What Makes Fact-Checking Effective? Exploratory Research into the Role of Linguistic Politeness

Abstract

This study examined how the linguistic politeness of fact-checking messages affects audiences’ belief accuracy and message acceptance by shaping their perceived face threat and psychological reactance. In 2 experiments with national convenience samples of South Korean adults, we manipulated linguistic politeness by using either honorific or non-honorific sentence endings—a well-established marker of linguistic politeness in Korean—in fact-checking news articles correcting misinformation about COVID-19 (Study 1) and refugees (Study 2). Study 1 (N = 240) found that fact-checking articles using honorific sentence endings were perceived as less face-threatening than those presenting identical content with non-honorific endings, which in turn increased belief accuracy and the perceived persuasiveness of the articles. Study 2 (N = 296) replicated the findings of Study 1 and extended them by demonstrating: (1) the effect of using honorific (vs. non-honorific) sentence endings on recipients’ perceived face threat was not moderated by their conspiratorial predisposition or need for cognition; (2) the effects of perceived face threat were mediated by psychological reactance; and (3) the sequential mediation also produced attitudinal effects. The findings are discussed in light of their theoretical and practical implications for designing effective fact-checking messages to combat misinformation in non-Western languages.

An accurately informed public is a fundamental pillar of a robust democracy. However, the proliferation of misinformation—whether inadvertent or deliberately spread—threatens this foundation. Such misinformation can lead to suboptimal decisions both at the societal and individual levels, impacting critical domains from electoral processes to public health (Bennett & Livingston, 2018; Lewandowsky, Ecker, & Cook, 2017; van der Linden, 2022). Fact-checking is defined as “the practice of systematically publishing assessments of the validity of claims made by public officials and institutions with an explicit attempt to identify whether a claim is factual” (Walter, Cohen, Holbert, & Morag, 2020, p. 351), emerges as a crucial tool to counter these falsehoods.

However, merely stating what is true and what is false does not guarantee successful misinformation correction. A meta-analysis (Walter et al., 2020) shows that while fact-checking on average reduces misperceptions, the effect sizes are heterogeneous, suggesting that its efficacy depends on various moderating factors. Message-level factors can serve as such moderators. Although research has started to explore this area (e.g., Cappella, Maloney, Ophir, & Brennan, 2015), little is still known about the message features of fact-checking that enhance its effectiveness, especially in non-Western countries. Most previous studies were conducted in Western, English-speaking countries. As a result, “fact checkers across the world are forced to rely on findings that may not speak to their particular cultures” (Dias & Sippitt, 2020, p. 607). This study aimed to fill this gap by examining the role of linguistic politeness in journalistic fact-checking (i.e., news articles reporting on fact-checking; Graves & Amazeen, 2019) in South Korea. We conducted two online experiments with national convenience samples of South Korean adults to test how the use of honorific (vs. non-honorific) sentence endings—a well-established indicator of linguistic politeness in Korean—in fact-checking articles influences recipients’ belief accuracy and both perceived and actual article persuasiveness by affecting their perceived face threat and psychological reactance.

The Role of Linguistic Politeness in Fact-Checking

Fact-checks can be viewed as persuasive messages in that they aim to evaluate the veracity of public claims and convince recipients about which are true or false, thereby reducing their misperceptions and enhancing belief accuracy (Garrett & Weeks, 2013). People’s misperceptions are influenced by four factors (Garrett, 2017): (a) motivated reasoning, (b) non-motivated information-processing biases (e.g., accessibility), (c) social dynamics (e.g., influentials), and (d) the information environment (e.g., news). Focusing on psychological processes linked to motivated reasoning, and considering perspectives from research on source and message factors that can affect recipients’ motivated biases (e.g., depoliticization, source credibility, self-affirmation; Garrett, 2017; Hovland, Janis, & Kelley, 1953), we tested the role of linguistic politeness in enhancing the efficacy of fact-checking messages for misinformation correction.

