Better Informed or Stay Naïve? Revisiting Different Types of Selective Exposure and the Impact on Political Learning

ABSTRACT

This study extends the concept of selective exposure by examining different types of selectivity, including topical selectivity (entertainment vs. hard news), information channel selectivity (social media vs. traditional media), and ideological selectivity (like-minded vs. cross-cutting exposure). Drawing on a two-wave panel survey of American netizens (N = 834), this study revisits the debate about the potential of selective exposure to enhance or erode political learning. Results show that both topical and information channel selectivity directly reduce audience’s political knowledge while ideological selectivity does not significantly affect political learning. Topical selectivity reduces political knowledge indirectly through decreased offline political discussion, while ideological selectivity enhances political knowledge gain indirectly through increased offline political discussion. Online political discussion fails to mediate the relationships between all three types of selectivity and political knowledge since it does not have a significant relationship with political knowledge. The findings imply that the prevalence of entertainment media and social media use may even enlarge existing political knowledge gaps. In addition, the political knowledge gained by ideological selective exposure is also dangerous because it could be less comprehensive, unbalanced, and attitudinally biased.

As information floods our daily lives and the media landscape continues to evolve with growing diversity, “selectivity” becomes an inevitable strategy to manage today’s high-choice media environment (Stroud, 2008). Selective exposure is the motivated selection of messages matching one’s interests, beliefs, and desirability (Stroud, 2018). Studies on selective exposure in political contexts have mainly focused on partisan selective exposure to talk shows, cable news channels, and news websites that present an unbalanced image with clear partisan affiliation (Stroud, 2008, 2010). However, selective exposure decisions could also be guided by other factors such as personal interests, medium preferences, and informational motivations. Thus, revisiting the concept of selective exposure is imperative as it can help explain how people respond to various kinds of information they encounter (Knobloch-Westerwick & Meng, 2011; Stroud, 2010).

Selective exposure can occur at different levels for various reasons (Ohme & Mothes, 2020). At the first level, selectively clicking a political news story online is particularly affected by political interest. However, the decision to spend more time reading the news deeply seems more strongly driven by other factors, such as social endorsement. It is important to examine the commonly adopted criteria for news audiences to select certain messages or media to consume. The consequences of different selective patterns for individuals and society are also worthy of investigation.

Selectivity is thought to occur whenever alternatives exist for media users. In addition to motivated selections, habitual news use patterns also represent selections. Thus, this study goes beyond partisan selective exposure and extends its conceptual map by examining the political implications of different types of selectivity. We believe this is important because the complex formation of one’s media diet comes from consecutive selections, swinging between choices, information seeking, and avoidance. Understanding selectivity in different dimensions provides an initial step for us to further explore the political outcomes of selective exposure in a holistic and integrated manner.

Following Stroud (2018), we discuss three types of selectivity (i.e., topical selectivity, information channel selectivity, and ideological selectivity) by examining how they affect political knowledge in different ways. From a holistic perspective, the three types of selectivity represent three separate but related dimensions of media use: what, where, and who. Specifically, topical selectivity touches upon one’s personal interests, that is, what themes and topics a user is interested in. Information channel selectivity refers to where a user prefers to seek information. Ideological selectivity implies who the users are and how they ascribe their partisan identity to their media selections. Thus, the three components of selectivity may together build a user’s media diet and subsequently affect political knowledge.

Political knowledge guides citizens to participate in democratic decision-making effectively. Most studies on selective exposure take political knowledge as an indicator of political predispositions that predict selective exposure (Stroud, 2008). However, from the information processing perspective, it is possible that additional information received during selective exposure could affect one’s understanding about politics (Knobloch-Westerwick & Meng, 2011). Thus, in this study, we treat political knowledge as an outcome variable that could be affected by different selective exposure patterns. Our results also highlight the different roles of reasoning behaviors (i.e., online and offline political discussions) that may happen in face-to-face and computer-mediated communication settings after selective exposure.

Conceptualizing Different Types of Selective Exposure

Selective exposure refers to individuals’ tendencies to favor information that reinforces their preexisting beliefs and to avoid contradictory information (Stroud, 2018). Scholars have drawn on various theoretical approaches to explain why selective exposure happens, among which the most common one is reinforcement-seeking to achieve cognitive equilibrium (Festinger, 1962). Selective exposure can reinforce media effects if the political ideology of the media is consistent with the audience’s mind (e.g., Knobloch-Westerwick & Meng, 2011).

