Sage Journals: Discover world-class research

Abstract

Many individuals do not seek news, believing instead that “news will find me” (NFM), implying that they trust their social networks to keep them informed, saving them the trouble of proactively seeking news from journalistic outlets. Does this mean that they trust social media algorithms to accurately filter and recommend content that is relevant to them? Do they trust their friends to keep them informed, just like they would journalists? To answer these questions, we conducted a pre-registered between-subjects experiment (N = 244) in which users with varying levels of NFM were randomly assigned to receive news recommended by either their social media friends, news editors, or an algorithm. We discovered that while users tend to act on news recommended by an algorithm mindlessly before reading it first, the type of cognitive heuristic triggered by a news source plays an important role in shaping their trust. Specifically, individuals high on NFM tend to trust algorithms because they trigger the “machine heuristic.” They also consider social media friends and algorithms to be as authoritative as news journalists and editors (“authority heuristic”). Our results advance theoretical knowledge about why high levels of NFM predict higher trust in social media friends and algorithms.

Keywords

news source News Finds Me perception algorithm machine heuristic authority heuristic

Today’s social media environment, where entertainment and news constantly mingle, has fueled the News Finds Me Perception (NFM), which refers to the belief among individuals that they will indirectly stay informed about public affairs without actively following the news, because they believe that they will be kept informed by their online social networks (Gil de Zúñiga et al., 2017). Studies using panel surveys and time series data have shown that NFM predicts lower levels of political learning (Lee, 2020), higher political cynicism (Song et al., 2020), and the spread of politically homogeneous opinions (Gil de Zúñiga & Cheng, 2021).

It appears that high-NFM individuals are more likely to rely on their friends or social media algorithms to receive news (Gil de Zúñiga & Cheng, 2021). Yet, limited empirical research has directly and experimentally tested the underlying mechanism of why high-NFM users trust these sources in finding news for them. Research has long documented that people tend to rely on the source of information as the proximal cue when evaluating online news credibility in an expedient manner (Sundar et al., 2007), in part due to cognitive overload. The reason behind this tendency is that the source cues trigger different cognitive heuristics (or mental rules of thumb) among individuals, such as the authority heuristic triggered when the source is a professional authority like a news editor (i.e. trained experts are more trustworthy), the bandwagon heuristic triggered when the news is recommended by other users (i.e. if others think that something is good, then I should, too), and the machine heuristic triggered when the news is recommended by an algorithm (i.e. machines are more objective and accurate than humans; Sundar, 2008). In addition to shaping perceptions of the news, these heuristics predict how online users consume information, including their tendency to verify the content before sharing it (Sun & Xie, 2024). They can also explain the widespread phenomenon of users acting on the news (e.g. forwarding, commenting on, or liking news) without reading the content. This is because their actions are based on peripheral cues, such as headlines, source signals, or social endorsement metrics, rather than engaging in effortful, central processing of the news content (Sundar et al., 2025).

Will high-NFM individuals be more likely to follow such heuristics when evaluating and consuming news recommended by different sources compared to low-NFM individuals? Although research has not directly tested whether high NFM predicts more shallow or heuristic processing of information, extant research seems to imply this possibility. For instance, studies have found that high-NFM individuals are more likely to rate misinformation as credible (Diehl & Lee, 2022) and tend to trust the news recommended by their social peers more (Gil de Zúñiga et al., 2022). These studies suggest that high-NFM users may be less likely to scrutinize the actual content of the news, relying instead on superficial heuristic cues to form credibility assessments and sharing decisions, compared to low-NFM users. Previous studies examining the detrimental effects of NFM have primarily relied on longitudinal survey data, in which participants report their general political knowledge of ongoing public affairs (e.g. Gil de Zúñiga et al., 2017; Song et al., 2020). However, there is a lack of research examining how individuals process or assess the credibility of different news recommendations in an experimental setting. This study, therefore, addresses the following questions: Do users evaluate news differently when it is recommended by different sources? And more importantly, will NFM predispose social media news consumers to trust social media friends and algorithms more than professional news editors? And if so, why?

To answer these questions, we conducted a pre-registered experiment by randomly assigning users with various levels of NFM perceptions to receive news recommended by different sources (i.e. algorithms vs. social media friends vs. news editors). Our results provide practical implications to combat the negative effects of NFM and mitigate high-NFM users’ overreliance on news of dubious provenance. Our study can shed light on how we can help high-NFM users become more discerning when they encounter news on their social media feeds.

Literature Review

Effects of News Recommendation Sources

Online users are overloaded with information given the vast amount of news feeds they are exposed to on a constant basis on their mobile devices and computers. To cope with overload, they tend to rely on sources as cues for evaluating the credibility of information they receive (Sundar et al., 2007). However, the concept of “source” in today’s news environment is quite murky and complex, with multiple layers (Sundar & Nass, 2001), including but not limited to news editors, social network friends, and, increasingly, news algorithms. Therefore, one question that arises is how these different sources affect users’ perceived news credibility. Understanding this question is of great importance, as source information plays a crucial role in their evaluation of the news recommendations, including perceived news credibility (i.e. the degree to which is news is considered accurate, authentic, and believable) (Appelman & Sundar, 2016), perceived bias (whether the news is objective and bias-free) (Waddell, 2019), perceived newsworthiness (whether the news is considered important, timely and interesting) (Xu, 2013), and the credibility of the system that recommends the news (Liao, 2023), with different sources triggering different cognitive heuristics or mental shortcuts (Sundar, 2008). Beyond credibility assessments, source cues may also shape news consumption behaviors. One such behavior is users’ intention to verify the news they encounter online. A recent meta-analysis showed that users’ tendencies to verify false news, or to share news without verification, are largely driven by cognitive heuristics, such as the social endorsement heuristic (Sun & Xie, 2024). In addition, there is increasing evidence to suggest that, beyond a lack of verification behaviors, many users directly forward or share information without first engaging with its content. For example, prior research found that more than half of shared news URLs from major news outlets were never clicked (Holmström et al., 2019). More recent evidence further shows that 75% of forwarded links in public Facebook posts containing URLs were shared without being clicked first (Sundar et al., 2025). The reason behind this phenomenon could be that users make snap judgments to like, comment, or share the news by relying solely on the news title, or contextual cues about the provenance of the news (i.e. friends, etc.), or just the interface instead of the actual news content. In this study, we would like to explore whether source cues contribute to this phenomenon.

Decades of research have been dedicated to understanding the effects of news sources, be they original sources or gatekeepers, that is, “selecting sources” (Sundar & Nass, 2001). Traditionally, news consumers receive news directly from news organizations and editors. These news organizations have established their reputation in the news business with strict gatekeeping standards. Therefore, news editors, copy editors, and proofreaders are usually perceived as the authorities in the news industry because they play a critical role in ensuring that news stories are well-written, accurate, and credible (Swasy et al., 2015). By providing these important editorial services, journalists help maintain the quality and integrity of the news content. National and local news organizations have gained a reputation for being a reliable and credible sources of news, especially in this era of misinformation (Allcott & Gentzkow, 2017). Therefore, when a news article presents itself as recommended by a news editor or a legitimate news organization, it is likely to activate the authority heuristic (Sundar, 2008), or a mental shortcut that news editors are trained experts and hence more authoritative in recommending news. In addition to affecting users’ evaluation of news credibility, authority cues can influence behavioral intentions, such as sharing (Weismueller et al., 2022). Therefore, we propose that the authority heuristic will serve as a mediator that explains the effects of news editors on both evaluative and behavioral outcomes:

H1a: Compared to receiving news recommended by an algorithm or social media friends, receiving news from editors is more likely to trigger the authority heuristic.

H1b: Authority heuristic will further predict more positive evaluations of the recommended news and less verification intention and behaviors among users.

Aside from relying on recommendations from news experts, online news readers also follow the wisdom of crowds (Surowiecki, 2005). More users have started to rely on their online social networks to keep themselves informed about important news. The same news shared by Facebook friends could generate more trust than directly receiving it from traditional news outlets (Turcotte et al., 2015). A study by Sundar and Nass (2001) found that news consumers preferred the news selected by “other users” over news selected by a news editor or a computer, even when the selected news articles were identical. In other words, simply seeing the source cue indicating that other users selected the news could make the news appear more appealing to news consumers. It is likely that the attribution of other users as the news source triggers the bandwagon heuristic, or the mental shortcut that “if others think something is good, I should think so, too” (Sundar, 2008). When a news feed had a high number of votes from other users, users perceived the news as more trustworthy and were more likely to share the news due to higher bandwagon perception (Xu, 2013).