Politeness and Persuasion

Attempts to persuade may lead to resistance. As Knowles and Linn (2004) put it, “[r]esistance is the tug-of-war partner with persuasion” (p. 8). Therefore, a key to successful persuasive messages, such as fact-checks, is to overcome this resistance (Garrett, 2017). Politeness theory (Brown & Levinson, 1987) is revealing in this regard as it explains why message recipients often resist persuasion (face threats) and suggests a messaging strategy to decrease their perceived face threats, thereby reducing resistance (using polite language; Jenkins & Dragojevic, 2013).

Politeness theory (Brown & Levinson, 1987) posits that during communication, individuals systematically use politeness strategies to support both their own and the recipients’ negative and positive face wants (the desire to be autonomous and the desire to be approved, respectively). In persuasion contexts, it is crucial to address recipients’ face needs, especially by mitigating potential face threats. Persuasive messages can pose potential face threats by steering recipients toward certain attitudes or behaviors (Jenkins & Dragojevic, 2013). Such messages may make recipients feel their freedom of choice is restricted (negative face threat), and/or that their ability to make the right decisions independently is questioned (positive face threat).

As persuasive messages, fact-checks can also be face-threatening (Garrett, 2017). Fact-checking messages that correct misinformation determine what is true or false, leaving little room for recipients to make their own judgments. Those who reject the messages risk being labeled as believing misinformation. Even those already skeptical of the misinformation may perceive these messages as questioning their ability to independently discern the truth.

When recipients perceive a persuasive message as face-threatening (e.g., a fact-checking news article), they might resist the persuasion attempt to protect their face, by rejecting the advocated position (e.g., the verdict of fact-checking; Garrett, 2017). For example, using forceful (i.e., impolite) language in messages promoting organic food consumption heightens perceived face threat, thereby increasing both source and message derogation, leading to less favorable attitudes toward organic foods and lower intentions to consume them (Jenkins & Dragojevic, 2013). This resistance can also be explained by psychological reactance, “the motivational state that is hypothesized to occur when a freedom is eliminated or threatened with elimination” (Brehm & Brehm, 1981, p. 37). A face-threatening persuasive message can induce reactance (Jenkins & Dragojevic, 2013), marked by anger and negative thoughts about the message and its source (Dillard & Shen, 2005). Those experiencing reactance seek to restore their threatened freedom and reject the message's position (Brehm & Brehm, 1981; Quick, Shen, & Dillard, 2013).

Then, how can persuasive messages be crafted to mitigate face threats to recipients and thereby reduce the likelihood of reactance (i.e., lowering motivated reasoning; Garrett, 2017)? Politeness theory and psychological reactance theory suggest similar strategies: using polite language (e.g., being indirect and tentative, offering options, and avoiding domineering language; Jenkins & Dragojevic, 2013; Quick et al., 2013). Prior research has centered on linguistic features that make persuasive messages “more impolite” (e.g., the use of forceful language; Jenkins & Dragojevic, 2013; see also Bode, Vraga, Tully, 2020; Kim & Masullo Chen, 2021). Yet, less is known about what constitutes “more polite” persuasive messages. This is crucial for fact-checkers because the politeness strategies outlined by politeness theory (Brown & Levinson, 1987) are not directly applicable to fact-checks. Making ambiguous or tentative conclusions for the sake of politeness is not feasible for fact-checks that seek clear verdicts based on transparent evidence (Graves, 2017). Offering recipients a freedom of choice conflicts with fact-checking’s primary objective: grounding everyone in shared facts, irrespective of differing opinions (Graves, 2017). To address this gap, we examined a well-established marker of linguistic politeness in East Asian languages: the use of honorifics.