While selective exposure is usually driven by ideology or social identity, it has been manifested in various contexts (e.g., political communication, health communication, Internet use, and entertainment consumption). It arises when media has been captured by political parties so that different outlets represent certain ideological wings (Iyengar & Hahn, 2009). Selective exposure also occurs at the crossroads of entertainment and hard news, driving scholars to worry that the plethora of media choices will result in increased audience exposure to entertainment content instead of political information (Prior, 2005). In addition, with the advent of new digital communication technologies, selective exposure has been endowed with new formats and meanings. This could happen as a distinct decision to adopt a certain channel before audiences select news stories they prefer (Knobloch-Westerwick, 2014). According to the Pew Research Center (2021), only 5% of American adults used at least one social media platform in 2005, but as of 2021, 72% were using some type of social media.

Incorporating these situations, we specify three types of selectivity: topical selectivity (entertainment vs. hard news), information channel selectivity (social media vs. traditional media), and ideological selectivity (like-minded vs. cross-cutting exposure). We use these types to conceptually and operationally extend the concept of selective exposure and understand how the three types of selectivity could inform the public.

Previous researchers viewed selective exposure either as choice and habit or as preferences and tendencies (Knobloch-Westerwick, 2014). From the media choice perspective, selective exposure is engaged in by an individual user at a specific time, when they select a certain media message while disregarding alternative ones. From the perspective of preferences, selective exposure captures an individual’s general tendency to favor a specific kind of media content. Conceptualizing selective exposure as preferences could better capture the psychological process that occurs when people are surrounded by abundant options. With this in mind, we conceptualize three types of selectivity based on individuals’ use preferences for one kind of message or medium over another.

Topical selectivity focuses on themes and topics of media preference, specifically the situation that audiences favor entertainment content over hard news information. Skipping hard news but approaching entertainment can be understood as avoidance of distressing political news (Kim et al., 2013; Villi et al., 2022). Moreover, from a mood management perspective, people who are experiencing low affective arousal or boredom are more likely to engage with topical selectivity, seeking thrilling entertainment stimuli rather than calming content (Luong & Knobloch-Westerwick, 2021). However, the relationship between entertainment and hard news is not necessarily a zero-sum game (Lelkes, 2020). Indeed, the growth of paid-for entertainment content online might benefit news publishers as news audiences may get used to paying for all forms of online media (Fletcher & Nielsen, 2020). In this study, we do not assume there are trade-offs between entertainment and hard news consumption, but our focus is on comparative preference for entertainment over hard news.

Information channel selectivity refers to users’ preference to approach social media channels as opposed to traditional media channels in news consumption. It targets media channels rather than topics or attitudes of the media content. As the offspring of digital media development, social media platforms have prompted a growing tendency of users to move from traditional media channels (e.g., newspapers, magazines, cable TV) toward social media for news consumption. One major reason for news audiences’ turn to social media is that they are seeking social connections, intensive information, and chances for self-disclosure (Nabity-Grover et al., 2020; Roberts & David, 2020). Even though many social media users do not primarily use it to consume news, they are incidentally exposed to more news sources than non-social media users (Fletcher & Nielsen, 2018). Some scholars are concerned that selective exposure could be further fostered in the social media context due to the greater ease of selecting like-minded networks (Sunstein, 2018) and the algorithmic information sorting and filtering on the platforms. However, others argue that people may wish to maintain awareness of diverse political views online (Garrett, 2009), and social media can help to expand connections with weak ties (Beam et al., 2020). Within this debate, information channel selectivity is generally believed to have political implications as well.

Ideological selectivity highlights how partisans ascribe their partisan positions to like-minded media entities and filter content based on their ideological congruence (Bennett & Iyengar, 2008; Stroud, 2008). While topical selectivity reflects preference between content types and information channel selectivity focuses on the selection between media channels or modalities, ideological selectivity is guided by social identity and political beliefs. There is empirical evidence supporting the existence of ideological selectivity, or partisan selective exposure (Chen et al., 2020; Stroud, 2010). For example, Iyengar and Hahn (2009) found that in the US, Republicans preferred to watch Fox News and tended to avoid news from CNN and NPR, while Democrats exhibited the opposite pattern. In a changing media and political environment, partisans can vary their political attentiveness and subject themselves to biased information flows even if their media diets are diversified and balanced (J. W. Kim & Kim, 2021). Such ideological selectivity can be leveraged by their political orientation (e.g., populist partisanship), level of political commitment, or level of affinity to their party (Mothes & Ohme, 2019).

