The Human Alarm System for Sensational News,Online News Headlines,and Associated Generic Digital Footprints: A Uses and Gratifications Approach

Abstract

By adopting the uses and gratifications approach to understand two evolutionary needs—the environmental surveillance need and social involvement need—this study investigated the use of alarm and prosocial words in news headlines and the associated generic digital footprints. We analyzed over 170,000 online news headlines and the number of associated clicks and “likes” for each news story on an online news platform. Our results support the idea of a human alarm system for sensational news as a psychological survival mechanism designed to detect and pay attention to threatening news such as catastrophes and diseases. News headlines with alarm words indirectly attracted more “likes,” indicating a concern with survival, through an increased number of clicks to select that news item. Furthermore, the results of a conditional indirect effect model showed that while online readers selectively clicked on news headlines with alarm words, the presence of a prosocial word in the headline increased the likelihood that readers would “like” it.

Keywords

Human alarm system News headlines Online news platform Prosociality Sensational news Uses and gratifications

News media exist to provide not only social, cultural, and economic information, but also sensational information such as news about accidents, diseases, and disasters. Individuals are attracted to sensational news because such events imply potential threats to the survival and reproduction of their genes (Davis & McLeod, 2003; Shoemaker, 1996). Sensational news satisfies a basic human instinct and desire to detect threats in the surroundings and around the world and find solutions to real and potential threats (Shoemaker, 1996). We found three gaps in the research literature on sensational news, and this study makes theoretical and methodological contributions accordingly.

First, the value of sensational news has not yet been specified. In communication studies, the question, “How do events become news?” has been a central concern of journalists since Galtung and Ruge (1965). Journalists’ selection of news depends on the news values that decide the story’s newsworthiness (Galtung & Ruge, 1965; Harcup & O’Neill, 2001, 2017). Because of the fundamental importance of news values, journalism scholars have developed different newsworthiness models to conceptualize their underlying factors (e.g., Galtung and Ruge, 1965; Schultz, 2007; Shoemaker & Reese, 2014). No one newsworthiness model can explain news values perfectly and comprehensively, but all of the models include one type of news: sensational news. While the definition and terminology of sensational news may be slightly different across these models, journalists have acknowledged the value of sensational news over time, although sensationalism has been surrounded by negative stigma (Kilgo, Harlow, García-Perdomo, & Salaverría, 2016). However, journalists may not understand why sensational news is attractive from an evolutionary point of view. Because of this, this study investigates sensational news selection by the audience and examines the specific aspect or element that leads sensational news to be considered “valued” by the journalists who select news.

Second, prosocial acts, such as help, assistance, donations, and charitable giving, are an important topic in disaster news research. Previous studies have focused on the direct media effect of disaster news attention, exposure, and involvement on the intention of audiences to donate (e.g., Bennett & Kottasz, 2000; Martin, 2013; Oosterhof, Heuvelman, & Peters, 2009; Seo, Sun, Merolla, & Zhang, 2012). However, it is unclear from the selective media effect whether sensational news with prosocial acts attracts more attention than sensational news alone. Sensational news that specifically includes prosocial acts may be valuable.

Third, sensationalism in traditional media, including television (e.g., Arbaoui, De Swert, & van der Brug, 2016; Hendriks Vettehen, Nuijten, & Beentjes, 2005; Hendriks Vettehen, Nuijten, & Peeters, 2008) and newspapers (Davis & McLeod, 2003), has been widely studied using traditional analytic techniques such as surveys, experimental designs, and content analysis. Little is known, however, about the selection of sensational news on online news platforms where users see only the headlines of each news story but must click on each to read the story. This study investigated whether the human alarm words and prosocial words in news headlines could predict the number of clicks and “likes” a news story would receive.

Following Davis and McLeod (2003), sensational news was defined as news related to death, disaster, disease, and accidents that affect the reproductive fitness of human beings. We proposed that people are more likely to be concerned with sensational news because human beings have a psychological mechanism designed to pay attention to news of life-threatening events such as catastrophes and diseases that could influence the survival and reproductive success of their genes.