Moreover, social media help users connect with others with similar interests or socioeconomic status. Users are more likely to read and share the news shared by others that match their own preferences due to a higher sense of perceived homophily (Kim & Ihm, 2020). The greater the perception of similarity between two people (Rocca & McCroskey, 1999), the higher the likelihood that they will perceive the information shared by each other as credible (Wang et al., 2008). Therefore, in addition to the bandwagon heuristic, it is possible that users might also be more likely to trust the news shared by their social media friends due to the triggering of the homophily heuristic. Liao (2023) found that if the news was recommended by similar others (collaborative filtering), it could trigger the homophily heuristic, or the mental shortcut that if others similar to me trust something, I should trust it as well, which was associated with higher perceived credibility of not only the news recommender system but also that of the news content. Homophily perceptions could also predict more news sharing behaviors (Halberstam & Knight, 2016). Based on these rationales, we propose the following two hypotheses:

H2a: Compared to receiving news recommended by the news editors or an algorithm, receiving news from social media friends is more likely to trigger the bandwagon heuristic.

H2b: Bandwagon heuristic will further predict more positive evaluations of the recommended news and less verification intention and behaviors among users.

H3a: Compared to receiving news recommended by the news editors or an algorithm, receiving news from social media friends is more likely to trigger the homophily heuristic.

H3b: Homophily heuristic will further predict more positive evaluations of the recommended news and less verification intention and behaviors among users.

Nowadays, social media platforms have also automated the process of news selection, recommending news based on different algorithms. Journalism is no longer exclusively controlled by human journalists, as AI systems today “perform, task solve, communicate, interact, and act logically as it occurs with biological humans” (Gil de Zúñiga et al., 2023, p. 4). Instead of following the traditional editorial news selection process, algorithmic news curation operates based on a pre-determined set of rules to prioritize certain factors, such as prior user engagement and user preferences. It can be applied consistently across all news stories (DeVito et al., 2017). These data-driven metrics can be perceived as more objective and accurate than human editors’ judgments, which are usually perceived as subjective decisions that could be influenced by personal biases or preferences (Gans, 2004). Thus, interface cues which inform readers that the news is selected and recommended by algorithms may trigger the “machine heuristic,” or the mental shortcut that machines are more objective, accurate and reliable than humans (Sundar, 2008), leading to an expedited positive judgment of information credibility.

Previous studies have demonstrated that users tend to assess the news more favorably, as being less biased and more trustworthy when a machine, rather than human, is the attributed written source of news (Graefe et al., 2018; Waddell, 2019). Although limited research has examined the effects of algorithms as the source of news recommendations compared to other sources, it is evident that online news readers are more likely to share stories when they believe the stories are accurate and objective (Altay et al., 2022). Feezell et al. (2021) revealed that getting news from algorithm-driven platforms facilitates online political participation when compared to receiving news from non-algorithmic platforms. Based on this theoretical rationale and empirical evidence, we propose machine heuristic to be a significant mediator that explains the effects of an algorithm as the recommendation source on users’ evaluations and news consumption behaviors. More formally:

H4a: Compared to receiving news recommended by news editors or social media friends, receiving news from an algorithm is more likely to trigger the machine heuristic.

H4b: Machine heuristic will further predict more positive evaluations of the recommended news and less verification intention and behaviors among users.

However, not all news consumers react to source cues in a similar manner. According to the Heuristic-Systematic Model (HSM), the likelihood for individuals to deploy the specific heuristic depends on whether this rule of thumb is accessible in their memory (i.e. the accessibility principle) as well as their cognitive abilities (Chaiken, 1987). Therefore, understanding how users with different characteristics react to these source cues differently becomes important. In this study, we aim to explore how an increasingly popular belief, News Finds Me perception, affects users’ reactions to different source cues.

Effects of News Finds Me Perception

First proposed by Gil de Zúñiga and colleagues (2017), News Finds Me Perception (NFM) refers to the “extent to which individuals believe they can indirectly stay informed about public affairs—despite not actively following the news” (Gil de Zúñiga et al., 2017, p. 3). NFM comprises three conceptual dimensions—being informed, peer reliance, and not actively seeking news (Gil de Zúñiga et al., 2017). Later, Gil de Zúñiga and Cheng (2021) proposed a fourth attribute, algorithmic reliance, motivated by evolving media structures and individuals’ inclination to accept algorithmic news gatekeeping. Gil de Zúñiga et al. (2020) reported that 49% of online users exhibit high-NFM perception, especially among individuals who prefer to receive news on social media sites. Such widespread prevalence can be attributed to the emergence of a high-choice media environment (Van Aelst et al., 2017), which provides news audiences with myriad options for getting their news across social media and other online platforms. Moreover, this media context fuels individuals’ belief that they can be informed about public affairs or issues of national interest without deliberately seeking news but deferring to their social media networks for their news information needs (Gil de Zúñiga et al., 2020).

A strong perception of NFM appears to have detrimental effects on traditional news consumption by audiences and users’ expectations of journalistic excellence (Park & Kaye, 2020). High-NFM users are receptive to news but less willing to devote their time, effort, and resources to seek it actively. NFM is said to enhance the “illusion of knowledge” and the notion of being network-dependent on accessing news because the relevant information will be channeled toward users regardless of explicit effort (Strauß et al., 2021). This implies that NFM involves passivity and low cognitive effort in obtaining learning gratification from social media news use (Diehl & Lee, 2022). Thus, it is not surprising that several studies have shown that NFM perception negatively relates to political knowledge and learning (Lee, 2020; Oeldorf-Hirsch & Srinivasan, 2022), quite contradictory to the assumption that social media provides “ambient awareness” among users (Hermida, 2010) and plenty of opportunities for individuals to seek and receive political news (Fletcher & Nielsen, 2019). Prior research seeking to explain why NFM has detrimental effects has primarily focused on the selective exposure hypothesis. Specifically, because NFM creates a false sense of being informed, individuals high in NFM perceive little need to actively seek out news, making them less likely to intentionally expose themselves to news content in the first place (Strömbäck & Shehata, 2018). For example, Skurka et al. (2023) found that individuals high in NFM are more likely to selectively expose themselves to soft news (e.g. entertainment and sports) rather than hard news (e.g. politics and science). Similarly, Zhang and Jiang (2024) showed, using cross-sectional survey data, that perceptions of NFM were positively associated with misperceptions about COVID-19, due to greater information avoidance. Lin et al. (2024) further demonstrated a negative relationship between general NFM perceptions and COVID-19 knowledge, which was mediated by reduced information seeking on social media.

Beyond the selective exposure hypothesis, more recent work suggests an additional mechanism: individuals high in NFM may also lack the motivation and cognitive capacity to deeply process news content (Skurka et al., 2025). Under this view, NFM reflects a low-effort cognitive style characterized by reliance on peripheral or heuristic processing and reduced motivation to scrutinize information carefully (Su et al., 2024). Therefore, although high-NFM users might increase their news consumption via social media, such consumption is only superficial and shallow, which does not contribute to actual knowledge gain. In other words, high-NFM users might be more likely to engage in heuristic rather than systematic processing of news they see online.

Although this possibility has not been directly tested, prior empirical evidence provides indirect support for the notion that individuals high in NFM are less likely to scrutinize news content than those with lower levels of NFM. Diehl and Lee (2022) tested the possibility that NFM was, in fact, a manifestation of a low-effort cognitive style of attention to news and public affairs, as high-NFM individuals might make little or no effort to scrutinize the accuracy of news content received from their social media peers. Results revealed that high-NFM individuals, who consume only fragments of news on social media, were more vulnerable to misinformation or fake news regardless of political valence (Diehl & Lee, 2022). Skurka et al. (2025) reported that when asked to identify false information, high-NFM individuals not only performed worse than low-NFM individuals but also substantially overestimated their performance on the task. This vulnerability may stem from the fact that high-NFM individuals are less likely to pay attention to or elaborate on factual stories. However, another possibility is that high-NFM users lack the motivation or ability to process information carefully and instead rely heavily on contextual cues like the source of news to make rapid credibility judgments. Supporting this interpretation, prior research has demonstrated that cognitive elaboration is negatively associated with NFM (e.g. Strauß et al., 2021). That is, individuals high in NFM are less likely to engage in effortful thinking, systematically process information, or integrate new information with existing knowledge when consuming news. Consistent with this pattern, Mosallaei et al. (2025) found that individuals who enjoy critical and analytical information processing (i.e. those with higher need for cognition) are less likely to endorse the belief that news will find them or to rely on social media feeds as a primary news source. Similarly, Tian and Willnat (2025) showed that high-NFM individuals invest less cognitive effort in evaluating the veracity of news content compared to their low-NFM counterparts. This latter mechanism is theoretically grounded in the notion that high-NFM individuals’ false sense of being sufficiently informed (i.e. perceived information sufficiency) may undermine their motivation to engage in effortful cognitive processing of political information, relative to low-NFM individuals (Petty & Cacioppo, 1986; Skurka et al., 2025).