The Use of Honorifics as a Marker of Linguistic Politeness in Korean

Politeness theory (Brown & Levinson, 1987) does not fully account for linguistic politeness in non-Western languages, especially those with honorifics like Korean and Japanese (see Byon, 2006; Ide, 1989; Matsumoto, 1989). In these languages, honorifics serve as a formal linguistic tool to express politeness. Specifically, in print news articles in Korean, the most direct and straightforward way to express linguistic politeness is through sentence endings (Byon, 2006). Formal forms of addressee honorification span four stages reflecting different levels of respect. For example, the sentence “He read a book” can be rendered in four distinct ways using different sentence endings: (a) Geuneun chaegeul ilgeutseupnida (the “hasipsio” form, which is extremely addressee-raising); (b) Geuneun chaegeul ilgeutso (the “hao” form, moderately addressee-raising); (c) Geuneun chaegeul ilgeutne (the “hage” form, moderately addressee-lowering); and (d) Geuneun chaegeul ilgeutda (the “haera” form, extremely addressee-lowering).

We compared the most polite sentence endings (the “hasipsio” form) with the least polite one (the “haera” form). Notably, the “haera” form is conventionally used in journalistic writing for both regular news articles and fact-checking articles, primarily because of its efficiency and clarity. Even though the “haera” form represents the baseline level of politeness in this context, it is still perceived as polite, rather than impolite. In sum, we predicted that fact-checking news articles using “more polite” language (honorific sentence endings) would be less likely to threaten recipients’ face and evoke psychological reactance than those using the conventional journalistic form, which, despite offering a baseline level of politeness, is still considered “polite” (non-honorific sentence endings). Consequently, the use of honorific endings should enhance belief accuracy and persuasion.

Study 1

In the context of exposure to fact-checking news articles correcting misinformation about COVID-19, Study 1 predicted that articles using honorific sentence endings would be perceived as less face-threatening than those using non-honorific ones (H1), and this perceived face threat would, in turn, decrease both belief accuracy (H2a) and perceived article persuasiveness (H2b).

Method

An online experiment was conducted with a one-way between-subjects design (linguistic politeness: honorific vs. non-honorific sentence endings). Participants were a national convenience sample of South Korean adults recruited via a survey company, Macromill Embrain. Out of the 251 participants who initially began the experiment, 10 dropped out, and one failed an attention-check item (“5 × 3 = ?”), leaving a final sample of 240 (50.8% female; Age M = 39.43, SD = 10.93; 74.2% BA degree holders). After reporting their gender and age, participants were randomly assigned to either the honorific or non-honorific condition and were presented with a fact-checking news article that corrected the misinformation suggesting that pets can be infected with COVID-19 and subsequently transmit it to humans. The non-honorific condition article employed sentence endings in the “haera” form, a typical writing style of Korean print news articles, whereas the honorific condition article used the “hasipsio” form, a more polite style of reader-honorific expression in Korean (Supplementary Appendix A). To ensure perceived authenticity and credibility, the article was crafted to appear as though it had been reported by a major newspaper, Hankook Ilbo, and published on Naver News, South Korea’s leading online news platform (Supplementary Appendix B). After reading the article, participants answered questions regarding perceived face threat, belief accuracy, perceived article persuasiveness, and education level.

All the posttest variables were measured on a seven-point scale (1 = strongly disagree, 7 = strongly agree). For perceived face threat, participants responded to two items: “The author of the article implies that I am not a very capable person,” “… is insulting” (Jenkins & Dragojevic, 2013; r = .81, M = 2.20, SD = 1.26). Belief accuracy was measured using a single item “Pets spread COVID-19 to humans” (reverse-scored; M = 5.65, SD = 1.51). For perceived article persuasiveness, participants indicated how strongly they agreed that the article was “persuasive,” “effective,” and “convincing” (Dillard, Shen, & Vail, 2007; α = .95, M = 4.38, SD = 1.38).

