than arguments that contradicted their beliefs (Adams, 1961). More recently, in the domain of political communication, conservatives in an experiment preferred to read articles from the conservative site Fox News, whereas liberals preferred to read articles from more liberal sources such as CNN and NPR (Iyengar & Hahn, 2009). These effects persist even with financial incentives on the line (Frimer, Skitka, & Motyl, 2017). Recent research has also examined how presentation order and structure moderate this phenomenon (Fischer et al., 2011, Jonas et al., 2001).
Prior empirical and theoretical work examining the origins of selective exposure has treated this tendency as an effort to mitigate cognitive dissonance – an unpleasant state of psychological arousal evoked by the presence of contradictory cognitions (Festinger, 1957, Festinger, 1964). Indeed the earliest demonstrations of selective exposure in social psychology were conducted under the umbrella of the theory of cognitive dissonance, and thus evidence of selective exposure was treated as confirming the theory’s predictions (e.g., Adams, 1961). Later meta-analytic evidence supports this interpretation (Hart et al., 2009). Specifically, selective exposure increases when individuals expect to experience greater feelings of threat and conflict, such as in situations when they recently affirmed their views (Jonas et al., 2001) or made an irreversible decision (Frey & Rosch, 1984).
Although the literature on selective exposure has inferred that people avoid opposing views because cognitive dissonance causes negative affect (e.g., Laurin, 2018, Wegener and Petty, 1994), this work has not, to our knowledge, directly examined the affective consequences of such exposure. Instead, these studies typically employ a choice paradigm, theorizing that the selected choice (pro-attitudinal information) is more enjoyable than the foregone one (counter-attitudinal information). Thus, research has not addressed whether individuals accurately predict how aversive exposure to opposing views actually is. Is it possible that individuals make daily choices about which news to read, which radio stations to listen to, and even which friends to socialize with based on faulty predictions about the affective consequences of these choices?
In the present research, we build on methodology from the literature on affective forecasting (Wilson and Gilbert, 2003, Wilson and Gilbert, 2005) to measure the positive and negative affect that political partisans expect to experience when confronted with opposing views. We focus our investigation on the accuracy of those expectations, testing whether exposure to opposing views is as aversive as anticipated, or whether individuals make a systematic error in forecasting their affect. In addition, we test whether mistakenly extreme affective forecasts partially underpin selective exposure – one of the most damaging drivers of contemporary political polarization.
An extensive body of research has demonstrated that individuals overestimate their negative affective reactions to a wide variety of events (for relevant reviews see Wilson and Gilbert, 2003, Wilson and Gilbert, 2005). Specifically, individuals systematically mis-forecast both the intensity of their affective reactions (i.e., the impact bias; Gilbert et al., 2002, Morewedge and Buechel, 2013, Sieff et al., 1999) and the duration of their affective reactions (i.e., the durability bias; Gilbert et al., 1998, Wilson et al., 2000). Studies examining the cause of these mispredictions have concluded that the error is often driven by focalism: the tendency to overweight the affective impact of the target event while underweighting the affective impact of other, non-target events (Mallett et al., 2008, Wilson et al., 2000; but see also Levine, Lench, Kaplan, & Safer, 2012).
Although research on affective forecasting has demonstrated that individuals overestimate their negative reactions to events ranging from divorce, to job rejection, to inter-racial social interaction (Gilbert et al., 2004, Gilbert et al., 1998, Mallett et al., 2008, Buechel et al., 2017; but see also Levine et al., 2012), this work has not addressed one of the most common sources of potential negative affect: exposure to the disagreeing views of others. Indeed, in our own pilot data (N = 208) examining individuals’ enjoyment of a variety of daily activities, we found that people who hold strong opinions on an issue rated policy discussion with holders of opposing views as more aversive than any other activity listed, including household chores, yard work, and a visit to the dentist (details available in the Supplementary Online Material).
In the present research, we test whether such aversion is partly unjustified. We use the methods and framework offered by the literature on affective forecasting to examine whether individuals systematically mis-predict their affective reactions to opposing views.