Although past research has not directly tested how users with different levels of NFM react to news recommended by different sources differently, it has revealed that high NFM is usually tied with a decline in use of traditional news outlets because people increasingly depend on the always-on mode of social media for up-to-date information (Park & Kaye, 2020). In addition, users with high NFM are found to be more likely to fall for the availability bias and rely on the availability heuristic when evaluating their level of exposure to news on a daily basis (Hermida, 2010; Strauß et al., 2021).

In sum, the presence of social media friends and algorithmic systems as news sources is more likely to activate specific cognitive heuristics among high-NFM individuals, which may help explain why they place greater trust in recommendations from social media friends and algorithms than do low-NFM individuals. Song et al. (2020) revealed that higher NFM is tied with higher political cynicism, or a general distrust toward authoritative institutions like the government and a democratic political system. Given their distrust of institutions, including news organizations, they might be more likely to trust their friends to receive news. Research has revealed that when users rely on news shared by their social media peers, they tend to trust such information and are possibly more influenced by their exposure, even if they are counter-attitudinal to their political beliefs and interest (Messing & Westwood, 2014). This finding suggests that although the presence of social media friends as the source of news recommendations might trigger social endorsement heuristics (i.e. bandwagon heuristic and homophily heuristic), compared to news editors and algorithms, this tendency may be particularly pronounced among those with high levels of NFM perception than among those with lower NFM, constituting a contributory moderation pattern (Holbert & Park, 2020). Instead of scrutinizing the content of the information, they are even more likely to trust the information shared by their friends due to the triggering of the bandwagon and homophily heuristics:

H5: Users’ NFM perceptions will moderate the effects specified in H2.

H6: Users’ NFM perceptions will moderate the effects specified in H3.

With the increasing integration of AI (artificial intelligence) in the process of news production and dissemination (Feezell et al., 2021), scholarship on NFM has extended to inquiries on the spread of politically homogeneous opinions and NFM’s effect on people’s attitudes toward algorithmic news selection and social media political homophily (Gil de Zúñiga et al., 2022). Empirical evidence reveals that users with higher levels of NFM are more likely to develop a positive attitude toward algorithmic news selection, as it represents their preconceived notion that algorithms, unlike journalists, efficiently curate the most relevant news to suit their informational needs through their social networks and online engagement (Gil de Zúñiga et al., 2022). This finding suggests the possibility that users with high NFM are more likely to apply the machine heuristic in their everyday news consumption. Therefore, when seeing a cue which indicates that the news is recommended by a machine compared to when it is recommended by news editors or their social media friends, users with high-NFM perceptions are even more likely to trust the news recommendations, and even engage with the news without reading it first due to the triggering of machine heuristic (i.e. a contributory moderation pattern). Therefore, we propose:

H7: Users’ NFM perceptions will moderate the effects specified in H4.

See Figure 1 for a summary of all the hypotheses.

Figure 1.

Summary of hypothesis.

Method

We conducted a three-condition between-subjects experiment (source of news recommendation: news editors vs. social media friends vs. algorithms) after receiving institutional IRB approval. All the hypotheses were pre-registered in Open Science Framework.¹

Participants

We recruited 244 usable participants² from Cloud Research, a platform consisting of high-quality workers from Amazon Mechanical Turk in the United States (Litman et al., 2017). Roughly half of our participants are male (45.5%, N = 111), half of them female (51.2%, N = 125), 0.8% indicated Others (N = 2), and 2.5% chose not to disclose their gender (N = 6). They have an average age of 44.02 (SD = 12.80, range: 19–77), with the majority being Caucasian (75.5%, N = 173), followed by Asian (10.0%, N = 23), African American (6.6%, N = 15), and Hispanic (4.4%, N = 10). They are relatively educated, with 59.8% having a bachelor’s degree or above (N = 146). Their political ideology generally leans slightly toward the liberal side (M = 4.42, SD = 1.84) (1 = very conservative, 7 = very liberal).

Procedure

After providing consent to the study, participants were first asked to fill out a pre-questionnaire in which we measured some of the control variables (i.e. automation bias, reciprocity belief) and their level of NFM (please refer to Supplementary Material III for detailed rationale for statistically controlling these two variables). Then they were randomly assigned to interact with one of the mockup news recommendation systems called the Daily Times, which recommended news selected by either news editors, social media friends, or an algorithm, depending on the condition.

After they were directed to the news system to which they were assigned, they landed on a home page that briefly described how the system recommends news for them (e.g. Figure 2 for the algorithm condition). Next, they were directed to the profile-building page (Figure 3). Participants assigned to social media friend selection conditions were asked to log in to their social media accounts, so the system could find news that their friends recommended. Similarly, participants assigned to algorithm conditions were asked to log in to their web browser account so that the system could understand their news consumption behavior. It should be noted that we provided instructions and hyperlinks to their social media accounts or browser accounts. However, in reality, these hyperlinks simply led to the web-based log-in page of those social media and web browser accounts to give users an illusion that the system is learning about users’ social media ecosystems or news consumption behaviors. For participants assigned to the news editors condition, however, there was no profile-building page since the news articles were ostensibly selected by editors. All participants were then directed to a loading page, which contained the manipulation of the source of recommendations one more time on the interface (Figure 4, left). Finally, all participants were recommended one top-pick news article among several news articles (Figure 4, right). To ensure our findings are generalizable across different news articles, we implemented stimulus sampling such that all participants were randomly assigned to view one of the six news blurbs selected based on a pre-test (described in detail in Supplementary Material I). We built the news recommendation page using Sosci Survey based on the code shared by Unkel (2021), which enabled us to unobtrusively observe users’ actual actions on the news recommendation page, including the time they spent reading the news blurbs, and the first action they took after seeing the news blurb.

Figure 2.

Home page of Daily Times (algorithm selection condition).

Figure 3.

Profile building page for Daily Times (Left: algorithmic selection condition; Right: Social media friends selection condition).

Figure 4.

Daily Times; News Editors condition (left: Loading page; right: News recommendation page).

After users completed their interaction with Daily Times, we provided them a finish code to enter in Qualtrics to access the post-questionnaire in which we asked manipulation check questions, questions about the mediating and dependent variables, as well as their demographic information. We gave participants in the social media friends and algorithms conditions a debrief that the system did not obtain any information from their social media or web browser accounts. All participants received $1.4 compensation upon completion of the study.

Manipulation of the Source of the Recommendation

As we described in the procedure, we embedded cues on the home page, the loading page, and the recommendation page respectively to communicate how the system recommends news to users. See Supplementary Material II for a summary of our manipulation.

Measurement

Except for the behavioral data, all the items were measured on a 7-point Likert-type scale. Please refer to Supplementary Material III for the full questionnaire.

Control Variables

Automation bias was measured using the 10-items Automation-induced Complacency Potential Scale developed by Merritt et al. (2019). Sample item: “When I have a lot to do, it makes sense to delegate a task to automated systems” (M = 4.35, SD = 0.77, α = .79).

Belief in positive reciprocity norms was also controlled as past research has revealed that this belief also predicts users’ tendency to create content (Holton et al., 2015) and share news on social media (Omar et al., 2023). It was measured by using the 10-item scale developed by Eisenberger et al. (2004). Sample item: “If someone does something for me, I feel required to do something for them” (M = 5.28, SD = 0.98, α = .89).

Political ideology was measured using one item “In politics, people sometimes talk about liberal and conservative. Where would you place yourself on a scale from 1 to 7 where 1 means very conservative and 7 means very liberal?” (M = 4.42, SD = 1.84).

The type of news blurbs was also controlled in our analysis. Although we randomly assigned our participants to view one of the news blurbs from the pre-test, we took an additional step by dummy coding this categorical variable and statistically controlling it in our model.

Moderating Variable

News-finds-me perception (NFM) was measured using the six-item scale developed by Gil de Zúñiga et al. (2017) that consists of three dimensions of NFM, including not seeking (e.g. “I do not have to actively seek news because when important public affairs break, they will get to me via social media”), peer reliance (e.g. “I rely on my friends to tell me what’s important when news happens”), and well-informed (e.g. “I can be well-informed even when I don’t actively follow the news”). We also incorporated the fourth dimension of NFM, algorithm reliance, which was later proposed by Gil de Zúñiga and Cheng (2021) (e.g. “I rely on social media algorithms to tell me what’s important when news happens”) (M = 3.38, SD = 1.23, α = .88).

