“Others Are Misled, Not Me”: Third-Person Perception and Misinformation Sharing Among Chinese Elderly on WeChat

2.1 The Third-Person Perception Theory

The third-person perception hypothesis was first introduced by sociologist Philip Davison in 1983. The hypothesis predicts that individuals tend to believe others are more easily influenced by mass media messages than themselves (Davison 1983). This hypothesis has been supported by numerous studies in different contexts. For instance, research on the behavioral component of the third-person effect on the preparations for Y2K (the year 2000) confirmed the idea that people believe others would overprepare for Y2K (Tewksbury, Moy, and Weis 2004). Pham, Shancer, and Nelson applied both qualitative and quantitative research methods and found third-person perception regarding undesirable behaviors on Facebook among user perceptions (Pham, Shancer, and Nelson 2019). There have been some studies in the Asian context as well. For example, Lv and Wu, drawing on the research design of P. C. Meirick and examined whether third-person perception exists in Chinese youth’s perception of cigarette advertisements. Moreover, within the Chinese cultural context, the “anti-first-person effect” was identified (Lv and Wu 2021).

Over the years, third-person perception has been supported in various research topics, such as advertising, political issues, news, and pornography. Generally, there are three main research domains: the entertainment domain (e.g., media violence, pornography, and media use (e.g. Tsay-Vogel 2016; Pham, Shancer, and Nelson 2019; Caron and Light 2016; Lev-On 2017); the persuasion domain (e.g., the effect of advertising), and the news domain (e.g., avian flu news and fake news (Ham and Nelson 2016; Henriksen and Flora 1999; Jang and Kim 2018; Ștefăniță, Corbu, and Buturoiu 2018; Wei, Lo, and Lu 2008). With the development of social media, researchers began to explore whether TPP exists within this context. Especially since 2015, the key research terms have been “social media,” “Facebook” and “network” and other words related to new media (Fu, Chen, and Yu 2020). For example, Nicoleta et al. studied the impact of TPP among Facebook young users (Corbu, Ştefǎnițǎ, and Buturoiu 2017). Research on comments in social interactions also confirmed the existence of TPP. It is found that people believed that uncivil comments had a greater negative impact on others than on themselves (Chen and Ng 2017). At the same time, the research scope has expanded further. For instance, Stavrositu and Kim (2014) analyzed the third-person effect of online health news, adding a new dimension to the traditional third-person effect research by examining social media indicators, such as shares and comments (Stavrositu and Kim 2014).

Although there have been many studies on social media and third-person effects, only a few studies focus on the third-person effect in the Chinese context. He Hongyan and Han Hong investigated WeChat rumors during the COVID-19 epidemic, exploring the manifestations and unique characteristics of the “third-person effect” on social media, and found a significant third-person effect in the spread of rumors on WeChat (He and Han 2021).

The third-person perception can be generally divided into two components: perceptual level and behavioral level. Studies on this hypothesis have been extended from the perceptual component to the behavioral component, including attitudes, behavioral intentions, and actual behaviors. The perceived impact disparity between self and others may lead to some behavior, such as media restrictions (Gunther 1995), support for censorship, and parental mediation (Hoffner and Buchanan 2002). Pham, Shancer and Nelson’s study confirmed TPP regarding undesirable behaviors on Facebook (Pham, Shancer, and Nelson 2019). Likewise, misinformation sharing behavior, a form of undesirable behavior, leads to the following hypothesis:

As for the possible reasons why TPP exists, many possible explanations have been proposed by numerous researchers by applying different theories. For example, Henriksen and Flora attributed third person perception to a superiority bias, the tendency to regard oneself as better or better off than others (Henriksen and Flora 1999). Further, when theorizing TPP and behavioral consequences such as support for censorship or regulations, paternalism is often suggested as a reason for why individuals with TPP would support censorship for others. However, Golan and Banning argued that paternalism only explains TPP for negative messages but lack of effectiveness when it comes to positive messages (Golan and Banning 2008). They suggested explaining the behavioral aspect of TPP with Fishbein and Ajzen’s theory of reasoned action, which posits that people are trying to provide a favorable outcome to meet the social expectations of others (Fishbein and Ajzen 1975). Similarly, Sun, Shen, and Pan (2008) suggested that such behavior aims to rectify problematic situations. In a comparative analysis of TPP between Asia and the rest of the world, it is found that the most commonly applied theories are social comparison theory and social identity theory (Lo et al. 2016). However, the explanations are broad, and no universally accepted explanation has been achieved.