In summary, we specify three distinct types of selectivity (i.e., topical, information channel, and ideological) to represent the “what,” “where,” and “who” dimensions of news consumption in today’s high-choice media flows. There is broad consensus that selective exposure to media information is a precondition of media effects (Stroud, 2008). Nonetheless, media effects research that details and specifies a broad selective exposure paradigm is scarce. Mapping out the effects of selective exposure on political learning in detail not only addresses the interplay between individual selectivity and Internet affordances, but also helps to explain how this interplay could secure or deteriorate a well-informed citizenry.

Selective Exposure and Political Knowledge

Democratic theorists believe that political participation is the cornerstone of democracy and citizens’ engagement in the democratic process should be based on relevant, abundant, and accurate political knowledge (Habermas & McCarthy, 1984). However, in the US, there has long been a concern that citizens are being entertained rather than politically informed by media (Postman, 2005). Studies have also identified a huge knowledge gap among US citizens (Liu & Eveland, 2005). Does selective exposure improve or hinder the acquisition of political knowledge? Sunstein (2001) proposed that the ability to customize political information will have a detrimental impact on democracy as media users become less likely to encounter and learn from information that challenges their viewpoints. Others tend to be more optimistic, believing that as media choice increases, incidental exposure to the news may work as a “knowledge leveler” (Oeldorf-Hirsch, 2018). While this debate is far from settled, we revisit the effect of different types of selectivity on political knowledge in a recent US sample.

The basic premise of this study is that people’s media use is governed by their preferences for topics, channels, and values, and they select programs to satisfy these preferences. People who desire entertainment are usually less motivated to read and learn about politics. Some scholars believe that mere exposure to news could produce learning, but a preference for entertainment exposure, reflecting the tendency to avoid hard news, decreases the possibility of passive learning (Prior, 2005). More critically, the preference for entertainment may lead people to pay more attention to the “entertaining” aspects of politics (e.g., soft news) and not actually produce any positive learning effects (Prior, 2005). Therefore, we assume a negative relationship between topical selectivity (entertainment content vs. hard news) and political knowledge.

Likewise, selectivity of information channels (social media vs. traditional media) questions the well-established positive effects of news use in political learning (Eveland, 2001). A meta-analysis reveals no evidence of positive learning effects on social media across platforms, knowledge types, time, and geographic locations (Amsalem & Zoizner, 2023), pointing to further inquiries on why social media is incapable of advancing learning. The prevalence of “brief, intermittent attendance to news,” or “news snacking” (Ohme & Mothes, 2023, p. 1) demonstrates how social media users allocate limited attention to digital news. This habit can impair the positive role of news exposure in political learning (Ohme & Mothes, 2023). The availability of multiple choices on social media could easily draw users’ attention, energy, and cognitive capacity from news reading to other activities, hindering their political learning. In addition, compared with traditional media run by professional journalists and editors, user-generated messages on social media may be of dubious quality. Exposure to low-credibility information deteriorates political learning as well. S. Lee and Xenos (2019) suggested that social distraction, low-quality social media content, and lower motivation to seek information can explain why social media use is detrimental to political knowledge.

The relationship between ideological selectivity and political knowledge has long been debated. One may doubt that like-minded exposure rules out alternative views and traps the reinforced audience in their echo chambers, which may limit their knowledge gains (Sunstein, 2018). A more convincing argument with empirical evidence is that ideological selectivity advances political learning because additional information received during selective exposure, even if like-minded, instigates elaboration and reorganization of political knowledge (Knobloch-Westerwick & Meng, 2011). Specifically, ideological selectivity contributes to partisan learning, especially securing knowledge of issues that are most relevant to one’s own group (e.g., abortion rights, racism, or immigration; Chen, 2018; Kleinberg, 2023).

We propose the following hypotheses on the direct effect of selective exposure on political knowledge:

H1a:

For topical selectivity, exposure to entertainment (vs. hard news) negatively predicts political knowledge.

H1b:

For information channel selectivity, social media use (vs. traditional media use) negatively predicts political knowledge.

H1c:

For ideological selectivity, like-minded exposure (vs. cross-cutting exposure) positively predicts political knowledge.

The Mediating Role of Online and Offline Political Discussion

The importance of political discussion was raised in the early work on personal influence by Katz and Lazarsfeld (1995), who described how the persuasive potential of mass messages is influenced by interpersonal interactions. Discussing political issues with others provides additional opportunities for political reasoning, which can be helpful in forming political beliefs and obtaining new knowledge. In line with the communication mediation model (Cho et al., 2009), selective exposure, which is a stimulus of new use, could initiate or demobilize the political reasoning process (i.e., political discussions), which then affects knowledge gains.