Mediating Variables

We measured four cognitive heuristics as mediators. According to Bellur and Sundar (2014), cognitive heuristics can be operationalized both in terms of the construct accessibility of the heuristic (i.e. belief in heuristic as a moderator or cognitive disposition variable), as well as the attitude accessibility of the heuristic after interaction via self-reported measure (e.g. bandwagon perception as a mediator, such as “Do you think many people like this?” in Sundar et al. (2008)). Bellur and Sundar (2014) suggested that to establish the causal effects of cues in triggering these heuristics, we could vary the cues and measure the perception as a proxy in experimental settings. Therefore, we measured cognitive heuristics as perceptual mediators.

Authority heuristic perception was measured by three self-developed items. Sample item: “The system is trustworthy as the news is recommended by experts” (M = 4.25, SD = 1.37, α = .92).

Bandwagon heuristic perception was measured by adopting six items from Sundar et al. (2008). Sample item: “Many users recommended the news I just saw” (M = 4.57, SD = 0.95, α = .87).

Homophily heuristic perception was measured by four items modified from the scale in Nowak and Rauh (2005). Sample item: “Users who are similar to me like the recommended news” (M = 4.70, SD = 1.16, α = .92).

Machine heuristic perception was measured by adopting four items in Molina and Sundar (2022). Sample item: “The system has machine-like precision” (M = 4.09, SD = 1.18, α = .86).

Dependent Variables

Evaluation of the recommended news was operationalized in terms of sub-dimensions (M = 4.70, SD = 0.88, α = .95). Specifically, the first dimension, perceived news credibility, was measured adopting the three-item message credibility scale (Appelman & Sundar, 2016), for example, “The recommended news article is accurate.” The second dimension was perceived news bias, which was measured with the seven-item scale adopted from Waddell (2019). Sample item: “The recommended news article is biased.” We reverse-coded the items to represent perceived objectivity of the news. The third dimension was perceived newsworthiness, which was measured by six items adopted from Xu (2013). Sample item: “The recommended news article is important.” Finally, perceived credibility of the news recommendation system was measured by seven items in the source credibility scale (McCroskey & Teven, 1999; Metzger & Flanagin, 2013), for example, “The Daily Times is trustworthy.”

Verification Behaviors

First, participants’ actual clicking behavior was recorded unobtrusively on the interface (see Figure 4).³ Participants could either click “click to read more,” “like,” “comment,” or “share” first. If participants selected “like,” “comment,” or “share,” the behavior was coded as 1, indicating that an action was taken prior to reading the news. In contrast, selecting “click to read more” was coded as 0, indicating that the participant chose to read the full news article before engaging in any subsequent action. Second, we measured participants’ perceived intention to verify the recommended news by using four items adopted from Khan and Idris (2019), for example, “I feel I do not have to check the news recommended by the system (reverse coded)” (M = 4.78, SD = 1.25, α = .82).

Data Analysis Plan

We used MANCOVA to test H1a, H2a, H3a, and H4a, and Model 4 of the Process Macro of SPSS (Hayes, 2017) to test H1b, H2b, H3b, and H4b. For H5, H6, and H7, we first used Model 3 of the Process Macro to test the interaction effect between source and NFM on the mediators. Then, we used Model 8 to test moderated mediation effects proposed in H5–H7.

Results

Our manipulation checks were successful (see Supplementary Material IV).

Main Effects of the News Source

Although not directly proposed in H1, H2, H3, and H4 (which were primarily about the indirect effects of recommendation sources on users’ evaluations and actions), we found a main effect of source on the evaluation of the recommended news, F (2, 232) = 4.09, p = .018, R² = .03. Specifically, post hoc analyses with Bonferroni correction revealed that the news recommended by social media friends led to more negative evaluation (M = 4.47, SD = 1.04) compared to news recommended by news editors (M = 4.86, SD = 0.77), p = .016. However, the difference between algorithmic selections (M = 4.73, SD = 0.80) and social media friends (p = .172), or with news editors (p = 1.0) was not significant. We did not find source cues to have a total effect on verification intention, F (2, 232) = .98, p = . 376, R² = .01. Results from the logistic regression revealed that the source of the news recommendations had total effects on users’ first action after seeing the news blurb, χ² (2, N = 244) = 6.86, p = .034. Specifically, when the news was recommended by the algorithm, participants were more likely to share, comment or like first without reading the news (46.3%, N = 38) compared to when the news was recommended by social media friends (31.6%, N = 24) or editors (31.4%, N = 27).

The more positive evaluation of the news recommended by the news editors can be attributed to the triggering of the authority heuristic. Consistent with what H1a proposed, we found that news attributed to news editors was more likely to trigger the authority heuristic (M = 4.74, SD = 1.17) than when it was recommended based on social media friends (M = 3.90, SD = 1.50) (p < .001) or algorithmic selections (M = 4.06, SD = 1.31) (p = .004), F (2, 232) = 8.53, p < .001, R² = .07 (Figure 5), while the difference between the social media friend recommendations and algorithmic selections was not significant, p = 1.00. Mediation analysis using Model 4 of the Process macro reveals that due to the triggering of authority heuristic, news attributed to news editors was evaluated more positively, but that did not translate to their actions or intention to verify the information (Table 1). Therefore, H1a was supported, and H1b was partially supported.

Figure 5.

The main effect of the source of news recommendations on authority heuristic.

Table 1.

Summary of Results From the Mediation Analysis (Mediator: Authority Heuristic).

Mediation path	B ^a	SE	CI^b
Mediation path	B ^a	SE	Lower CI	Upper CI
Editors (0) vs. Social media friends (1) --> Authority Heuristic --> Recommendation evaluation***	−.19	.07	−.33	−.07
Editors (0) vs. Social media friends (1) --> Authority Heuristic --> Intent to verify	−.00	.07	−.13	.14
Editors (0) vs. Social media friends (1) --> Authority Heuristic --> Action^c	−.13	.12	−.40	.08
Editors (0) vs. Algorithm (1) --> Authority Heuristic --> Recommendation evaluation***	−.16	.06	−.27	−.06
Editors (0) vs. Algorithm (1) --> Authority Heuristic --> Intent to verify	−.00	.06	−.11	.12
Editors (0) vs. Algorithm (1) --> Authority Heuristic --> Action^c	−.11	.10	−.34	.06

Note. All four mediators (i.e. bandwagon heuristic, homophily heuristic, machine heuristic, and authority heuristic) were entered in the model.

Unstandardized path coefficient.

Bias-correlated and accelerated 95% confidence interval (CI).

Action is coded as 0 (first action: read more), 1 (first action: like, share, or comment).

***

p < .001.

However, contrary to what was proposed in H2a, we found no significant differences across the three conditions in terms of bandwagon heuristic perception (news editors: M = 4.64, SD = 0.84; social media friends: M = 4.58, SD = 1.04; algorithms: M = 4.48, SD = 0.93), F (2, 232) = 0.66, p = .521, R² = .01, so H2a was not supported. In a similar vein, we did not find significant differences across the three conditions in terms of homophily heuristic perception (news editors: M = 4.68, SD = 1.11; social media friends: M = 4.80, SD = 1.18; algorithms: M = 4.64, SD = 1.21), F (2, 232) = 0.64, p = .529, R² = .01, meaning H3a was also not supported.

H4 proposed that users will be more likely to evaluate the system and the recommended news more positively, and even like, share, or comment on the news without reading it first, if an algorithm recommends the news, due to the triggering of the machine heuristic. Results from the MANCOVA revealed that there was no significant difference across the three conditions (algorithm: M = 4.21, SD = 1.11; social media friends: M = 4.04, SD = 1.30; news editors: M = 4.03, SD = 1.14), F (2, 232) = .59, p = .558, R² = .01. Therefore, H4 was not supported.

The Moderating Effects of NFM Perceptions

First, we found that NFM and source has a significant interaction effect on news recommendation evaluations, F (2, 230) = 3.20, p = .043, R² = .03 (see Figure 6), but not intention to verify, F (2, 230) = 2.47, p = .087, R² = .02, or first action taken, χ² (2, N = 244) = 1.01, p = .604. Specifically, we found that when the source was an algorithm or social media friends, NFM was significantly positively related to participants’ evaluation of the news recommendations; however, this relationship was not significant when the recommendation source was news editors (Figure 6).

Figure 6.

Interaction effect between the source of news recommendations and NFM on evaluation of news recommendations.

By analyzing the data using Model 1 of the Process macro, we found that source and NFM did not have an interaction effect on the bandwagon heuristic, F (2, 230) = 0.46, p = .634, R² = .00, or the homophily heuristic, F (2, 230) = 1.31, p = .273, R² = .01. Therefore, H5 and H6 were not supported.