Multiple research programs suggest that individuals may underestimate their level of agreement with a piece of communication from across the political aisle. First, prior research on partisan conflict has documented the phenomenon of “naïve realism” – the tendency to consider one’s own views to be essentially accurate and based in reality, and free of the biases that plague the views of disagreeing others (Pronin, Gilovich, & Ross, 2004). To the extent that individuals consider the holders of opposing views to be subject to misinformation, bias or nefarious motives, they see those views as more extreme and more homogenous than they actually are. This tendency toward exaggerating the extremity of views on the other side has been termed “false polarization” (Keltner and Robinson, 1993, Robinson et al., 1995, Sherman et al., 2003). Research on false polarization suggests that partisans may expect specific communications from the other side to be more extreme in their partisanship and have less overlap with their own views than is actually the case.
Second, prior research on affective forecasting has documented the phenomenon of “focalism” – the tendency to overweight the affective impact of the target event and underweight the affective impact of all other non-target events (Mallett et al., 2008, Wilson et al., 2000). To the extent that partisans maintain a greater-than-warranted focus on the partisan identity of the speaker they might mis-predict how much the other side will talk about political content (vs. other tangentially related topics). Such a misprediction about the balance of political vs. non-political content in a communication might lead to a misprediction of one’s agreement with the communication content overall.
Across our studies, we use a variety of stimuli to test the prediction that individuals underestimate their level of agreement with communications supporting opposing political perspectives. We find evidence for both this underestimation of agreement, and the hypothesis that it leads to an affective forecasting error: individuals expect to experience more negative affect when consuming content from across the aisle because they expect to have fewer beliefs in common with the other side than is actually the case.
Section snippets
Research overview
Across five studies, we test three overarching hypotheses. First, in Studies 1–3, we test whether individuals systematically overestimate their negative affective reactions. We do so by comparing forecasted affect to experienced affect. Second, in Study 1 and Study 3, we test whether this affective forecasting error (the difference between forecasted affect and experienced affect) is underpinned by an underestimation of agreement (the difference between forecasted agreement and experienced
Study 1
Study 1 took place minutes after President Trump’s inauguration speech on January 20, 2017. We launched the study immediately after the video recording of the speech became publicly available in order to minimize the chance that our participants would have previous exposure to the speech. Participants were Clinton voters who first imagined watching President Trump’s inauguration speech and forecasted their affective reactions to it, as well as the degree to which they expected to agree with the
Study 2
Study 2 extends our investigation by testing participant reactions to content produced by voters instead of a professionally authored speech. It also used written, rather than spoken, content. Finally, Study 2 used a between-subjects design.
Study 3
Study 3 builds upon Studies 1 and 2 in several ways. First, we examined affective forecasts from both sides of the political aisle. Second, we again change stimuli in order to test the generalizability of our phenomenon. Participants in Study 3 viewed videos by current United States senators. Third, we again tested the extent to which participants underestimate their agreement with the content of a communication from the opposing side, now in a between-subjects design.
Study 4
In Study 1, we documented that the faulty affective forecasts predicted participants’ reported level of selective exposure. However, that evidence suffers from the shortcomings of a correlational design and the fact that the measure of selective exposure was hypothetical. In Study 4, we directly test this causal mechanism by manipulating people’s affective forecasts and observing how this changes real information consumption choices. We further test whether the change in selective exposure is
Study 5
We designed Study 5 to provide a conceptual replication of Study 4 while alleviating concerns of experimental demand. Specifically, participants chose whether to watch a video by an agreeing vs. disagreeing senator. We predicted that participants would be more willing to watch a video by a disagreeing senator when they saw that other participants had rated the video as not as aversive as anticipated.
General discussion
Across five experiments we demonstrate a robust error in individuals’ affective forecasts when faced with the prospect of consuming opposing political views. Specifically, participants find exposure to opposing views to be substantially less aversive than expected. We document this phenomenon across a variety of stimuli, communication modalities, and across the political spectrum. We also document this pattern with both within- and between-subjects experimental designs.