The development of new media has brought together processes that were formerly ascribed to mass or interpersonal domains (Walther, 2017). Interpersonal political discussion can now be performed in both online and face-to-face settings in a more frequent and timely manner. Empirical studies have documented that large social network size, the strength of weak ties online, and the quality of reasoning discussion that one experiences through online political discussion all have positive effects on political outcomes, such as increasing online participation and political knowledge (Valenzuela et al., 2012).

Selective exposure, regardless of type, could trigger discussions on topics people have been frequently exposed to, as selectivity implies one’s interests and preferences. Intuitively, people like to talk about things they are passionate about. In this way, topical selectivity with a preference for entertainment provides sufficient talking points regarding entertaining events among family members and peers but can squeeze out deliberative discussions on politics due to limited time and capacity. Political entertainment programs such as late-night talk shows do not sufficiently grab users’ attention or facilitate cognitive learning processes (Hollander, 2005; Prior, 2005). Y. M. Kim and Vishak (2008) found that compared to news media, entertainment media are less effective in helping audiences acquire factual information. These entertainment products may not be able to act as prerequisites for deliberative reasoning by expression, taking others’ viewpoints, and creating mutual understanding (Goodin, 2000). Therefore, we anticipate that the more topical selectivity people exhibit, the less political discussion they engage in both online and offline.

The relationship between information channel selectivity with social media preference and political discussion is debatable. Some worry that social media creates more like-minded rather than cross-cutting discussions because social media has brought about greater alignment between individuals’ preferences and the media content they consume (Prior, 2007). On the other side, an optimistic view holds that informational social media use could provide diverse opinions to users, thereby encouraging people to fulfill their civic obligations, including online political expression and offline discussion (Gil de Zúñiga et al., 2012). The debate still continues with no consensus, but a more agreeable argument is that heavy social media use mobilizes online political discussion, no matter whether it happens with like-minded peers or among cross-cutting networks. Moreover, such online discussion patterns can be carried over to offline settings (Gil de Zúñiga et al., 2012).

Ideological selectivity is closely related to strong political attitude. It is conceivable that citizens whose views have been confirmed by repeated like-minded exposure tend to have greater opinion strength (H. Lee et al., 2015). People with stronger political attitudes more easily retrieve partisan cues from memory, and their attitudes are more readily available to serve as a basis for discussion and expressive behaviors, either to show in-group support or to counterargue the out-groups (Moon, 2013). In addition, agreement embedded in partisan selective exposure can serve as a crucial resource for political discussions that consolidate one’s political views (Mutz, 2006). Thus, ideological selectivity may enhance political discussion online and offline.

We propose the following hypotheses on the relationship between different types of selective exposure and political discussion:

H2:

For topical selectivity, exposure to entertainment (vs. hard news) negatively predicts a) online political discussion and b) offline political discussion.

H3:

For information channel selectivity, social media use (vs. traditional media use) positively predicts a) online political discussion and b) offline political discussion.

H4:

For ideological selectivity, like-minded exposure (vs. cross-cutting exposure) positively predicts a) online political discussion and b) offline political discussion.

Previous research suggests that political discussion, as the key reasoning process, contributes to political learning (Liu & Eveland, 2005). As a meta-analysis by Amsalem and Nir (2021) shows, when people talk more about political issues, no matter whom they talk with, their political learning is improved. Empirical studies offer evidence that the more people discuss politics or news offline, the more they know about politics (Eveland & Hively, 2009). People not only learn from the news they read, but also get additional information from their trusted interpersonal sources.

In online settings, users have additional space to discuss and participate in politics. Some empirical evidence supports that online sharing could be an effective way to induce reasoning (Beam et al., 2016). However, others doubt that an ideal online public sphere can be achieved (Dahlberg, 2001). Online discussion may be narrow, isolated, and superficial, failing to promote deliberative and in-depth thinking. Therefore, it is still uncertain whether online political discussion could significantly enhance political knowledge. To address this uncertainty, we propose the following hypothesis and research question:

H5:

Offline political discussion positively predicts political knowledge.

RQ1:

What is the relationship between online political discussion and political knowledge?

Taken together, we propose a mediating role of offline political discussion on the relationships between three types of selective exposure and political knowledge. Political knowledge is not directly gained only through news exposure; the learning process could be boosted through induced political talks and reflections. Building on H2 to H5, we propose the following hypotheses:

H6a:

For topical selectivity, exposure to entertainment (vs. hard news) indirectly and negatively affects political knowledge through offline political discussion.