H7 further proposed that high-NFM individuals were more likely to trigger the machine heuristic when receiving news recommended by an algorithm compared to low-NFM individuals. We found an interaction effect between source and NFM on machine heuristic, in support of H7’s prediction, but this effect fell just short of statistical significance, F (2, 230) = 2.85, p = .06, R² = .03 (see Figure 7). We found that algorithmic news recommendation was more likely to trigger the machine heuristic among high-NFM individuals compared to low-NFM individuals. Yet the relationship between NFM and machine heuristic was not significant when the sources are social media friends or news editors. Model 8 of the Process Macro also revealed that machine heuristic perception was a significant mediator between NFM, source, and evaluation of the news recommendation and intent to verify. Specifically, algorithmic news recommendation was more likely to trigger the machine heuristic among high-NFM individuals, and was further associated with more positive evaluation outcomes of the news and the system and lower intention to verify the information, although it did not predict participants’ actions (Table 2). Please also refer to supplementary material V for more details of the moderation effect.

Figure 7.

Interaction effect between the source of news recommendations and NFM on machine heuristic.

Table 2.

Summary of Results From the Moderated Mediation Analyses.

Mediation path	B ^a	SE	CI^b
Mediation path	B ^a	SE	Lower CI	Upper CI
Editors (0) vs. Social media friends (1) x NFM --> MH^c --> Recommendation evaluation	−.00	.04	−.08	.07
Editors (0) vs. Social media friends (1) x NFM --> MH --> Intent to verify	.01	.06	−.11	.12
Editors (0) vs. Social media friends (1) x NFM --> MH --> Action^d	.00	.03	−.06	.07
Editors (0) vs. Algorithm (1) x NFM --> MH --> Recommendation evaluation***	.06	.03	.00	.13
Editors (0) vs. Algorithm (1) x NFM --> MH --> Intent to verify***	−.10	.05	−.22	−.00
Editors (0) vs. Algorithm (1) x NFM --> MH --> Action^d	−.03	.05	−.14	.06
Editors (0) vs. Social media friends (1) x NFM --> AH^e --> Recommendation evaluation	.04	.04	−.04	.13
Editors (0) vs. Social media friends (1) x NFM --> AH --> Intent to verify	−.01	.02	−.06	.03
Editors (0) vs. Social media friends (1) x NFM --> AH--> Action^d	.03	.05	−.05	.15
Editors (0) vs. Algorithm (1) x NFM --> AH --> Recommendation evaluation***	.10	.04	.03	.18
Editors (0) vs. Algorithm (1) x NFM --> AH --> Intent to verify***	−.02	.03	−.09	.05
Editors (0) vs. Algorithm (1) x NFM --> AH --> Action^d	.08	.06	−.02	.23

Unstandardized path coefficient.

Bias-correlated and accelerated 95% confidence interval (CI).

MH = Machine heuristic.

Action is coded as 0 (first action: read more), 1 (first action: like, share, or comment).

AH = Authority heuristic.

***

p < .001.

Exploratory Analyses

Given the overwhelming evidence for authority heuristic (over bandwagon and homophily heuristics) in mediating the effects of source, we conducted an exploratory analysis to further test if NFM moderates these mediation effects on the dependent variables.⁴ Results revealed a significant interaction effect between source and NFM on the triggering of the authority heuristic, F (2, 230) = 3.63, p = .028, R² = .03. Specifically, we found that users with lower levels of NFM were more likely to perceive news editors as more of an authority compared to the other two sources. However, users with high NFM tended to perceive a high level of authority regardless of the recommendation source. In other words, high-NFM users were more likely to perceive the algorithm and their social media friends to be as reliable and authoritative in recommending news as the news editors (Figure 8). Please also refer to supplementary material V for more details of the moderation effect.

Figure 8.

Interaction effect between source and NFM on authority heuristic.

Results from Model 8 of the Process Macro revealed that the authority heuristic mediated the interaction effects of NFM and source on users’ evaluation of the system (Table 2).

Summary of Findings

In summary, we found that users are likely to perceive the news recommended by news editors to be more credible, newsworthy, and less biased compared to when the news was recommended by their social media friends or an algorithm, due to the triggering of the authority heuristic. However, algorithmic news recommendations are more likely to be shared/liked/commented by users without first reading them.

More interestingly, we found that the higher trust toward news editors does not hold true for users with higher NFM. On the contrary, they are more likely to trust the system and the news recommended by an algorithm due to the triggering of the machine heuristic. In addition, we found that, different from users with low NFM who usually think news editors are the authority in news recommendations, high-NFM users were more likely to perceive the algorithm and their social media friends to be as reliable and authoritative as news editors in recommending news, which explains why they are more likely to trust the news recommended by an algorithm or their social media friends.

Discussion

Our study extends the current literature on the detrimental effects of NFM by investigating the underlying psychological mechanisms of high-NFM users’ processing and sharing of news received from different sources. First, we found that general online readers tend to view editors as more trustworthy than other news selection sources such as social media friends and algorithms because of the operation of the authority heuristic. Theoretically, this finding suggests that baseline credibility at the psychological level rests on a perception of “authority,” that is, online users are more likely to trust information presented by sources perceived as authoritative or credible (Sundar, 2008). Although it has been reported that users’ confidence in traditional media has declined over the years (Lee, 2010), our study supports their relative appreciation of recommendations from experts—news editors, in our case—rather than social media friends or algorithms. This also corroborates previous research showing that news organizations with established reputations and editorial oversight are generally perceived as more trustworthy and credible by readers (Swasy et al., 2015).

Interestingly, although users perceive the news recommended by news editors as more credible, newsworthy, and less biased, they engage with the news (i.e. share/comment/like, often without reading it) more when it is recommended by an algorithm. The presence of a recommendation algorithm appears to be the reason underlying findings in the literature which suggest that sharing without clicking is a widespread phenomenon on social media. For instance, a recent study analyzed over 35 million URLs shared on Facebook from 2017 to 2020 and revealed that as much as 75% of news links were shared without being read (Sundar et al., 2025). However, contrary to what is proposed in the literature, our data suggest that machine heuristic is not the underlying reason for this behavioral effect, even though it did explain other perceptual outcomes of algorithms among certain individuals (i.e. users with higher NFM). One possible explanation could be that users’ stance toward algorithm-recommended news is different from their stance toward news recommended by news editors or social media friends. When a news professional curates the news for users, the story could be considered a verified publication that represents the ground truth, which might increase the tendency for users to try and read it more before taking any actions. By contrast, users might view the news recommended by an algorithm as lacking such careful and effortful curation, but rather as a suggestion for consumption in the public domain. This could potentially explain why their immediate reaction is to take actions, such as sharing, liking, and commenting without perusing it first. This is corroborated by the fact that our participants engaged in more mindless sharing/liking/commenting behaviors when the algorithm is the attributed source of the news, despite perceiving news editors as a more trustworthy source. This finding also demonstrates the importance of measuring actual behavioral engagement with online news content in addition to perceptual measures, as it provides a more accurate and comprehensive understanding of how users consume and interact with news in the online environment.

The trust toward—and overreliance on—news algorithms is especially true for users with higher levels of NFM perceptions. First, this finding explains the inconsistent results in the literature regarding user trust toward news algorithms, with some showing that users tend to trust editorial choices made by journalistic professionals over algorithmic curation (Swasy et al., 2015) and others showing the exact opposite (Thurman et al., 2019). Our study highlights the importance of taking users’ personal characteristics and prior beliefs, such as NFM, into consideration when evaluating users’ trust in different news sources. According to the Heuristic–Systematic Model (Chaiken, 1989), the activation of cognitive heuristics is shaped by individuals’ prior beliefs and the cognitive accessibility of those heuristics. In this regard, NFM perceptions may function as a boundary condition that represents the accessibility of different cognitive heuristics in individuals’ minds. Indeed, a recent study by Campbell and Hawkins (2025) demonstrated that individuals with varying levels of NFM hold distinct social media mindsets, that is, core beliefs about their relationship with social media and the role it plays in their lives. Specifically, individuals who perceive themselves as having low agency over their social media use, while attributing high agency or control to algorithms over news exposure, are more likely to develop NFM beliefs such as “not seeking” and “being informed,” largely due to habitual social media use. Similar to mindsets, cognitive heuristics are shaped by prior experiences, and function as mental shortcuts that guide individuals’ judgments and decision-making (Chaiken, 1980). Extending this line of research, the present study demonstrates that, beyond differences in social media mindsets, individuals with varying levels of NFM may also differ in the cognitive accessibility of heuristics triggered by source cues, which helps explain why high-NFM users place greater trust in social media friends and algorithms to find news for them.

