Influence of News-Finds-Me Perception on accuracy, factuality and relevance assessment. Case study of news item on climate change

This paper analyses the effect of ‘News-Finds- Me' Perception on the user's assessment of a news item on climate change, both directly and indirectly through news-sharing habits, news-consumption habits, uses and gratifications on social media sites, prior knowledge and attitudes regarding the decision to share said news item. Ninety-six Spanish university students took a survey and were asked to read and rate the quality of a news item on climate change, then to indicate whether they would share that news item on social media. The results show that students with a higher news-finds-me perception tended to rate accuracy and factuality more highly than students with a lower news-finds-me perception. However,


Influence of News-Finds-Me Perception on accuracy, factuality and relevance assessment. Case study of news item on climate change
Francisco Segado-Boj, Jesús Díaz-Campo, Enrique Navarro-Asensio, Lorena

Introduction
Social media has potentially become the most important news distribution channel ever (Dafonte-Gómez, 2018) and a source for political and civic information among citizens (Anspach, Jennings & Arceneaux, 2019). This trend has gained global dimensions in western countries such as the United States, where at least 68% of people follow news on social media at least occasionally (Matsa & Shearer, 2018), as well as in developing countries (Mitchell, Simmons, Matsa & Silver Laura, 2018). Almost half of Spanish Internet users (48%) usually get news through Facebook (Newman, Fletcher, Kalogeropoulos, Levy & Nielsen, 2018).
Social media have also displaced the traditional, unidirectional concept of distribution by the idea of dissemination (Noguera-Vivo, 2018). In this environment, users are no longer the final end of a transmission line but a node inside a wider network (Carlson, 2016). This change means that users play an active role in spreading news by deciding what content and articles they share with their contacts on their different social media profiles.
In this sense, the importance of social media for finding news has led to the so-called 'News-Finds-Me Perception' (NFM Perception), that is, the belief that there is no need to actively follow traditional mass media to keep abreast of events, since social media provides the user with the relevant information (Gil de Zúñiga, Weeks & Ardèvol-Abreu, 2017). It has been warned that this reliance on social media has negative civic effects, as it is linked to lower levels of political knowledge and political participation (Gil de Zúñiga & Diehl, 2018; Shehata & Strömbäck, 2018).
On the other hand, due to the fast pace of change in news consumption patterns driven by social media and new technologies, we know little about the possible sharing patterns and the impact of social media on peoples' experiences of news distribution.
Previous research shows that the decision to share news items on social media is deeply related to message characteristics (Trilling, Tolochko & Burscher, 2017). Specifically, the perceived quality of content [Singer, 2014] and the perceived relevance of the topic (Ma, Sian Lee & Hoe-Lian Goh, 2014) have been highlighted as characteristics that stimulate news sharing on social media.
Message characteristics alone, however, do not fully explain the phenomenon of news sharing. To obtain a fuller picture of this behaviour, it is necessary to take into account personal traits. For example, research has shown that people tend to consume media and information that are consistent with their own views and beliefs, a phenomenon known as 'selective exposure' or 'confirmation bias' in psychology (Cappella, Kim & Albarracín, 2015). Today's news media environment, exemplified by the relevance of userdistributed content on social media, has seen the emergence of a parallel trend: 'selective sharing' or the extent to which individuals share primarily attitude-consistent content with their social networks (Shin & Thorson, 2017). According to this phenomenon, prior attitudes towards a certain topic influences the kind of media content that users share and distribute on social media (Barnidge, 2015).
Users are not only influenced by the type of media they consume but also by the type of content and concrete uses they seek through these media (McLeod, 2000). In this sense, the Uses and Gratifications Theory (U&GT) helps to explain different patterns of media behaviour and different effects (Segado-Boj, Díaz-Campo & Quevedo-Redondo, 2019). This theory posits that individuals have an innate need that can be satisfied through media usage.