Results

Consistent with H1, linguistic politeness reduced perceived face threat (Table 1). Participants perceived fact-checking articles using honorific sentence endings as less face-threatening than those with non-honorific ones, b = −0.37, 95% CI [−0.69, −0.05], p = .024. This decrease in perceived face threat subsequently led to an increase in both (a) belief accuracy, b = −0.48, 95% CI [−0.63, −0.34], p < .001 and (b) perceived article persuasiveness, b = −0.25, 95% CI [−0.39, −0.12], p < .001, supporting H2a and H2b. A statistical mediation analysis with 5,000 bootstrap samples revealed that using honorific (vs. non-honorific) sentence endings increased belief accuracy and perceived article persuasiveness by mitigating perceived face threat, b = 0.18, 95% CI [0.04, 0.36], b = 0.09, 95% CI [0.01, 0.23], respectively.

Study 2

The results of Study 1 supported the hypothesis that linguistic politeness, as indicated by the use of honorific sentence endings in Korean, can enhance the efficacy of fact-checking news articles by reducing recipients’ perceived face threat. Study 2 was designed to replicate and extend Study 1 in the following ways. First, we examined whether the effects of linguistic politeness are contingent upon recipients’ cognitive proclivities shown to modulate message effects on belief accuracy (conspiratorial predisposition; Carey et al., 2022; RQ1a) and attitude change (need for cognition; Petty & Cacioppo, 1986; RQ1b), while acknowledging the existence of other potential moderators (e.g., deference to expert authority; Brossard & Nisbet, 2007; see the Discussion section). Second, drawing on research linking politeness theory to psychological reactance theory (Jenkins & Dragojevic, 2013; Quick et al., 2013; Zhang & Sapp, 2013), we tested whether perceived face threat would increase reactance (H3a), subsequently reducing belief accuracy (H3b), and perceived article persuasiveness (H3c). Lastly, we employed message stimuli addressing a more controversial issue (misinformation about refugees) and measured both actual (attitudes toward refugee acceptance; H3d) and perceived message effectiveness to more directly examine the persuasive impact of linguistic politeness in fact-checking messages.

Method

Using a one-way between-subjects design (linguistic politeness: honorific vs. non-honorific sentence endings), we conducted an online experiment with a national convenience sample of South Korean adults recruited via Macromill Embrain. From an initial pool of 326 participants, 28 dropped out and two failed an attention-check item (“5 × 7 = ?”), resulting in a final sample of 296 (48.6% female; Age M = 39.60, SD = 10.84; 72.0% BA degree holders). The procedure of Study 2 was the same as that of Study 1, except the following. First, Study 2 used a fact-checking news article that corrected two misperceptions: (a) accepting refugees in South Korea would result in a significant expenditure of taxes and (b) it would lead to increased crime rates (Supplementary Appendix A). Second, Study 2 measured conspiratorial predisposition and need for cognition as potential moderators in the pretest (Supplementary Appendix C).

All variables described below were measured on a seven-point scale (1 = strongly disagree, 7  =  strongly agree), unless otherwise noted. To more thoroughly measure perceived face threat, in addition to the two items used in Study 1, we incorporated the following two items: “The author of the article thinks he or she has the right to tell people what to do,” “… people are required to do what he or she says” (Cupach & Carson, 2002; Jenkins & Dragojevic, 2013; Johnson, Roloff, & Riffe, 2004; α = .66, M = 3.01, SD = .97). For psychological reactance (Dillard & Shen, 2005; Reynolds-Tylus, Bigsby, & Quick, 2021), we assessed anger (“While reading this article, I felt [irritated, angry, annoyed, aggravated]”; 1 = not at all, 7 = very much) and negative cognitions (“The thoughts I had about this article were [favorable(1)–unfavorable(7), positive–negative, good–bad])”: α = .89, M = 3.50, SD = 1.10. Belief accuracy was measured using two items: “If South Korea accepts refugees, [a vast amount of taxes would be spent, crime rates would increase]” (1 = definitely false, 6 = definitely true; reverse-scored; r = .65, M = 3.31, SD = 1.05). For perceived article persuasiveness, we used the same items as those in Study 1 (α = .94, M = 4.05, SD = 1.24). Attitude toward hosting refugees was measured using three bipolar scales ([disagreeing(1)–agreeing(7), unfavorable–favorable, negative–positive]: α = .96, M = 3.55, SD = 1.50. Conspiratorial predisposition was assessed using four items (e.g., “Much of our lives are being controlled by plots hatched in secret places”; Uscinski, Klofstad, & Atkinson, 2016; α = .84, M = 4.45, SD = 1.18). Need for cognition was measured using five items (e.g., “I would prefer complex to simple problems”; Petty & Cacioppo, 1986; α = .77, M = 3.76, SD = .97).