Second, our study makes a valuable contribution to the current literature on NFM by emphasizing the significance of incorporating news algorithms alongside peer reliance when investigating this phenomenon. Gil de Zúñiga and Cheng (2021) added a fourth attribute, algorithmic reliance, to the original NFM concept to capture individuals’ increasing dependence on algorithms to receive news, and even create their own echo chamber of ideologically-congenial information (Gil de Zúñiga et al., 2022). Our study extends this line of research by revealing that high-NFM users are not only more likely to rely on algorithms to receive news but also more likely to perceive the news recommended by an algorithm as credible. This higher evaluation of algorithmically curated news can be concerning, given past research showing that news algorithms tend to drive political polarization (Barberá, 2020). We also uncovered the underlying psychological mechanism of this effect, which is the triggering of the machine heuristic among users with high NFM, thus supporting our hypothesis that high-NFM users are more likely to be influenced by certain source cues and engage in heuristic rather than systematic information processing of the news content. This underscores the importance of implementing additional measures to prevent high-NFM users from relying solely on heuristic cues when evaluating the credibility of online information.

Finally, although we did not find support for our hypotheses that high-NFM users tend to trust their social media friends’ news recommendations more due to the triggering of the bandwagon heuristic or the homophily heuristic, our exploratory analysis revealed that it is because users with high NFM think social media friends and algorithms are as authoritative as news editors in recommending news, quite unlike users with lower NFM perceptions, who believe that news editors have more authority and expertise in selecting news than algorithms and social media friends. This higher authority perception further explains why they are more likely to trust the news recommended by an algorithm or their social media friends. This result is particularly concerning, as only news editors have the requisite training and experience in ensuring news gatekeeping standards. Social media friends and algorithms, on the other hand, are “pseudo cognitive authorities” (Froehlich, 2019) in that they might appear credible and trustworthy on the surface, but upon closer scrutiny, fail to possess these qualities; they tend to promote a biased or partisan agenda, often disregarding truth, evidence, logic, and facts. Their greater trust in social media friends and algorithms as experts and authorities might also explain why high-NFM users are more vulnerable to misinformation (Diehl & Lee, 2022). The findings of our study underscore the need for designing more comprehensive literacy programs that educate high-NFM individuals about news gatekeeping standards. This will help prevent the misattribution of expertise to peers and algorithms, which can lead to a distorted assessment of the credibility of online information.

Limitations

Our study is not free from limitations. First, even though we used stimulus sampling by randomly presenting participants with news we selected from the pre-test, we only provided each participant with one news recommendation as their top pick for the day. The reason for this design is to avoid overwhelming our participants, minimize incidental confounds (such as the layout and positions of the recommended news), and ensure that we could record all the behaviors users perform accurately on the interface. Yet, this interaction might have been too short for some participants. Future studies could come up with more creative ways to unobtrusively track users’ engagement with the news and randomize the display of various news recommendations. Another limitation could be that we operationalized NFM as a single-dimensional construct, whereas an increasing number of studies have examined the effects of its constituent dimensions separately (Campbell & Hawkins, 2025; T. Lee et al., 2025). Future research could disentangle these dimensions to examine how each uniquely contributes to the activation of cognitive heuristics, ideally using larger samples. More generally, our sample size may be limited and potentially underpowered to fully detect moderation effects involving NFM. Our prospective power analyses were based on the assumption of a medium effect size for the main effects found in the literature, but our analyses involved more complex effects, including moderation and mediation, for which we may have lacked power even though we employed 5000 bootstrapped samples via PROCESS, which reduces the sample-size requirements (Hayes & Scharkow, 2013; Sim et al., 2022). We would like to caution readers in interpreting the null findings reported in this study and encourage more future studies to replicate and extend our findings with larger samples. Finally, we only included participants from the United States, a significant limitation considering that NFM is known to have different outcomes across different cultural contexts (Gil de Zúñiga et al., 2020). Future studies would do well to replicate our study in other cultural contexts.

Conclusion

While prior research has documented the prevalence and negative effects of the News Finds Me perception (NFM) among online news consumers, we lacked a theoretical understanding of why high-NFM users tend to rely on their social media friends or algorithms for news. Our study has revealed that users’ overreliance on these sources might be rooted in their higher trust in news algorithms due to the triggering of machine heuristic, and their mistaken belief that social media friends and algorithms are as authoritative as news editors in recommending news. By shedding light on why users form NFM perceptions, our study offers a new way forward for better mitigating its negative effects, such as reduced knowledge of public affairs and rampant spread of misinformation.

Supplemental Material

sj-docx-1-sms-10.1177_20563051261434801 – Supplemental material for When We Think “News Will Find Me”: Relative Credibility of Social-Media Friends, Algorithms, and Editors

Supplemental material, sj-docx-1-sms-10.1177_20563051261434801 for When We Think “News Will Find Me”: Relative Credibility of Social-Media Friends, Algorithms, and Editors by Mengqi Liao, Yuan Sun, Timilehin Durotoye, Homero Gil de Zúñiga and S. Shyam Sundar in Social Media + Society

Footnotes

ORCID iDs

Mengqi Liao

Yuan Sun

Timilehin Durotoye

Homero Gil de Zúñiga

S. Shyam Sundar

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported in part by the MSIT (Ministry of Science, ICT), Korea, under the Global Scholars Invitation Program (RS-2024-00459638) administered by IITP (Institute for Information & Communications Technology Planning & Evaluation).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The data underlying this article will be shared on reasonable request to the corresponding author.

Supplemental material

Supplemental material for this article is available online.

Notes

Author biographies

Mengqi Liao (PhD in Mass Communication, Penn State University) is an Assistant Professor in the Department of Advertising and Public Relations at Grady College of Journalism & Mass Communication, University of Georgia. Her research explores the process and psychological effects of human interactions with communication technology, including social media, virtual reality, mobile devices, and emerging artificial intelligence (AI) applications. Her research aims to gain a deep understanding of the effects of different technological affordances on persuasion, users’ cognitive information processing, psychological well-being, and trust.

Yuan Sun (PhD in Mass Communication, Penn State University) is an Assistant Professor in the Department of Advertising at the University of Florida College of Journalism and Communications. Her research primarily focuses on the role and impact of artificial intelligence (AI) in strategic communication, including how user trust is developed and calibrated in the context of digital health and advertising. She also studies the psychological effects of social media and advertising campaigns on prosocial behaviors.

Timilehin Durotoye (MA in Media Studies, The University of Ibadan, Nigeria) is a PhD candidate in the Donald P. Bellisario College of Communications at the Pennsylvania State University. She earned her bachelor’s and master’s degrees from the Department of Communication and Language Arts, University of Ibadan, Nigeria. Durotoye’s research centers on the effects of social media, algorithms, and artificial intelligence (AI) on human-social interaction, long-debated and emergent political behavior capable of influencing democratic processes and media policies.

Homero Gil de Zúñiga (PhD in Politics at Universidad Europea de Madrid and PhD in Mass Communication, University of Wisconsin–Madison) is Distinguished Research Professor–Chair in Political Behavior at University of Salamanca, Distinguished Professor of Media Effects & AI at Pennsylvania State University, and Senior Research Fellow at Universidad Diego Portales. In general, his work draws from theoretically driven research, aiming to shed an empirical social scientific light over how social media, algorithms, AI, and other technologies affect society and democracy.

S. Shyam Sundar (PhD in Communication, Stanford University) is Evan Pugh University Professor, James P. Jimirro Professor of Media Effects, and Director of the Center for Socially Responsible Artificial Intelligence at Penn State University. He is also the founding director of the Media Effects Research Laboratory at Penn State’s Bellisario College of Communications. His research examines social and psychological effects of interactive media. Specifically, his experiments investigate the role played by technological affordances in shaping user experience of mediated communications. Current research pertains to psychological effects of Human-AI interaction in a variety of contexts, ranging from personalization and recommendation to fake news and content moderation.

References

Allcott

Gentzkow

(2017). Social media and fake news in the 2016 election. Journal of Economic Perspectives, 31(2), 211–236.

Altay

De Araujo

Mercier

(2022). “If this account is true, it is most enormously wonderful”: Interestingness-if-true and the sharing of true and false news. Digital Journalism, 10(3), 373–394.

Appelman

Sundar

S. S.

(2016). Measuring message credibility: Construction and validation of an exclusive scale. Journalism & Mass Communication Quarterly, 93(1), 59–79. https://doi.org/10.1177/1077699015606057

Barberá

(2020). Social media, echo chambers, and political polarization. In Persily

Tucker

J. A.

(Eds.), Social media and democracy: The state of the field, prospects for reform (pp. 34–55). Cambridge University Press.