Objectives and hypothesis
Previous studies have found that people relying on social media have different political views and opinions (Lee, Lindsey & Kim, 2017) and that people with heavier exposure to political information on Facebook attribute different salience to political issues (Feezell, 2018). Even though NFM Perception has been assumed to lower users' standards for the news sources they receive information from (Spohr, 2017), there is a gap in the literature regarding how dependence on social media influences the perception and evaluation of news items. Thus, the first aim of this study is to understand how NFM Perception influences this assessment of journalistic products. We suggest that users with different levels of NFM Perception will assess the quality of a news item differently.
A second objective for the research is to measure how this evaluation of the news item influences the decision to share said news article on social media. Understanding the logic of "why people share news" is a key to understanding the societal impact of social media (Singer, 2014) and how news travels (Khuntia, Sun & Yim, 2016). Granted that the quality perception of news influences the intention to share news (Ma et al., 2014), we expect that users' different qualities of assessment will have an influence on the decision to share (or not to share) the proposed new item.
The study thus seeks to evaluate how 'NFM Perception' influences the perceived relevance, accuracy and factuality of news about climate change. In addition, it measures the moderating role of uses and gratifications, news habits, prior knowledge and attitudes on climate on that influence. Finally, it aims to check whether changes on the evaluation of that piece of news increase or decrease the individual's decision to share that item on social media.
Hence, we posit hypothesis H1: 'NFM Perception influences the perceived relevance, accuracy and factuality of news on climate change'.
Additionally, the present paper explores the impact of the moderating role where by some aspects, as expounded in the introduction, affect the user's perception of news. Therefore, we decided to test two additional hypotheses: H1.1. Use & Gratification (U&G) moderates the effect of NFM Perception regarding the perceived relevance, accuracy and factuality of the news item on climate change.
H1.2. News habits on social media moderate the effect of NFM Perception regarding the perceived relevance, accuracy and quality of the news item on climate change.
As mentioned, social media not only serve as a news consumption venue but also as a secondary distribution channel. Choi (2016a) suggests that users' social media activity is differentiated between news internalizing (how people receive news on social media) and news externalizing (how users share news on social media). It has been confirmed that prior knowledge of a topic has a strong role on the decision to read news that users are incidentally exposed to on social media ( Lastly, we consider that higher perceptions of quality lead to higher chances of considering a news item.
H2. Higher perceived relevance, accuracy and factuality correlates to higher chances that a news item is shared on social media.

Justification
While much of the literature focuses on the information source and the message characteristics, it omits the receiver's role in sharing news (Shi, Lai, Hu & Chen, 2017). The present study, however, attempts to evaluate the role played by the user's assessment of a news item when deciding whether to share an item on climate change.
This research also considers the role of self-presentational concerns, which has been linked to users' selfdisclosure on Facebook of political concerns (Liu, Rui & Cui, 2017). In this context, we predict that uses and gratifications associated with self-presentation would moderate the role of NFM Perception, since such content is a strategic subject in building and projecting self-identity on the Internet (Jensen Schau & Gilly, 2003). Lastly, by considering a set of factors that includes users and message characteristics, our study takes a multifaceted approach that overcomes the limitations of monotonous approaches that preclude us from identifying how different factors influence one another (Son, Lee & Kim, 2013).

Material and methods
An online self-administered questionnaire was used, as this method allowed us to include relational and social aspects of news sharing on social media, as suggested by previous studies (Kalsnes & Larsson, 89 2018). The design of the questionnaire, the informed consent and the procedures to collect and analyse the data gathered from the participants were approved by the research ethics office in the corresponding author's University. The first page of the online questionnaire provided information about the study, to which all the participants gave their informed consent to participate by clicking "next" to begin the survey. Data were later analyzed anonymously.
The questionnaire was sent to a sample (n = 96) of undergraduate students selected from universities in the Madrid region, recruited by a social studies firm (Societae). To avoid similar academic backgrounds, the students were selected from different degree courses. To encourage participation, each participant was offered a €25 gift voucher upon successfully completing the survey. The sample comprised 61.46% women and 38.54% men, with an average age of 20.23 years (SD = 3). Young people provide a relevant subject for study, not only because younger adults are particularly active news sharers on social media but also because this news sharing is likely to remain a key feature of news distribution models (Bobkowski, 2015). The responses to the survey were collected from May 5th to July 3rd, 2017.
The participants responded to several sets of questions detailed in Tables 1-5. After those questions were submitted, participants were asked to read a news item on the effects of climate change (La Vanguardia, 2017). The text and images of the item included no reference to its source or any other context. After reading the news item, the participants were asked whether they would share that 'news item' on social media. Finally, they were asked to rate the perceived relevance of the new item.