Results

As in Study 1, using honorific (vs. non-honorific) sentence endings in fact-checking articles reduced participants’ perceived face threat, b = −0.24, 95% CI [−0.45, −0.03], p = .028 (Table 2), supporting H1. This effect was not moderated by conspiratorial predisposition, b = −0.06, 95% CI [−0.24, 0.12], p = .538 (RQ1a), or need for cognition, b = −0.12, 95% CI [−0.34, 0.10], p = .275 (RQ1b). The reduction in perceived face threat subsequently decreased reactance, b = 0.48, 95% CI [0.36, 0.60], p < .001, thereby (a) enhancing belief accuracy, b = −0.37, 95% CI [−0.48, −0.27], p < .001; (b) increasing perceived article persuasiveness, b = −0.40, 95% CI [−0.54, −0.26], p < .001; and (c) fostering more favorable attitudes toward refugee acceptance, b = −0.49, 95% CI [−0.65, −0.32], p < .001. A sequential mediation analysis using 5,000 bootstrap samples showed that articles with honorific (vs. non-honorific) sentence endings enhanced belief accuracy, as well as both perceived and actual article persuasiveness, by leading participants to view the articles as less face-threatening and thereby to experience less reactance: b = 0.04, 95% CI [0.01, 0.10] for belief accuracy; b = 0.05, 95% CI [0.01, 0.11] for perceived article persuasiveness; b = 0.06, 95% CI [0.01, 0.13] for attitudes. Thus, H3a to H3d were supported.

Discussion

The results of the two experiments indicate that linguistic politeness enhances the efficacy of fact-checking news articles that correct misinformation by reducing perceived face threat and psychological reactance. Study 1 demonstrated that articles employing honorific sentence endings (vs. those with non-honorific ones conveying identical content) mitigated perceived face threat. This subsequently increased belief accuracy and perceived article persuasiveness.

Using news articles that fact-checked misinformation about a more controversial issue as stimuli, Study 2 replicated the face-threat-reducing effect of honorific sentence endings. It also revealed that linguistic politeness enhanced both the actual persuasiveness (i.e., inducing message-consistent attitudes) and perceived persuasiveness of the articles. This underscores the importance of assessing the attitudinal impacts of fact-checking, beyond its immediate effects on belief accuracy. Furthermore, Study 2 showed that the influence of perceived face threat on belief accuracy, as well as on both perceived and actual article effectiveness, was mediated by psychological reactance. This supports the proposition that politeness theory and psychological reactance theory together provide a comprehensive framework for the role of linguistic politeness in persuasion (Jenkins & Dragojevic, 2013; Quick et al., 2013; Zhang & Sapp, 2013).

Additionally, Study 2 found that using honorific sentence endings enhanced the efficacy of fact-checking articles, regardless of recipients’ conspiratorial predisposition or need for cognition. This finding contrasts with past studies where such cognitive proclivities acted as contingencies (e.g., the effects of fact-check exposure were greater among those with high conspiracy predispositions than among their low counterparts; Carey et al., 2022). This discrepancy suggests that the effects of linguistic politeness might operate through subtler pathways, distinct from those influenced by vulnerability to misinformation and motivation for elaboration. Furthermore, other individual characteristics, such as deference to scientific/expert authority and trust in fact-checkers, might also moderate these effects (Brossard & Nisbet, 2007; Jenkins & Dragojevic, 2013).