2.2 TPP in the Post-Truth Era

With the emergence of the “post-truth” era and the development of social media, some studies have shifted their focus to TPP in relation to fake news and misinformation. For instance, a study conducted in Romania confirmed the existence of TPP, finding that individuals perceived the impact of fake news as stronger on others than on themselves (Ștefăniță, Corbu, and Buturoiu 2018). Another study examined TPP in fake news and found that a higher TPP inclination led individuals to support media literacy (i.e., the ability to access, analyze, evaluate, and create media) over media regulation as a way to combat fake news online (Jang and Kim 2018). In earlier studies, social distance, perceived knowledge, and media exposure have been the main factors examined in TPP studies (Conners 2005). However, with the proliferation of social media and the empowerment of information producers, critical thinking factors, such as media trust, fact-checking habits, and misinformation verification ability have become essential in the post-truth era.

Previous studies have shown that the perceived credibility of media content can mitigate TPP. For instance, Wei, Lo, and Lu’s research on TPP in the context of tainted food product recall news found that higher perceived message credibility reduces the perceived effect gap. With higher perceived credibility of the messages, the perceived effect increases on both oneself and others, but the gap between the two narrows (Wei, Lo, and Lu 2008). In a study on WeChat rumors and the third person effect, it was found that perceived credibility significantly impacts message-forwarding behavior and support for regulation. However, whether credibility influences the third-person effect itself remains unclear (He and Han 2021). Another study on TPP in social media sites found that lower perceived information credibility led individuals to believe they were less likely to be influenced while perceiving others as more susceptible (Li et al. 2021). Furthermore, the perceived effect on oneself promotes information-seeking behavior and protective actions (Wei, Lo, and Lu 2008). Therefore, it can be hypothesized that individuals with lower trust in a specific media source may seek information from alternative sources. Additionally, individuals with better fact-checking habits might exhibit higher levels of critical thinking and a stronger ability to verify information. Hence, based on the research questions and literature review, the following hypotheses are proposed:

Moreover, demographic factors are important, as individual differences may impact TPP on social media (Tsay-Vogel 2016). Among those demographics, gender differences have been widely discussed. Previous studies have identified consistent patterns in gender-related differences concerning the third-person perception effect. One typical example is research on pornography, where gender has consistently been identified as a key factor in the strength of the TPP effect (Gunther 1995; Lo and Wei 2002). It was found that females tend to perceive greater negative effects of pornography on males than on females (Lo and Wei 2002). Similarly, studies on TPP in the context of online gaming have shown that the effect is stronger among females than males (Zhang 2013). Thus, it is essential to examine if the third person perception is stronger among males or females regarding misinformation sharing behavior on WeChat. Given past findings that females are more likely to perceive others as influenced by negative media messages, we propose the following hypothesis:

Additionally, examining potential gender differences in both self-reported and objectively measured misinformation verification abilities is valuable. We aim to explore whether traditional gender stereotypes——suggesting that males are generally more independent and rational, while females are considered more dependent and emotional (Bardwick and Douvan 1971; Maccoby and Jacklin 1974)—are reflected in individuals’ self-reported and objectively measured abilities. Based on this, we propose the following hypotheses:

Figure 1:

Hypothesized research model for third-person perception (TPP) and misinformation sharing behavior on WeChat.