Bellur

Sundar

S. S.

(2014). How can we tell when a heuristic has been used? Design and analysis strategies for capturing the operation of heuristics. Communication Methods and Measures, 8(2), 116–137. https://doi.org/10.1080/19312458.2014.903390

Campbell

S. W.

Hawkins

(2025). Social (media) psychology of the “news-finds-me” perception: Habits, mindsets, and beliefs. Journal of Computer-Mediated Communication, 30(5), zmaf010. https://doi.org/10.1093/jcmc/zmaf010

Chaiken

(1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of Personality and Social Psychology, 39(5), 752–766. https://doi.org/10.1037/0022-3514.39.5.752

Chaiken

(1987). The heuristic model of persuasion. In Zanna

M. P.

Olson

J. M.

Herman

C. P.

(Eds.), Social influence: The Ontario symposium (Vol. 5, pp. 3–39). Lawrence Erlbaum Associates.

Chaiken

Liberman

Eagly

A. H.

(1989). Heuristic and systematic information processing within and beyond the persuasion context. In Uleman

J. S.

Bargh

J. A.

(Eds.), Unintended thought (pp. 212–252). The Guilford Press.

10.

DeVito

M. A.

Gergle

Birnholtz

(2017). “Algorithms ruin everything”: #RIPTwitter, folk theories, and resistance to algorithmic change in social media. In Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems (pp. 3163–3174). Association for Computing Machinery. https://doi.org/10.1145/3025453.3025659

11.

Diehl

Lee

(2022). Testing the cognitive involvement hypothesis on social media: “News finds me” perceptions, partisanship, and fake news credibility. Computers in Human Behavior, 128, 107121.

12.

Eisenberger

Lynch

Aselage

Rohdieck

(2004). Who takes the most revenge? Individual differences in negative reciprocity norm endorsement. Personality and Social Psychology Bulletin, 30(6), 787–799. https://doi.org/10.1177/0146167204264047

13.

Faul

Erdfelder

Lang

A.-G.

Buchner

(2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39(2), 175–191. https://doi.org/10.3758/BF03193146

14.

Feezell

J. T.

Wagner

J. K.

Conroy

(2021). Exploring the effects of algorithm-driven news sources on political behavior and polarization. Computers in Human Behavior, 116, 106626. https://doi.org/10.1016/j.chb.2020.106626

15.

Fletcher

Nielsen

R. K.

(2019). Generalised scepticism: How people navigate news on social media. Information, Communication & Society, 22(12), 1751–1769. https://doi.org/10.1080/1369118X.2018.1450887

16.

Froehlich

T. J.

(2019). The role of pseudo-cognitive authorities and self-deception in the dissemination of fake news. Open Information Science, 3(1), 115–136. https://doi.org/10.1515/opis-2019-0009

17.

Gans

H. J.

(2004). Deciding what’s news: A study of CBS evening news, NBC nightly news, Newsweek, and Time. Northwestern University Press.

18.

Gil de Zúñiga

Cheng

(2021). Origin and evolution of the News Finds Me perception: Review of theory and effects. Profesional de La Información, 30(3), 1-17.

19.

Gil de Zúñiga

Cheng

González-González

(2022). Effects of the News Finds Me perception on algorithmic news attitudes and social media political homophily. Journal of Communication, 72(5), 578–591.

20.

Gil de Zúñiga

Goyanes

Durotoye

(2023). A scholarly definition of Artificial Intelligence (AI): Advancing AI as a conceptual framework in communication research. Political Communication, 41(2), 317–334. https://doi.org/10.1080/10584609.2023.2290497

21.

Gil de Zúñiga

Strauss

Huber

(2020). The proliferation of the “News Finds Me” perception across societies. International Journal of Communication, 14, 1605–1633.

22.

Gil de Zúñiga

Weeks

Ardèvol-Abreu

(2017). Effects of the News-Finds-Me perception in communication: Social media use implications for news seeking and learning about politics. Journal of Computer-Mediated Communication, 22(3), 105–123. https://doi.org/10.1111/jcc4.12185

23.

Graefe

Haim

Haarmann

Brosius

H.-B.

(2018). Readers’ perception of computer-generated news: Credibility, expertise, and readability. Journalism, 19(5), 595–610.

24.

Halberstam

Knight

(2016). Homophily, group size, and the diffusion of political information in social networks: Evidence from Twitter. Journal of Public Economics, 143, 73–88. https://doi.org/10.1016/j.jpubeco.2016.08.011

25.

Hayes

A. F.

(2017). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach (2nd ed.). Guilford Press.

26.

Hayes

A. F.

Scharkow

(2013). The relative trustworthiness of inferential tests of the indirect effect in statistical mediation analysis: Does method really matter? Psychological Science, 24(10), 1918–1927. https://doi.org/10.1177/0956797613480187

27.

Hermida

(2010). Twittering the news: The emergence of ambient journalism. Journalism Practice, 4(3), 297–308. https://doi.org/10.1080/17512781003640703

28.

Holbert

R. L.

Park

(2020). Conceptualizing, organizing, and positing moderation in communication research. Communication Theory, 30(3), 227–246. https://doi.org/10.1093/ct/qtz006

29.

Holmström

Jonsson

Polbratt

Nilsson

Lundström

Ragnarsson

Forsberg

Andersson

Carlsson

(2019). Do we read what we share? Analyzing the click dynamic of news articles shared on Twitter. Proceedings of the 2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (pp. 420–425). IEEE.

30.

Holton

A. E.

Coddington

Lewis

S. C.

Zúñiga

H. G. de

. (2015). Reciprocity and the news: The role of personal and social media reciprocity in news creation and consumption. International Journal of Communication, 9, 22.

31.

Khan

M. L.

Idris

I. K.

(2019). Recognise misinformation and verify before sharing: A reasoned action and information literacy perspective. Behaviour & Information Technology, 38(12), 1194–1212. https://doi.org/10.1080/0144929X.2019.1578828

32.

Kim

E.-M.

Ihm

(2020). Online news sharing in the face of mixed audiences: Context collapse, homophily, and types of social media. Journal of Broadcasting & Electronic Media, 64(5), 756–776.

33.

Lee

(2020). Probing the mechanisms through which social media erodes political knowledge: The role of the news-finds-me perception. Mass Communication and Society, 23(6), 810–832.

34.

Lee

Johnson

Jia

Lacasa-Mas

(2025). How social media users become misinformed: The roles of news-finds-me perception and misinformation exposure in COVID-19 misperception. New Media & Society, 27(3), 1730–1751. https://doi.org/10.1177/14614448231202480

35.

Lee

T.-T.

(2010). Why they don’t trust the media: An examination of factors predicting trust. American Behavioral Scientist, 54(1), 8–21.

36.

Liao

(2023). Understanding the effects of personalized recommender systems on political news perceptions: A comparison of content-based, collaborative, and editorial choice-based news recommender system. Journal of Broadcasting & Electronic Media, 67(3), 294–322. https://doi.org/10.1080/08838151.2023.2206662

37.

Lin

Chen

Lee

S. Y.

S. H.

Chen

Tandoc

E. C.

Goh

Z. H.

Salmon

C. T.

(2024). Understanding the effects of News-Finds-Me perception on health knowledge and information seeking during public health crises. Health Communication, 39(2), 352–362. https://doi.org/10.1080/10410236.2023.2165750

38.

Litman

Robinson

Abberbock

(2017). TurkPrime.com: A versatile crowdsourcing data acquisition platform for the behavioral sciences. Behavior Research Methods, 49(2), 433–442. https://doi.org/10.3758/s13428-016-0727-z

39.

McCroskey

J. C.

Teven

J. J.

(1999). Goodwill: A reexamination of the construct and its measurement. Communication Monographs, 66(1), 90–103. https://doi.org/10.1080/03637759909376464

40.

Merritt

S. M.

Ako-Brew

Bryant

W. J.

Staley

McKenna

Leone

Shirase

(2019). Automation-induced complacency potential: Development and validation of a new scale. Frontiers in Psychology, 10, Article 225. https://doi.org/10.3389/fpsyg.2019.00225

41.

Messing

Westwood

S. J.

(2014). Selective exposure in the age of social media: Endorsements trump partisan source affiliation when selecting news online. Communication Research, 41(8), 1042–1063. https://doi.org/10.1177/0093650212466406

42.

Metzger

Flanagin

(2013). Credibility and trust of information in online environments: The use of cognitive heuristics. Journal of Pragmatics, 59, 210–220. https://doi.org/10.1016/j.pragma.2013.07.012

43.