Relatedly, more research is warranted to investigate the role of linguistic impoliteness (or incivility; cf. Papacharissi, 2004)—as indicated by, for example, using insults, name-calling, and profanity (Bode et al., 2020; Kim & Masullo Chen, 2021). This will help identify how a broader range of linguistic politeness shapes the psychological mechanisms observed herein and other potential routes such as perceived source credibility (Hovland, Janis, & Kelley, 1953) and perceived depoliticization (Garrett, 2017).

In sum, future research should delve deeper into the moderated mediation mechanisms associated with linguistic politeness in the context of misinformation correction.

This study has some limitations worth noting. First, our experiments utilized a single message as a stimulus, making it uncertain whether the findings pertain specifically to the individual messages tested or are generalizable to the broader categories of messages we are theoretically interested in (i.e., fact-checking articles with honorific vs. non-honorific sentence endings). To address this potential case-category confounding, future research should employ a multiple-message design (Jackson, 1992; Yang, Maloney, Tan, & Capella, 2018), using multiple stimuli for each category that vary by message characteristics, including topical areas shown to moderate the effects of corrective messages (Walter & Murphy, 2018).

Second, our experiments, which compared fact-checking articles using honorific sentence endings with those using non-honorific ones, lacked a no-message control condition (e.g., Carnahan & Bergan, 2022). Consequently, it remains uncertain whether reading each type of article leads to higher or lower belief accuracy compared to not reading any articles at all. Additionally, we employed a one-way experimental design without manipulating other message features (e.g., exemplification; Zillmann, 2006), thus limiting our understanding of the robustness and potential boundary conditions of our findings. To more thoroughly evaluate the impact of linguistic politeness in fact-checking messages on misinformation correction, future studies will need to use a multifactorial design plus a separate no-message control condition.

Third, Study 2 did not measure prior attitudes toward refugee acceptance to prevent carryover effects. Thus, we could not examine whether the observed effects were contingent upon recipients’ prior attitudes. Future research should employ an experimental design that measures prior attitudes while minimizing carryover effects (e.g., setting a sufficient time interval between the pretest and posttest).

Fourth, with approximately 120 participants per condition in Study 1 and 150 in Study 2, our study may be underpowered. This raises concerns about the robustness and generalizability of our findings. Future research employing larger samples is suggested to validate our results.

Lastly, we did not conduct a manipulation check on the use of honorific (vs. non-honorific) sentence endings in fact-checking articles, as this is an intrinsic message feature independent of recipient perceptions (O’Keefe, 2003). However, it may be questioned whether this variation indeed influenced the perceived politeness of the articles as intended. Although our mediator, perceived face threat, can serve as an indicator of perceived (im)politeness (Jenkins & Dragojevic, 2013), directly measuring perceived politeness in future research could further clarify the psychological processes observed herein.

Despite the noted limitations, this study advances research on fact-checking by identifying a linguistic feature—using honorific sentence endings in Korean—that enhances the efficacy of fact-checking messages by reducing recipients’ face threat and psychological reactance.

Supplementary Data

Supplementary data are available at the International Journal of Public Opinion Research online.

Acknowledgements

This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2021S1A5B8096358).

Conflict of Interest

None declared.

Biographical Notes

Hogeun Lee (MA, Seoul National University) is a research associate at the Institute of Communication Research at Seoul National University. His research focuses on understanding the cognitive mechanisms underlying message effects on persuasion and misinformation correction.

Hyun Suk Kim (PhD, University of Pennsylvania) is an associate professor in the Department of Communication at Seoul National University. His research centers on message effects in viral and persuasive communication; current projects focus on identifying message features that enhance the effectiveness of fact-checking.

References