3.1 Research Design

A questionnaire in Chinese and English was designed for research on the use and misinformation on WeChat, and the present research is part of this project. This questionnaire was approved by three experts from both China and Spain. Given the complexity of Chinese society and the significant urban–rural divide, this research focuses on urban China only.

Considering that the targeted population is older adults, the standard 5-point Likert scale may be difficult for them to understand. Therefore, we adapted the 5-point scales to be more understandable. For example, instead of asking “my peers often share rumors” with answers ranging from “strongly disagree” to “strongly agree,” we adapted this question to rate the frequency of sharing rumors from peer, reducing the difficulty in understanding the questions.

Before conducting the full-scale survey, a pilot test with 30 respondents was first carried out. Based on the results of the pilot test, some minor modifications were made. The full-scale survey lasted nearly one month.

A total of 345 responses were collected. Since we limited the scope to Chinese people aged 45 and above (including 45) who live in urban area, responses from individuals aged under 45 and people not living in urban area were excluded. After removing ineligible responses, the number of valid questionnaire replies was 317, which met the expected number (300 responses).

3.2 Participants

A sample of 317 participants (N = 317, 153 females and 164 males) aged 45–74 (M = 52.46, SD = 5.06) participated in this research. Most participants (87.7 %) had no prior experience in media-related work, and the average monthly income was 6,848 RMB. The sample was drawn from 28 provinces, autonomous regions, and municipalities in China (Figure 2), with only urban samples retained, excluding rural samples.

Figure 2:

Geographic and distribution of participants.

3.3 Measures

4.1 Third-Person Perception (TPP)

The first hypothesis predicted the existence of TPP. H1 proposed that third person perception would be observed in misinformation sharing behavior, such that others are perceived as performing this undesirable behavior more frequently than oneself on WeChat. To investigate whether TPP exists in misinformation sharing behavior, a paired t-test analysis was conducted. Results show that the scores for participants’ self-perceived misinformation-sharing behavior (M = 1.89, SD = 0.95) were significantly lower than those for perceived misinformation-sharing behavior of their peers (M = 2.81, SD = 0.90, p < .01). Thus, the third-person effect was observed in misinformation-sharing behavior, and H1 is supported.

4.2 TPP and Fact-Checking Habits

H2 focuses on third-person perception and fact-checking habits. To test whether the third-person perception effect is stronger among individuals with active fact-checking habits, an ANOVA was conducted. Results show significant differences between groups, so post hoc tests were conducted to identify whether each group differs significantly from the others. The results show that the TPP effect among individuals who usually do not search for information in other media (M = 0.67, SD = 1.04) is significantly larger than among those who usually look for information from different sources and compare them (M = 1.26, SD = 1.26), F (2, 314), p < .01 (see Table 1). The effect among individuals who typically seek information from other media they consider reliable is medium (M = 0.88, SD = 1.27) in between the other two groups (see Figure 3). Thus, the third-person perception effect is stronger among those with better fact-checking habits. H2 is supported.

Figure 3:

Mean TPP among individuals with different fact-checking habits.

Since TPP is measured by two aspects, further analysis was conducted to investigate whether the gap of TPP primarily arises from participants’ evaluations of others or themselves. As shown in the figure (Figure 4), individuals with better fact-checking habits are more likely to report a lower frequency of their own misinformation sharing behavior, F (2, 314) = 4.04, p < .05. However, in terms of the evaluation of others, there is no statistically significant difference between groups with varying levels of fact-checking habits (p > .05). Nevertheless, there is a sight trend suggesting that individuals with better fact-checking habits tend to rate the frequency of misinformation sharing by peers slightly higher (Table 1).

Figure 4:

Mean estimates of misinformation sharing behavior for self and others by fact-checking habits.

4.3 TPP and Trust Degrees in Information on WeChat

H3 proposed that the third person perception effect is stronger among individuals with lower levels of trust in information on WeChat. After comparing the mean value, we observed a clear decreasing trend as the trust level in information on WeChat increased (Figure 5). Results indicate that the TPP effect is stronger among individuals with lower trust in the information on WeChat. Thus, H3 is supported.