Measurements
This section details the measures and scales used in the study. Cronbach's alpha is indicated in each Table by the symbol α. 'News-finds-me' Perception was measured by the sum of the responses to a set of four questions (Gil de Zúñiga et al., 201)] through a 7-point Likert-type scale (1 = lowest agreement, 7 = highest agreement) (see Table 1).  Table 2 shows the measurements for social media Uses & Gratifications, which were determined by the items developed by Gao and Feng (2016); each U&G was determined by the sum of several independent values, and each was measured on a 7-point Likert-type scale (1 = lowest agreement, 7 = highest agreement). In line with Hyun and Kim (2015), prior knowledge about climate change was measured as an additive index of four items (average = 1.54, SD = 0.87), obtained by summing the correct answers to four true/false questions regarding climate change (see Table 3), where each correct response was coded as 1.

Table 1. Measurements for News-Finds-Me Perception
In addition, concern about climate change was measured as a single item question (1 = "not concerned at all", 7 = "very concerned"; average = 5.81, SD = 1.27) as well as interest in news about science (1 = "not interested at all", 7 = "very interested"; average = 4.92, SD = 1.59).

Table 3. Measurements for climate change knowledge
Global temperatures were first recorded 150 years ago. Since then these temperatures have risen unremittingly (True). CO2 is the only industrial gas responsible for the greenhouse effect (False). USA has signed the Kyoto Protocol (False). Every country that signed the Kyoto Protocol has agreed to reduce its CO2 emissions by 5% (False).

Source: Own elaboration
Attitudes towards climate change were independently measured through the items shown in Table 4 (1= completely disagree, 7 = totally agree).

Source: Own elaboration
Social Media news attendance was operationalized through the scale used by Choi and Lee (2015), shown in Table 5, through a Likert type scale (1 = never, 7 = every day). Quality, understood as the perception of excellence of a news story by the public, is a complex attribute with diverse layers (Sundar, 1999). We measured three different constructs (accuracy, factuality and relevance) on a 7-pointLikert-type scale (1 = lowest agreement; 7 = highest agreement), as shown in Table 6. The news item provides accurate and exact information The news item provides fact-based information The news item provides credible information Relevance (α = 0.87) The news item provides information that should be known by the general public The news item provides information that I should know The news item provides information that should be known by political leaders Source: Authors, based on Sundar, 1999 Regarding data analysis, linear regression was chosen when the dependent variable was continuous.
Ordinal regression was applied when the dependent variable was discreet. To determine whether a variable is a moderator, the biostatistical Penn State University model was used. According to this model, when a coefficient considering a third variable changes more than 10% from the coefficient without that third variable, itis taken as a moderator. The formula applied was: To validate data, the assumption of multivariate normality has been tested by analyzing the multivariate kurtosis index and the corresponding critical ratio. A critical ratio below 5 indicates that the multivariate normality assumption is fulfilled (Bentler, 2004). In this case, the critical ratio is 1.890, which points to maximum likelihood for the estimation. However, as the sample size is small and the model includes categorical variables, a Bayesian estimation process has been followed to avoid assumptions on the a priori distribution of the estimated parameters (Byrne, 2009).
The analysis has been developed into two main stages. In the first, the initial, null, and saturated models were estimated. The first model reflects the relationship among variables. The null model assumes no relationships among the variables, thus reflecting the worst possible outcome. The saturated model assumes that every parameter is significantly different from zero, which implies full adjustment of the whole dataset (Gaviria Soto, Bienciento López & Navarro, 2009). In the second phase, the initial model was modified to improve the adjustment by discarding non-significant effects and studying the modification indexes derived.
The model adjustment was tested through the following indexes: • Incremental adjustment indexes. These indexes assess the improvements to the independence model through the CFI index. Values above 0.95 are considered good (Byrne, 2009).
• Root Mean Square Error of Approximation (RMSEA). This index compares the adjustment between the proposed model and hypothetical population data. In this case, values above 0.05 are considered very good, and those between 0.05 and 0.08, acceptable (Byrne, 2009).

Influence on the perceived accuracy of the climate change news item
As Figure 2 shows, the NFM Perception influences the perceived accuracy of the news item under study. Based on a simple ordinal regression model, the associated odds ratio was 1.1366, indicating it reached statistical significance (95% confidence interval: 1.0526, 1.2298). That means that the higher the users rate the NFM Perception, the more likely (13% for each point of the scale) it is they will perceive the quality of the news item as higher.
This is not valid for all the values of the 7-point Likert-type scale (see Figure 2), however, because users are not able to determine very precisely what "accuracy" is, since this trend is clearer for values lower and higher than 4.