H6b:

For information channel selectivity, social media use (vs. traditional media use) indirectly and positively affects political knowledge through offline political discussion.

H6c:

For ideological selectivity, like-minded exposure (vs. cross-cutting exposure) indirectly and positively affects political knowledge through offline political discussion.

Since the relationship between online political discussion and political knowledge is inconclusive, we propose the following research question:

RQ2:

What, if any, is the indirect effect of topical selectivity (entertainment vs. hard news) (RQ2a), information channel selectivity (social media vs. traditional media) (RQ2b), and ideological selectivity (like-minded vs. cross-cutting exposure) (RQ2c) on political knowledge through online political discussion?

The theoretical framework of this study is visualized in Figure 1.

Figure 1. Theoretical Framework.

Method

Sample

The data for this study were drawn from a two-wave panel study of US adults (aged 18 and older). The two-wave panel design was administered by Dynata, a professional online survey company. The two-wave panel design enables us to trace the changes in respondents’ behaviors and perceptions over time and helps to address causal inferences. As we wanted to survey people who use social media to consume news, we employed quota sampling to make the first wave sample match the US Facebook user population in terms of gender and age. For detailed sampling procedure and sample profile, please see Appendix A.

Measures

Selective Exposure

We operationalized three types of selectivity as described above. To measure entertainment selective exposure, we asked respondents to indicate their interests in reading different news types. To measure social media selective exposure, we asked the respondents to report how often they get news from a list of traditional media platforms compared with a list of social media platforms. To measure partisan selective exposure, we divide respondents’ frequency of media use that aligns with their partisanship using scores for the use of like-minded and cross-cutting media. For details about the measurements and statistics of three types of selective exposure, please see Appendix B.

Political Discussion

Following Yamamoto and Nah (2018), we specified political discussion into online and offline modes. In W1, respondents were asked to rate on a 7-point scale (1 = strongly disagree; 7 = strongly agree) their agreement with the following: “In general, I often talk about politics or current events with my friends or family online/offline.” The scores were used to measure political discussion online (M = 3.11, SD = 1.83) and offline (M = 4.26, SD = 1.85).

Political Knowledge

We adopted measures from Shehata and Strömbäck (2021) to ask respondents a list of factual questions about politics and current affairs without guidance from others or the Internet. Correct answers to each question were given the value 1, whereas incorrect and don’t know responses were coded as 0. W1 includes eight questions about general political information and specific issues. An index of political knowledge was constructed by summing up the scores (M = 5.10, SD = 2.06, W1: Cronbach’s α = .74; W2: Cronbach’s α = .75). W2 includes ten questions mostly about issues and events that occurred between the two waves (i.e., in July and August 2022) to capture how much new political information the respondents had gained since the first wave (M = 5.11, SD = 2.98, W2: Cronbach’s α = .84). Appendix B presents the political knowledge items from the two waves.¹

Results

To test the hypotheses, PROCESS macro template 4 was adopted with 10,000 bias-corrected bootstrap resamples and 95% confidence intervals (CIs) (Hayes, 2017). Statistical significance (p < .05) is achieved when lower-bound and upper-bound CIs do not include zero. For detailed model specifications, please see Appendix C.

The results (summarized in Table 1) show a significant and negative relationship between topical selectivity and political knowledge (β = −.10, SE = .69, p < .001), indicating that preference for entertainment content directly leads to lower political knowledge. There is also a significant negative relationship between information channel selectivity and political knowledge (β = −.11, SE = .64, p < .001), indicating that preference for social media over traditional media directly decreases users’ political knowledge as well. H1a and H1b are both supported. The relationship between ideological selectivity and political knowledge shows the opposite pattern (RQ1: β = .07, SE = .48, p < .05), which implies that like-minded partisan selective exposure directly enhances political knowledge. H1c is also supported.

Table 1. Regression Coefficients in the Mediation Models.