Figure 5:

Average TPP among individuals with different levels of trust in information on WeChat.

Following the same approach used to analyze TPP and fact-checking habits, we conducted further analysis on the effect of varying trust levels in information on WeChat. As shown in Figure 6, individuals with lower trust in WeChat information, tend to rate others as having a higher likelihood of sharing misinformation on WeChat while rating themselves as having a lower likelihood. In other words, the perception gap arises from the perception of others, and the self-perception of the participants.

Figure 6:

Reliability of information on WeChat and mean estimates of misinformation sharing behavior for self and others.

4.4 TPP and Information Verification Ability

H4a examines the relationship between TPP and self-reported information verification ability, while H4b focuses on TPP, and information verification ability assessed through a set of tests. As shown in Table 2, there are no significant trends or relationships between the strength of TPP and self-reported verification ability. Across different levels of self-reported verification ability, the mean value of third-person perception fluctuates irregularly. Therefore, H4a cannot be supported here.

However, in the test of information verification ability, as the mean value of third-person perception increases, the test score also increases (Figure 7), and the differences are significant, F (3, 113) = 8.19, p < .05 (see Table 3). This result indicates that the third-person perception effect is stronger among individuals with higher information verification ability. Thus, H4b is supported.

Figure 7:

Mean TPP among individuals with different misinformation verification test scores.

Unlike the findings above, as the test score increases, the mean value of the perceived frequency of misinformation sharing behavior by others also rises. However, as shown in Figure 8, the evaluation of their own frequency of sharing uncertain information remains relatively stable. In other words, individuals with higher test scores, which indicates have higher misinformation verification ability in this study, tend to believe that there is a higher possibility of others to share misinformation on WeChat, while their self-evaluation remains consistent across different levels of misinformation verification ability. Therefore, this perception gap may primarily result from their evaluation of others rather than from both sides.

Figure 8:

Misinformation verification ability and mean estimates of misinformation sharing behavior for self and others.

4.5 TPP and Gender

H5 proposed that the TPP effect would be stronger among females than males. To investigate whether any gender differences exist, a t-test analysis was conducted. Results show no significant gender differences in TPP related to misinformation-sharing behavior (p = 0.17), with the mean TPP value for males (M = 1.02, SD = 1.15) slightly higher than that for females (M = 0.83, SD = 1.26) (Table 4). Thus, H5 was rejected.

Since TPP consists of two components, we further explored whether any gender differences exist in self-reported frequency of sharing information without verification or in the perceived frequency of misinformation-sharing by peers.

Firstly, we tested for gender differences in the perceived frequency of sharing misinformation behavior from peers by running a t-test. Results show no significant differences in mean values between males (M = 2.76, SD = 0.87) and females (M = 2.87, SD = 0.93) on this item (p > .05). Then we conducted another t-test to examine gender differences in self-reported misinformation-sharing behavior on WeChat. A significant gender difference was observed in this item: females’ self-reported frequency of sharing information without verification (M = 2.04, SD = 1.01) is higher than that of males (M = 1.74, SD = 0.87, p < .01).

4.6 Gender and Misinformation Verification Ability

H6a examines the relationship between gender and self-reported information verification ability, while H6b focuses on gender and information verification ability based on a set of tests.

Firstly, we compared the mean values of self-reported verification ability between genders using a t-test. Results show that males rated their self-reported verification ability slightly higher (M = 3.56, SD = 0.82) than females (M = 3.46, SD = 0.71), but the difference is not significant (p = 0.24). Therefore, this result cannot fully support H6a.

Next, from the crosstabulation of gender and tests-based misinformation verification ability, we found that 22.6 % of male participants answered all the test questions correctly, which is higher than the rate of females (20.3 %). Additionally, 50.6 % of male participants answered 2/3 of the test questions correctly, and this rate is much higher than that of females, too (45.8 %). Thus, H6b can be supported.