Evolutionary Needs: A Uses and Gratifications Approach

The theory of uses and gratifications explains the social and psychological origins of needs that drive media exposure and selection to fulfill individuals’ needs for gratification (Katz, Blumler, & Gurevitch, 1973, 1974). This approach holds that individuals are active and goal-oriented media users, their motivations explain media exposure, and they select media to fulfill their motives (Blumler, 1979; Katz, Haas, & Gurevitch, 1973). Rooted in this approach, the concept of the need for orientation offers a psychological explanation for why individuals seek information from media—people need to become conversant with the surrounding environment through news media (Matthes, 2005; Weaver, 1980). In particular, humans pay attention to the surroundings that involve potential threats to their lives (Davis & McLeod, 2003; Hendriks Vettehen et al., 2008; Shoemaker, 1996). Shoemaker (1996) argued that the surveillance function of news can best be explained with an evolutionary perspective. Apart from the need for orientation, the need for social involvement, which is also rooted in the uses and gratifications approach, represents a motivation for showing sympathetic and cooperative tendencies that can also be satisfied by the consumption of news (Ruggiero, 2000; Valenzuela, Bachmann, & Aguilar, 2016). Numerous studies have shown that from the evolutionary point of view, it is human nature to be good to others (e.g., Goetz, Keltner, & Simon-Thomas, 2010; Keltner, Kogan, Piff, & Saturn, 2014; Kurzban, Burton-Chellew, & West, 2015; McAndrew, 2012; Penner, Dovidio, Piliavin, & Schroeder, 2005; Tomasello, 2014).

This study intends to examine these two evolutionary needs as displayed through interaction with online news media: the environmental surveillance need and the social involvement need. News exposure satisfies the need to know and understand what happens in the world and the need to show sympathy and empathy with others (Ruggiero, 2000; Valenzuela et al., 2016). From the evolutionary perspective, people understand potential threats in the world so they can avoid them and demonstrate compassion and altruism from news consumption so they can receive direct and indirect benefits by promoting cooperative signaling and maintaining good reputations, both of which are ultimately aimed at maximizing the survival chances of humans’ genes (Boyd & Richerson, 1989; Kurzban et al., 2015; Shoemaker, 1996; Trivers, 1971). Need fulfillment through news media use, according to the uses and gratifications theory, leads to gratification. The two fundamental evolutionary needs can be gratified through media engagement. In this study, we focus on media engagement, including clicks and likes on an online news platform.

Environmental Surveillance Need: Human Alarm System

Humans need a psychological alarm system to detect and handle alarming and threatening situations (Eisenberger & Lieberman, 2004; Eisenberger, Lieberman, & Williams, 2003; van den Bos et al., 2008). The alarm system detects threat cues so that individuals can respond quickly and increase the survival chances of their genes. For example, social exclusion causes social pain and triggers the human alarm system, which alerts people to pay attention to the prolonged damage caused by social disconnection (Eisenberger & Lieberman, 2004; Eisenberger et al., 2003). This alarm system is evolutionarily adaptive and built into the biology of modern humans.

Human beings are concerned about events in their environment that are unusual and lethal, such as disaster and disease, because these threats significantly affect the chance of survival and reproduction of their genes. Sensational news stories about negative events, such as accidents, crime, fires, and natural disasters, satisfy the human survival instinct to detect environment threats and allow a quick reaction to deal with a potential or real threat (Davis & McLeod, 2003; Hendriks Vettehen et al., 2008; Shoemaker, 1996). The propagation of sensational news would have provided adaptive functions for our ancestors to detect predators, making their genes more likely to gain a survival and reproductive advantage in the same population. After millions of years, humans now have biological and surveillance mechanisms designed to pay attention to sensational news content (Shoemaker, 1996). Sensational news has attracted the concern of human beings from 1700 to the 20th century (Davis & McLeod, 2003).

The sensational news alarm system model that explains the environmental surveillance need is supported by a few empirical studies. Sensational content potentiates selective exposure (Stanca, Gui, & Gallucci, 2013). For example, sensational television news, such as stories about death, disasters, riots, fires, and terrorism, moderately enhance the emotional arousal of viewers, which in turn attracts their attention to like these television news stories (Hendriks Vettehen et al., 2008). Using intensifiers (e.g., “extremely dangerous” for “dangerous”) in sensational printed news articles directs readers to regard the news as more important and to pay greater attention to such life-threatening events (Burgers & de Graaf, 2013). Thus, this study proposed that human alarm-related news represents sensational news.