	Online discussion (W1)	Offline discussion (W1)	Political knowledge (W2)
Model 1: Outcome of topical selectivity
ESE (W1)	−.08 (.52)*	−.20 (.52)***	−.10 (.66)***
Online discussion (W1)			.01 (.05)
Offline discussion (W1)			.06 (.05)*
Political knowledge (W1)	.06 (.03)	.20 (.03)***	.52 (.05)***
Age	−.05 (.00)	−.03 (.02)	.12 (.01)***
Gender (male = 1)	.09 (.12)**	.02 (.13)	.07 (.16)**
Income	−.03 (.04)	.01 (.04)	.05 (.05)
Education level	.03 (.04)	.07 (.04)*	.06 (.06)*
Ethnicity (white = 1)	−.04 (.16)	−.02 (.17)	−.06 (.22)*
Partisanship	.02 (.03)	.04 (.03)	.04 (.04)
Social media use	.43 (.05)***	.30 (.05)***	−.04 (.06)
Constant	2.52 (.48)***	3.77 (.49)***	−.23 (.66)
R^2	.23***	.20***	.49***
Model 2: Outcome of information channel selectivity
SMSE (W1)	.17 (.54)***	.02 (.51)	−.11 (.64)***
Online discussion (W1)			.02 (.05)
Offline discussion (W1)			.08 (.05)*
Political knowledge (W1)	.07 (.03)	.24 (.04)***	.54 (.04)***
Age	−.09 (.01)*	−.09 (.01)*	.09 (.01)**
Gender (male = 1)	.09 (.13)*	.02 (.13)	.08 (.16)**
Income	−.04 (.04)	−.02 (.04)	.03 (.05)
Education level	.02 (.04)	.06 (.04)	.06 (.06)*
Ethnicity (white = 1)	−.05 (.17)	−.04 (.17)	−.06 (.21)*
Partisanship	.02 (.03)	.03 (.03)	.02 (.04)
Entertainment news use	.26 (.04)***	.18 (.04)***	−.03 (.05)
Constant	1.42 (.54)***	2.67 (.55)***	−.08 (.64)
R^2	.17***	.13***	.49***
Model 3: Outcome of ideological selectivity
PSE (W1)	.05 (.36)	.12 (.36)**	.07 (.48)*
Online discussion (W1)			.00 (.06)
Offline discussion (W1)			.08 (.06)*
Political knowledge (W1)	.05 (.04)	.17 (.04)***	.51 (.05)***
Age	−.04 (.01)	−.01 (.01)	.15 (.01)***
Gender (male = 1)	.10 (.14)*	.04 (.15)	.08 (.19)**
Income	−.04 (.05)	.04 (.05)	.04 (.06)
Education level	.03 (.05)	.06 (.05)	.07 (.06)*
Ethnicity (white = 1)	−.01 (.20)	−.02 (.20)	−.06 (.25)
Partisanship	.02 (.03)	.05 (.03)	.04 (.04)
Entertainment news use	.10 (.05)*	.08 (.05)	.00 (.07)
Social media use	.38 (.06)***	.29 (.06)***	−.04 (.08)
Constant	1.37 (.51)**	1.40 (.51)**	−2.74 (.68)***
R^2	.22***	.16***	.47***

Notes. Cell entries are standardized coefficients with standard errors in parentheses; *p < .05; **p < .01; ***p < .001.

In addition, topical selectivity is negatively related to both offline political discussion (β = −.20, SE = .52, p < .001) and online political discussion (β = −.08, SE = .51, p < .05). H2a and H2b are both supported. Information channel selectivity is positively related to online political discussion (β = .17, SE = .49, p < .001) but has a non-significant relationship with offline political discussion (β = .02, SE = .51, p = .68). H3a is supported, while H3b is rejected. Ideological selectivity is positively related to offline political discussion (β = .12, SE = .36, p < .01), but has a nonsignificant relationship with online political discussion (β = .05, SE = .39, p = .22), leading us to reject H4a and support H4b.

Concerning the relationship between political discussion and political knowledge, the results only demonstrate a positive relationship between offline political discussion and political knowledge (Model 1: β = .06, SE = .05, p < .05; Model 2: β = .08, SE = .05, p < .05; Model 3: β = .08, SE = .06, p < .05), supporting H5. The relationship between online political discussion and political knowledge is not significant (Model 1: β = .01, SE = .05, p = .74; Model 2: β = .01, SE = .05, p = .63; Model 3: β = .00, SE = .06, p = .91).

Mediation analysis results (Appendix C Table C1) show that both topical selectivity (B = −2.45, SE = .69, 95% CI = −.3.80 to −1.10) and information channel selectivity (B = −2.36, SE = .64, 95% CI = −3.61 to −1.11) directly decrease political knowledge. However, ideological selectivity increases political knowledge (B = 1.14, SE = .48, 95% CI = .20 to 2.08).