Source: Own elaboration
Finally, according to the results of the present analysis, none of the considered items (uses and gratifications, users' expectations, news internalizing and externalizing, gender, size of discussion network, and climate change perception) were revealed as mediator or moderator variables in the perceived quality of a news item (see Table 7), since none of them produced a variation higher than 10%.

Influence on the perceived factuality of the climate change news item
The linear coefficient ratio between the NFM Perception and the perceived factuality of the news item also reached statistical significance, with 0.1939 (95% confidence interval: 0.0949, 02.929). This means that when users rated a higher NFM Perception, they also rated the news item with higher factuality (0.1939 points as standard average growth) (see Figure 3).

Source: Own elaboration
As Table 8 shows, seeking information, entertainment, self-expression, social interaction, impression management, news internalizing and attitudes indicating that climate change is a consequence of human action that can be halted moderates the perceived factuality of the news item. The strongest moderator influences are Information Seeking and Attitudes, indicating that climate change is a consequence of human action and therefore can be halted.

Influence on the perceived relevance of the climate change news item
As Figure 4 shows, the News-Finds-Me Perception does not correlate with the perceived relevance of the news item. The linear coefficient was 0.01; hence, it did not reach statistical significance (95% confidence interval: -0.0117, 0.1318).

Source: Own elaboration
As Table 9 shows, seeking information, entertainment, self-expression, and social interaction moderate the perceived relevance of the news item.

Influence of perceived accuracy, factuality and relevance on the decision to share the news item
No significant correlation was found between the perception of the considered news item attributes and the participant's decision to share the article.
According to the simple logistic regression model, the associated odds ratio with accuracy evaluation did not reach statistical significance, being 0.786 (95% confidence interval: 0.5740, 1.0438) (See Figure 5).

Figure 5. Relationship between the perceived accuracy of the news item and the decision to share it.
Vertical axis: Perceived accuracy of the news item. Horizontal axis: Decision to share the news item.

Source: Own elaboration
A simple logistic regression model did not reach statistical significance for the relationship between the perception of the news item's factuality and the decision to share, which scored 0.9416 with a 95% trust rate of 0.7936,1.1031 (see Figure 6).

Source: Own elaboration
A simple logistic regression model did not reach statistical significance for the odds ratio between the perceived relevance of the news item and the user's decision to shared it, which scored 1.0806 with a 95% trust rate of 0.9841, 1.1919 (see Figure 7).

Figure 7. Relationship between the perceived relevance of a news item and the decision to share it.
Vertical axis: Perceived relevance of the news item. Horizontal axis: Decision to share the news item. Figure 8 shows the final research model. Standardized regression weights are shown at Table 10. Standardized indirect effects can be seen at Table 11.   Source: Own elaboration Table 12 shows the adjustment values of the model. All the indexes assume values above the acceptable margins. The CMIN/df value is below 3, while GFI is above 0.9. This value of the CFI index indicates that the model has improved as compared to the independence model, as seen in the CFI index, which in the final model stands above 0.9. The RMSEA value also falls within the acceptable margins.