The negativity bias also plays a role in explaining why people are attracted to sensational news. Species that are more responsive to bad events are more likely to avoid threats, thus maximizing the survival and reproduction of their genes throughout evolutionary history. Thus, it is evolutionarily adaptive for humans to be more likely to pay attention to negative events than positive events (Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001; Ito, Larsen, Smith, & Cacioppo, 1998; Vaish, Grossmann, & Woodward, 2008). Bias toward potential threats in the world in the form of sensational news serves the evolutionarily adaptive purpose of helping people to securely explore the environment while avoiding threatening situations.

Social Involvement Need: Prosocial Tendencies

News consumption can satisfy the social involvement motivation: the need to connect to other individuals (Ruggiero, 2000; Valenzuela et al., 2016). The most effective way to maintain relationships with others is to demonstrate prosociality. An understudied question worth investigating is whether prosociality is involved in sensational news selection. Prosociality is defined as the attitudes, beliefs, behavior, traits, and values people direct toward the benefit of others, such as altruism, cooperation, donation, reciprocity, and helping (Keltner et al., 2014; Penner et al., 2005). For example, do readers pay more attention to news stories about people helping the victims of a disaster?

Empirical studies have shown that news attention, exposure, and involvement are positively related to prosocial acts such as donations of money. For instance, if TV and newspaper reports of a catastrophe show quick and efficient responses from aid agencies and photos of victims looking pathetic, people are more likely to make a donation to an emergency relief appeal (Bennett & Kottasz, 2000) because reports of innocent victims during natural disaster coverage increases their sense of intimacy with their significant others and their social trust of strangers (Seo et al., 2012). People who followed news coverage specifically about the 2004 Indian Ocean earthquake and tsunami, the 2010 Haiti earthquake, and the 2011 Japan earthquake, tsunami, and nuclear meltdown were more likely to plan to make a donation (Martin, 2013).

According to the selective exposure self- and affect-management (SESAM) model (Knobloch-Westerwick, 2015), the motivation for selective media exposure in combination with media influence directs people’s selective exposure, so that individuals are dynamically influenced by their exposure and responses to selected media messages. People react to the sensational news they select by showing concern and donating money. Their responses may dynamically shape their selection of sensational news containing prosocial acts. As a result, while the above empirical research has shown that the direct and indirect effect of disaster news on prosocial behavior is that people perform charitable acts such as donating after exposure to disasters news, our hypotheses, in line with the SESAM model, examines whether people are more likely to pay attention and show concern to sensational news that contains prosocial acts.

From the evolutionary perspective, the ultimate functions of prosociality and the human alarm system are the same—to gain survival and reproductive benefits. Human beings are equipped with psychological mechanisms designed to help both relatives who carry their genes and nonrelatives who can offer delayed benefits through reciprocal cooperation (Hamilton, 1964a, 1964b; Kurzban et al., 2015; Trivers, 1971). To explore this research question, this study conceptualized an overlap between prosociality (Keltner et al., 2014; Ng, 2016; Penner et al., 2005), the human alarm system (Eisenberger & Lieberman, 2004; Eisenberger et al., 2003; van den Bos et al., 2008), and interest in sensational news to investigate whether prosociality and the alarm system intensify the psychological mechanism for detecting threatening cues.

Online News Headlines and Generic Digital Footprints

This study analyzed news headlines in terms of the number of clicks and “likes” for each news story on an online news platform where users could “like,” share, and comment on news stories. News headlines have important functions in news communication (Dor, 2003; Ecker, Lewandowsky, Chang, & Pillai, 2014; Ifantidou, 2009). Rather than reading all of the news stories, modern newspaper readers often scan the headlines and then read only the stories they are interested in. Dor (2003) therefore proposed that the communicative function of newspaper headlines is to optimize the relevance of the information for readers. The headline acts as a selection device to guide them to invest cognitive effort to receive additional information on those stories they have a special interest in (Dor, 2003). Based on the proposed human alarm system for sensational news, alarm-related news headlines should drive readers to apply cognitive effort by clicking news headlines to obtain information related to threats to their survival. We therefore analyzed headlines that would increase reproductive fitness in an environment of evolutionary adaptedness, including news about accidents and natural disasters (Davis & McLeod, 2003).