Moreover, topical selectivity negatively affects political knowledge through decreased offline political discussion (H6a: B = −.33, SE = .16, 95% CI = −.67 to −.04), but not through online political discussion (RQ2a:B = −.02, SE = .07, 95% CI = −.18 to .10). Information channel selectivity fails to indirectly affect political knowledge through either online (H6b: B =.07, SE = .11, 95% CI = −.14 to .31) or offline political discussion (RQ2b: B = .03, SE = .07, 95% CI = −.11 to .17). Ideological selectivity indirectly enhances political knowledge through offline political discussion (H6c: B = .15, SE = .08, 95% CI = .02 to .34), but not through online political discussion (RQ2c: B = .00, SE = .03, 95% CI = −.06 to .08). In sum, H6a and H6c are supported, and H6b is rejected. The finalized model is presented in Figure 2.

Figure 2. Tested Mediation Models.

Notes. Cell entries are standardized coefficients; *p < .05; **p< .01; ***p < .001.

Discussion

In this study, we go beyond the narrow concept of partisan selective exposure by incorporating three types of selective exposure into our analysis to provide a detailed and integrated map that delineates their effects on political learning. The conceptual framework of the study views selective exposure from two perspectives. First, by differentiating three types of selectivity (i.e., topical [what], information channel [where], and ideological [who]), this study provides explicit and detailed analyses of their effects on political knowledge. Specifically, both topical selectivity and information channel selectivity erode political knowledge, while ideological selectivity not only increases political knowledge directly but also enhances political learning through increased offline political discussion. These analyses help to further clarify the conceptual differences of different types of selectivity within the broad selective exposure paradigm.

Second, in addition to differentiation and specification, our conceptualization also views selective exposure in an integrated framework by understanding [what], [where], and [who], which are inextricably intertwined in the media consumption process. When consuming hard news, political ideology can play a role in deciding partisan media use, while media channels affect the availability of news information at the same time. The situation becomes even more complicated in social media settings where news readers’ attentiveness can be allocated to endless choices, from hard news to entertainment and from like-minded news to alternative views. In this sense, our integrated framework of the three types of selectivity is more of a dynamic and fluid cycle rather than being simplistic and isolated.

In addition to advancing a conceptual framework of selective exposure for future research, this study takes a pioneering step in investigating its effects on political learning. While previous studies mainly addressed the effect of prior knowledge as a stable trait that could guide users’ media selections (Johnson et al., 2009), we found that selective exposure further influences people’s knowledge levels. Less-informed citizens are more likely to prefer entertainment content over other content (i.e., topical selectivity), which will further decrease their political knowledge. Young people and digital natives have not yet developed a deep political understanding, but they prefer to use social media for news rather than reading news from other traditional sources (i.e., information channel selectivity). The fragmented attention and dubious information quality from entertainment and social media exposure further deteriorate their political learning, especially for active social media entertainment seekers.

The effects of ideological selectivity on political knowledge show a different pattern. Strong partisans are inclined to use their existing political knowledge and beliefs to guide their media selections (i.e., partisan selective exposure). Although it can be biased, such exposure can increase political knowledge for several reasons. First, it increases the opportunities for exposure to factual political information and helps to refresh one’s knowledge structure. Second, information is especially memorable for partisans when they agree with the messages (Eagly et al., 2000). Third, people who selectively expose themselves to partisan media tend to be strong partisans who are often more likely to counterargue attitude-discrepant messages. Counterarguing dissonant information can raise people’s awareness and memory of the information, policy, and political figures from the opposite party as well because they need to organize and structure their reasoning (Taber & Lodge, 2006). These together build up the opportunity structure of ideological selectivity for political learning.

These results are alarming because these types of selective exposure are widely adopted and combined by news users under the growing trend of partisan media prevalence in the digital age. Only ideological selectivity, which has long been criticized as a major cause of political polarization, is likely to contribute to political learning. Topical selectivity with a preference for entertainment decreases political knowledge not only directly but also indirectly by undermining offline political discussions. As strong ties are more likely to be from homogenous networks, discussing entertainment content, even the entertaining aspects of political news, will further amplify the problem of “amusing ourselves to death” (Postman, 2005).