We find both
Acknowledgements
The authors gratefully acknowledge support from the Harvard Mind-Brain-Behavior Initiative and the Harvard Foundations of Human Behavior Initiative. The authors wish to thank Jennifer Lerner, Will Hart, and two anonymous reviewers.
References (53)
Recent research on selective exposure to information
Advances in Experimental Social Psychology
(1986)- et al.
Liberals and conservatives are similarly motivated to avoid exposure to one another's opinions
Journal of Experimental Social Psychology
(2017) - et al.
Affective forecasting
Advances in Experimental Social Psychology
(2003) Reduction of cognitive dissonance by seeking consonant information
The Journal of Abnormal and Social Psychology
(1961)- et al.
The short-run and long-run effects of behavioral interventions: Experimental evidence from energy conservation
American Economic Review
(2014) - et al.
Polarized political communication, oppositional media hostility, and selective exposure
The Journal of Politics
(2012) - et al.
Exposure to opposing views on social media can increase political polarization
Proceedings of the National Academy of Sciences
(2018) - et al.
The relationship between mental representations of welfare recipients and attitudes toward welfare
Psychological Science
(2017) - et al.
Impact bias or underestimation? Outcome specifications predict the direction of affective forecasting errors
Journal of Experimental Psychology: General
(2017) - et al.
The novelty penalty: Why do people like talking about new experiences but hearing about old ones?
Psychological Science
(2017)
A theory of cognitive dissonance
Conflict, decision, and dissonance (Stanford studies in psychology; 3)
Threat and selective exposure: The moderating role of threat and decision context on confirmatory information search after decisions
Journal of Experimental Psychology: General
Information seeking after decisions: The roles of novelty of information and decision reversibility
Personality and Social Psychology Bulletin
What comes to mind
The Quarterly Journal of Economics
The trouble with Vronsky: Impact bias in the forecasting of future affective states
The surprising power of neighborly advice
Science
The peculiar longevity of things not so bad
Psychological Science
Immune neglect: A source of durability bias in affective forecasting
Journal of Personality and Social Psychology
Information avoidance
Journal of Economic Literature
Feeling validated versus being correct: A meta-analysis of selective exposure to information
Psychological Bulletin
Red media, blue media: Evidence of ideological selectivity in media use
Journal of Communication
Fear and loathing across party lines: New evidence on group polarization
American Journal of Political Science
Confirmation bias in sequential information search after preliminary decisions: An expansion of dissonance theoretical research on selective exposure to information
Journal of Personality and Social Psychology
Estimating and testing mediation and moderation in within-subject designs
Psychological Methods
Imagined ideological differences in conflict escalation and resolution
International Journal of Conflict Management
Cited by (38)
But what if I lose the offer? Negotiators’ inflated perception of their likelihood of jeopardizing a deal
2024, Organizational Behavior and Human Decision ProcessesExperimental studies of conflict: Challenges, solutions, and advice to junior scholars
2023, Organizational Behavior and Human Decision ProcessesWhy is exposure to opposing views aversive? Reconciling three theoretical perspectives
2022, Current Opinion in PsychologyYou can't handle the truth! Conflict counterparts over-estimate each other's feelings of self-threat
2022, Organizational Behavior and Human Decision ProcessesCitation Excerpt :We used Senator Bernie Sanders and Senator Ted Cruz as our Liberal and Conservative target senators, respectively because at the time they held the most liberal and conservative voting records in the Senate and were highly familiar to the participants in our pool. Video clips were the most recent speeches uploaded to the YouTube channels of the respective senators at the time of the study (for a similar methodology, see Dorison et al., 2019). Both threat and anger were measured using four Likert items, presented in a randomized order and answered on 9-point scales from 0: “Not even the slightest bit” to 8: “More strongly than ever before.”
Toward a psychology of attitude conflict
2022, Current Opinion in PsychologyCitation Excerpt :In line with prior research, we define attitude conflict as the competitive disagreement between individuals (rather than groups) with regard to beliefs, values, and preferences [11,24,33]; see also [3].
False polarization: Cognitive mechanisms and potential solutions
2022, Current Opinion in Psychology