We operationalized “likes” as survival concern related to a particular news story. Facebook provides “likes” as a generic digital footprint for users to show enjoyment from or interest in something without leaving a comment (Gan, 2017). Twitter uses “likes” (a small heart) to express appreciation. Previous studies have used “likes” to predict personal attributes and personality traits (Kosinski, Stillwell, & Graepel, 2013; Youyou, Kosinski, & Stillwell, 2015), as a social endorsement of online peer influence (Sherman, Payton, Hernandez, Greenfield, & Dapretto, 2016), and as an online gift exchange that serves as an impression-management strategy (Hong, Chen, & Li, 2017). Youyou and colleagues (2015) argued that social media users use “likes” to express a relationship between online and offline objects. For example, liking a music video offers a proxy for the user’s music preferences and purchasing behavior. Like-based models offer a proxy of what people care about based on different digital footprints, such as clicking a news headline. As a result, clicks and “likes” can represent a proxy for survival selection and concern that can be linked to offline adaptive behavior after reading alarm-related news online. According to the human alarm system for sensational news, we predicted that after reading an alarm-related news headline, individuals would expend cognitive effort to read the article about the threat and then “like” the article, signaling survival concern.

In the realm of online word use, the increasing availability of online data has provided researchers with the opportunity to investigate large amounts of data about psychological states and behavior online. For example, negative words used to express anger and anxiety on social media were found to predict heart disease mortality (Eichstaedt et al., 2015), and future-oriented words used, such as “will,” “could,” and “gonna,” were shown to predict lower HIV prevalence (Ireland, Schwartz, Chen, Ungar, & Albarracín, 2015).

The main purpose of this study was to empirically examine human alarm words and prosocial words in online news headlines and the associated generic digital footprints. Paying attention to words, such as “accident,” “disaster,” and “death” in news headlines (i.e., sensational news), can enhance the chance of survival and reproductive success of one’s genes (Davis & McLeod, 2003). Words relating to helping, caring, and cooperation imply the adaptive human mechanism of altruistic behavior (Kurzban et al., 2015). In addition, considering the communicative function of newspaper headlines as relevance optimizers (Dor, 2003), the sensational news alarm system should be activated after the investment of cognitive effort. By overlapping the above concepts using both theory- and data-driven methods, the present study tested the following hypotheses:

Hypothesis 1a (H1a): People will be more likely to read a news article if the headline contains alarm words.

Hypothesis 1b (H1b): People will be more likely to “like” a news article if the headline contains alarm words.

Hypothesis 1c (H1c): People will be more likely to read and then “like” a news article if the headline contains alarm words.

Hypothesis 2 (H2): News headlines with prosocial words will moderate the indirect effect of alarm-related headlines on the number of “likes” via the number of clicks, such that headlines with alarm words will attract more likes than headlines without alarm words indirectly by enhancing the number of clicks when headlines also have prosocial words compared with headlines without prosocial words.

Method

Data Source

We chose TouTiao.com (Headline News), a news aggregator and dissemination platform based in Beijing, China. TouTiao employs editors to rewrite the headlines of news stories and makes some of the data available for crawling, including the data we used for this study: headlines, news categories, number of clicks (clicks), and number of likes (likes). Users first see the news headlines on the web page, and if they are interested in a news story, they can click (i.e., clicks) that news headline to read the full news content. Users can like, dislike, comment, and share the news after reading the full news content. They can also like, dislike, and share the news after reading the headlines only without clicking on to read the full story. Content from such a site cannot be a representative sample of all content or all cyber behavior of all Chinese; however, the data we analyzed captures all of the content on a major news site for nearly a month and the behavior of 250 million users (for details of data source and crawling, see the Supplemental Method in the Online Supplemental Materials).