In addition to specifying three types of selectivity and examining their effects on political knowledge, we also identify the different effects of online and offline political discussion on political knowledge. Contradicting previous findings based on the communication mediation framework (e.g., Jung et al., 2011), our results do not support the path from online political discussion (reasoning) to political knowledge (orientation). This finding also rebuts the expected democratic value of the online public sphere (Habermas et al., 1989): even though online political discussion could be triggered by information channel selectivity, it is not powerful enough to improve political knowledge (Park, 2017). The incapability of online discussions to advance political learning can be due to different reasons, such as repetitive like-minded discussions (Brauer et al., 1995) or the spread of misinformation and hate speech online, which hinders the deliberative potential of political discussions (Quandt, 2018). It could also be because of a lack of in-depth exchange of ideas where the information flow is dominated by algorithms, digital traffic business models, and the attention economy (D. H. Kim & Desai, 2021). Overall, this finding could inspire future research on the quality and deliberative potential of online political talk (Papacharissi, 2004).

Before concluding the study, several limitations should be acknowledged. First, scholars believe the relationship between selective exposure and political knowledge is reciprocal (Knobloch-Westerwick, 2014). Although we employed two-wave panel data to test the effects of selective exposure on political knowledge, definite claims about causality cannot be made. To check the robustness of our study, we conducted additional analyses of reversed models. The results (Appendix D) support that our proposed model has more robust support than the reverse ones. In addition, theoretically, the proposed model was based on the theory and evidence from prior research with the most logical, plausible, and general causal ordering of variables. Future researchers could implement longitudinal multi-wave designs to further understand the reciprocal relationship between selective exposure and political knowledge.

Second, although we employed quota sampling, we could not control the representativeness of the final sample in the analyses. The retention rate between waves was not that high, which also undermines the generalizability of our findings. However, the retention rate should be acceptable as it is similar to those of other two-wave panel studies conducted in similar time periods (e.g., S. Lee & Jones-Jang, 2022). The single measurement of political discussion leaves us unable to examine what types of political discussion (e.g., cross-cutting discussion and like-minded discussion) are more likely to be generated by a certain type of selective exposure and then affect political knowledge in different ways. A refinement of the political discussion measurement could help to further the development of the literature.

Last but not least, we used dichotomous items to measure people’s preference for entertainment over hard news, social media over traditional media, and like-minded over cross-cutting views. However, in real life, users combine different types of selectivity when consuming news. Topics, channels, and political ideology may work together to constitute people’s media landscape and affect their knowledge scope. Future researchers could consider more sophisticated measurements of selective exposure and explore how the integration of selectivity could affect political learning.

Despite the limitations, this study extends the conceptual map of selective exposure with its clear explication of the concepts and detailed investigation of their political consequences. People select their media diets and modalities not only based on partisanship, but also out of their personal interest, taste, and digital media literacy. Individual preferences in different dimensions work in an integrated form and interplay with digital media affordances, ultimately affecting people’s knowledge gains. We find a pessimistic situation in that neither topical selectivity nor information channel selectivity contribute to political learning and may even enlarge existing political knowledge gaps. In addition, the political knowledge gained through ideological selectivity is also dangerous because it could be unbalanced, less comprehensive, and attitudinally biased.

The effects of selective exposure on political learning do not stop at individual levels. The ripple effects (Hsu et al., 2015) of informedness backsliding can spread to both offline and online communities and may further enhance political polarization at the macro-level (Brauer et al., 1995). These results alert us to consider the real and present dangers of the onslaught of social media and entertainment media. Before we hand over education, journalism, and politics to the business demands of comedies, soap operas, game shows, and tailored social media content, we need to acknowledge how these media shape our lives and identify the ways we can, in turn, utilize them to serve our ultimate goals of achieving a well-informed citizenry.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/08838151.2024.2341031

Additional information

Funding

The work was supported by the Direct Grant, Faculty of Social Science, Chinese University of Hong Kong project (no. 4052233).

Notes on contributors

Jing Guo

Jing Guo is a doctoral candidate at the School of Journalism and Communication, The Chinese University of Hong Kong. Her research interests include social media and political communication.

Hsuan-Ting Chen

Hsuan-Ting Chen (PhD, The University of Texas at Austin) is an Associate Professor at the School of Journalism and Communication, The Chinese University of Hong Kong. Her research addresses the uses of digital media technologies and their impact on individuals’ daily lives, political communication processes, and democratic engagement.

Shuning Lu

Shuning Lu (PhD, University of Texas at Austin) is an Assistant Professor in the Department of Communication at North Dakota State University, USA. Her research focuses on media uses and effects, political communication, and social and political implications of digital media technologies.

Notes

¹ Independent sample t-tests were conducted to identify if partisanship (Republican vs. Democrat) made a difference in people’s knowledge on certain issues that may affect their knowledge scores. Knowledge scores did not differ between Republicans and Democrats (W1: t = −1.56, p > .05; W2: t = −1.54, p > .05). Invariance across partisanship was supported.