Molina

M. D.

Sundar

S. S.

(2022). When AI moderates online content: Effects of human collaboration and interactive transparency on user trust. Journal of Computer-Mediated Communication, 27(4), zmac010. https://doi.org/10.1093/jcmc/zmac010

44.

Mosallaei

Wang

Ognyanova

(2025). From politics to entertainment: Exploring “News Finds Me” perceptions across news topics. Social Media + Society, 11(4), 20563051251382442. https://doi.org/10.1177/20563051251382442

45.

Nowak

K. L.

Rauh

(2005). The influence of the avatar on online perceptions of anthropomorphism, androgyny, credibility, homophily, and attraction. Journal of Computer-Mediated Communication, 11(1), 153–178. https://doi.org/10.1111/j.1083-6101.2006.tb00308.x

46.

Oeldorf-Hirsch

Srinivasan

(2022). An unavoidable convenience: How post-millennials engage with the news that finds them on social and mobile media. Journalism, 23(9), 1939–1954.

47.

Omar

Apuke

O. D.

Nor

Z. M.

(2023). The intrinsic and extrinsic factors predicting fake news sharing among social media users: The moderating role of fake news awareness. Current Psychology, 43, 1235–1247. https://doi.org/10.1007/s12144-023-04343-4

48.

Park

C. S.

Kaye

B. K.

(2020). What’s this? Incidental exposure to news on social media, news-finds-me perception, news efficacy, and news consumption. Mass Communication and Society, 23(2), 157–180.

49.

Petty

R. E.

Cacioppo

J. T.

(1986). The elaboration likelihood model of persuasion. In Petty

R. E.

Cacioppo

J. T.

(Eds.), Communication and persuasion: Central and peripheral routes to attitude change (pp. 1–24). Springer. https://doi.org/10.1007/978-1-4612-4964-1_1

50.

Rocca

K. A.

McCroskey

J. C.

(1999). The interrelationship of student ratings of instructors’ immediacy, verbal aggressiveness, homophily, and interpersonal attraction. Communication Education, 48(4), 308–316.

51.

Sim

Kim

S.-Y.

Suh

(2022). Sample size requirements for simple and complex mediation models. Educational and Psychological Measurement, 82(1), 76–106. https://doi.org/10.1177/00131644211003261

52.

Skurka

Cheng

Goyanes

Gil de Zúñiga

(2025). News Finds Me as the illusion of competence: Evidence for overconfidence in discernment of political misinformation. Human Communication Research, 52, 11–23. https://doi.org/10.1093/hcr/hqaf015

53.

Skurka

Liao

Gil de Zúñiga

(2023). Tuning out (political and science) news? A selective exposure study of the news finds me perception. Communication Research. https://doi.org/10.1177/00936502231215528

54.

Song

Gil de Zúñiga

Boomgaarden

H. G.

(2020). Social media news use and political cynicism: Differential pathways through “news finds me” perception. Mass Communication and Society, 23(1), 47–70.

55.

Strauß

Huber

Gil de Zúñiga

(2021). Structural influences on the News Finds Me perception: Why people believe they don’t have to actively seek news anymore. Social Media+ Society, 7(2), 20563051211024966.

56.

Strömbäck

Shehata

(2018). Political journalism. In Nussbaum

J. F.

(Ed.), Oxford research encyclopedia of communication. https://doi.org/10.1093/acrefore/9780190228613.013.859

57.

Luo

Borah

(2024). Learning about climate change with algorithmic news? A two-wave panel study examining the role of “news-finds-me” perception. Journal of Computer-Mediated Communication, 29(5), zmae010. https://doi.org/10.1093/jcmc/zmae010

58.

Sun

Xie

(2024). Do heuristic cues affect misinformation sharing? Evidence from a meta-analysis. Journalism & Mass Communication Quarterly, 103(1), 249–272. https://doi.org/10.1177/10776990241284597

59.

Sundar

S. S.

(2008). The MAIN model: A heuristic approach to understanding technology effects on credibility. In Metzger

M. J.

Flanagin

A. J.

(Eds.), Digital media, youth, and credibility (pp. 73–100). MIT Press.

60.

Sundar

S. S.

Knobloch-Westerwick

Hastall

M. R.

(2007). News cues: Information scent and cognitive heuristics. Journal of the American Society for Information Science and Technology, 58(3), 366–378.

61.

Sundar

S. S.

Nass

(2001). Conceptualizing sources in online news. Journal of Communication, 51(1), 52–72. https://doi.org/10.1111/j.1460-2466.2001.tb02872.x

62.

Sundar

S. S.

Oeldorf-Hirsch

(2008). The bandwagon effect of collaborative filtering technology. In CHI ’08 Extended Abstracts on Human Factors in Computing Systems (pp. 3453–3458). Association for Computing Machinery. https://doi.org/10.1145/1358628.1358873

63.

Sundar

S. S.

Snyder

E. C.

Liao

Yin

Wang

Chi

(2025). Sharing without clicking on news in social media. Nature Human Behaviour, 9, 156–168. https://doi.org/10.1038/s41562-024-02067-4

64.

Surowiecki

(2005). The wisdom of crowds. Knopf Doubleday.

65.

Swasy

Tandoc

Bhandari

Davis

(2015). Traditional reporting more credible than citizen news. Newspaper Research Journal, 36(2), 225–236.

66.

Thurman

Lewis

S. C.

Kunert

(2019). Algorithms, automation, and news. Digital Journalism, 7(8), 980–992. https://doi.org/10.1080/21670811.2019.1685395

67.

Tian

Willnat

(2025). From news disengagement to fake news engagement: Examining the role of news-finds-me perceptions in vulnerability to fake news through third-person perception. Computers in Human Behavior, 162, 108431. https://doi.org/10.1016/j.chb.2024.108431

68.

Turcotte

York

Irving

Scholl

R. M.

Pingree

R. J.

(2015). News recommendations from social media opinion leaders: Effects on media trust and information seeking. Journal of Computer-Mediated Communication, 20(5), 520–535.

69.

Unkel

(2021). Measuring selective exposure in mock website experiments: A simple, free, and open-source solution. Communication Methods and Measures, 15(1), 1–16. https://doi.org/10.1080/19312458.2019.1708284

70.

Van Aelst

Strömbäck

Aalberg

Esser

De Vreese

Matthes

Hopmann

Salgado

Hubé

Stępińska

. (2017). Political communication in a high-choice media environment: A challenge for democracy? Annals of the International Communication Association, 41(1), 3–27.

71.

Waddell

T. F.

(2019). Can an algorithm reduce the perceived bias of news? Testing the effect of machine attribution on news readers’ evaluations of bias, anthropomorphism, and credibility. Journalism & Mass Communication Quarterly, 96(1), 82–100. https://doi.org/10.1177/1077699018815891

72.

Wang

Walther

J. B.

Pingree

Hawkins

R. P.

(2008). Health information, credibility, homophily, and influence via the Internet: Web sites versus discussion groups. Health Communication, 23(4), 358–368.

73.

Weismueller

Harrigan

Coussement

Tessitore

(2022). What makes people share political content on social media? The role of emotion, authority and ideology. Computers in Human Behavior, 129, 107150. https://doi.org/10.1016/j.chb.2021.107150

74.

(2013). Social recommendation, source credibility, and recency: Effects of news cues in a social bookmarking website. Journalism & Mass Communication Quarterly, 90(4), 757–775. https://doi.org/10.1177/1077699013503158

75.

Zhang

Jiang

(2024). “I know news will find me”: Examining the relationship between the “News-Finds-Me” perception and COVID-19 misperceptions. Health Communication, 39(13), 3032–3043. https://doi.org/10.1080/10410236.2023.2300902

When We Think “News Will Find Me”: Relative Credibility of Social-Media Friends,Algorithms,and Editors

Abstract

Keywords

Literature Review

Effects of News Recommendation Sources

Effects of News Finds Me Perception

Method

Participants

Procedure

Manipulation of the Source of the Recommendation

Measurement

Control Variables

Moderating Variable

Mediating Variables

Dependent Variables

Verification Behaviors

Data Analysis Plan

Results

Main Effects of the News Source

The Moderating Effects of NFM Perceptions

Exploratory Analyses

Summary of Findings

Discussion

Limitations

Conclusion

Supplemental Material

sj-docx-1-sms-10.1177_20563051261434801 – Supplemental material for When We Think “News Will Find Me”: Relative Credibility of Social-Media Friends, Algorithms, and Editors

Footnotes

ORCID iDs

Funding

Declaration of conflicting interests

Data availability statement

Supplemental material

Notes

Author biographies

References