Discussion
The results show that higher NFM Perception is related to higher evaluations of accuracy and factuality. We interpret this finding as evidence that NFM lowers the user's expectations towards news items.
Information seeking has the most relevant moderating role regarding the influence of NFM on this regard (see Table 8), but when considering relevance on its own, the most important moderating role comes from social interaction (see Table 9). This could be explained by the different aspects implied in each value. As relevance measures many factors, among them an external aspect ('The news item provides information that should be known by the general public'), social interaction U&G may encourage this attitude and behaviour. On the other hand, information seeking could moderate the influence of NFM Perception because users with higher informational needs might have more criteria to evaluate the factuality of the news item, given their higher information literacy or familiarity with journalism products. Further research is needed.
Although self-presentational U&G led to higher self-disclosure in political issues (Liu et al., 2017), selfexpression and impression management play a minor moderating role compared to the most influential U&G (Information Seeking and Social Interaction).
Thus, relevance is more related to social interaction, whereas the link between self-presentational U&G and political news sharing might be better explained by personal and psychological aspects.
News internalizing plays a stronger moderating role than news externalizing, thus substantiating the suggestion that consumption habits are more relevant to the perception of news and topics than sharing habits are. In other words, users who attach greater importance to social media as a means to stay informed and receive more news through social media, tend to have different perceptions and assessments of a news story.
Attitudes towards climate change were shown to have a greater moderating role than prior knowledge did. In fact, the idea that climate change is a phenomenon provoked by humans that can be halted led to a higher factuality and relevance assessment. Prior knowledge, in contrast, did not exert a moderating role. This could imply a link between attitudes and perception towards the topic and the relevance of climate change, thus suggesting a need to share information about the topic.
The perception of responsibility as well as the possibility to halt climate change leads to different attitudes towards this topic (Chang, Kim, Shim & Ma, 2016), which as demonstrated here, also moderate the other influences regarding the willingness to share the information about this topic.
In contrast, the NFM Perception does not directly influence the relevance of the news topic. Although higher exposure to political news on social media led to higher levels of interest in political issues (Feezell, 2018), our findings suggest that such exposure must occur to be effective. In other words, the simple dependence users have on social media for receiving news does not increase users' interests or relevance given to social or political issues.
Regarding news sharing, our model shows that the assessment of the news item had almost no relevance in the decision to share or not to share it on social media. This is consistent with previous findings that maintain that topic relevance takes precedence over news credibility when users decide to share a link or news item [Ma et al., 2014]. This is especially relevant, given that NFM Perception might increase the reliance on untrustworthy sources of information, as misinformation and low quality information circulates through social media (Gil De Zúñiga, Huber & Strauß, 2018).

Conclusion
Our findings could mean that NFM Perception might affect the assessment of journalistic pieces and products, since accuracy and factuality are aspects inherent to news reports. However, topical relevance is not so dependent on the way people acquire their news.
In this sense, almost every other factor considered in our model (media habits, prior knowledge and attitudes, uses and gratifications) moderates the influence of NFM Perception in topic relevance assessment (see Table 8 and Figure 8). This points that relevance attribution is a complex phenomenon that requires models of greater complexity to be understood and that NFM Perception does not play a relevant role.
Although NFM Perception might explain differences in the perception of textual characteristics of news (accuracy, factuality), aspects related to topic relevance are not directly affected by this construct.
According to our results, users might share information regardless of its accuracy or factuality. In other words, other factors external to the journalistic quality of the news item itself might be what truly matter in news sharing. This is meaningful for understanding the way in which misleading information is distributed on social media

Limitations and further research
Cronbach`s α measurement of news internalizing and NFM Perception, ranging from 0.6 to 0.7, was below the ideal value, although it qualifies them as acceptable constructs (DeVellis, 2017).
This study used a single-message design, which does not reflect the complexity of media effects and reception. Future studies could develop a design in which multiple messages are used as a stimulus.
Additionally, given the size of the sample and the variability on the considered variables, which combine numeral and ordinal variables, it is impossible to normalize the measures and build an integral research model that could reflect marginal changes.
Regarding the sample, all the participants are undergraduate students. This implies than results could only be extrapolated to this sociodemographic sector. Regardless, young adults are a relevant subject of study given their strong orientation towards online news and news sharing on social media (Antunovic, Parsons & Cooke, 2018).
Although prior attitudes and knowledge were measured, prior general interest towards the topic was not surveyed. This construct has been shown to be a motivating factor for engaging with news (Kümpel, 2018); therefore, future research should incorporate it to better understand the role of NFM Perception in news assessments.
As already mentioned, attitudes towards the topic were shown to play a relevant moderating role. Future studies should respond to or adapt the study to ascertain whether the identified tendencies appear in other socially relevant or controversial news topics.
We also mentioned that psychological aspects may help to improve our understanding of the link between the willingness to share information and uses and gratifications, which has been observed in studies on the effects of media (Ngai, Tao & Moon, 2015). Considering some psychological features and constructs such as the big five personality traits might help to gain further insight into the current topic.

Funding
This work was supported by the Spanish Ministry of Economy and Competitiveness, the Spanish Research Agency and the European Regional Development Fund, under grant CSO2017-86312-R. Ministerio de Economía y Competitividad [CSO2017-86312-R].