We crawled the data from January 27, 2015 to April 30, 2015. Overall, we collected about 220,000 online news headlines. After removing duplicates, we had 171,846 headlines. Every online news headline contained the name of the headline and the number of clicks, likes, dislikes, shares, and comments. The distributions of the generic digital footprints were positively skewed: clicks (skewness = 20.59, SE = .006), shares (skewness = 59.48, SE = .006), likes (skewness = 58.43, SE = .006), dislikes (skewness = 45.08, SE = .006), and comments (skewness = 62.92, SE = .006). Generic digital footprints that were more than 6 SD away from the mean were excluded, resulting in a final sample of 170,503 online news headlines.

Coding

We coded the main variables, including the moderating variable, prosocial words, and the independent variable, alarm words, from each online news headline, with the exception of the mediating variable, clicks (i.e., the number of clicks on each headline), and the dependent variable, likes (i.e., the number of likes for each news story). We included punctuation marks in the news headlines, news type (soft vs. hard news), and title length as the control variables.

Prosocial words dictionary

The Linguistic Inquiry and Word Count (LIWC) closed vocabulary approach has been widely used in the behavioral and social sciences to analyze words in a given text (Tausczik & Pennebaker, 2010) and, recently, to analyze word use in tweets (e.g., Dehghani et al., 2016; Ireland et al., 2015). However, prosocial words are not an existing category in LIWC. Therefore, we created the Prosocial Words Dictionary in our study to investigate the function of prosocial words online.

We developed the Prosocial Words Dictionary with reference to the first few steps of the LIWC 2015 dictionary development (Pennebaker, Boyd, Jordan, & Blackburn, 2015) and the Moral Foundations Dictionary (Graham, Haidt, & Nosek, 2009). Based on the previous reviews on prosociality (Goetz et al., 2010; Keltner et al., 2014; Ng, 2016; Penner et al., 2005; Zaki, 2014), we used the following prosocial words as a starting point to generate the dictionary development: care, help, donation, empathy, compassion, sympathy, moved, tender, softhearted, touched, pity, gratitude, forgiveness, altruism, trust, reciprocity, generosity, cooperation. Then, two judges individually looked up these 16 prosocial words in three dictionaries: Cambridge English Dictionary (http://dictionary.cambridge.org/), Merriam-Webster Dictionary (http://www.merriam-webster.com/), and Oxford English Dictionary (http://www.oxforddictionaries.com/). After generating all of the prosocial words, the two judges discussed and compared the results. All of the disagreements were resolved after discussion. English words in the completed Prosocial Word Dictionary are displayed in online supplemental Table S1.

Following Frimer, Schaefer, and Oakes (2014), to validate the English Prosocial Words Dictionary, we invited 35 MTurk workers who received US$0.4 in return for their participation to describe three prosocial goals, three proself goals, and three future goals. We excluded 19 MTurkers whose responses were unusable (e.g., only numbers were reported). We then conducted content analysis on the remaining 16 participants’ English goal texts. The results of repeated measures ANOVA showed that prosocial goal prompts generated about 10 times higher prosocial word density (M = 10.9%, SD = 13.8%) than proself goal prompts (M = 1.4%, SD = 2.7%) and future goal prompts did (M = 1.4%, SD = 2.2%), F(2, 30) = 8.25, p = .001, η² = .35.

We translated the English prosocial words into Chinese. We tried our best to ensure the construct validity of the Chinese Prosocial Words Dictionary. All of the authors and coders were native Chinese speakers who were good to native English speakers and expert users. We were aware that the English and Chinese languages differ in conceptualizing certain concepts. Thus, during the process of translation, we were careful to make sure that the terms had the same connotations in both English and Chinese.

We used the Cambridge English Dictionary (http://dictionary.cambridge.org/dictionary/english-chinese-traditional/) and Google Translate (https://translate.google.com/?hl=zh-TW#en/zh-TW) to translate the English terms into Chinese. We realized that Google Translate might generate poor and unintelligible translations in the target language (although our experiences tell us that Google Translate is quite good at translating English to Chinese). As a result, we also used another dictionary, the Cambridge English Dictionary, for the translation. Also, all of the Chinese authors and coders with good to excellent English competency checked the translation procedure to make sure that the translations made sense in Chinese.

Verb conjugation is not an issue in the translation in our study. Unlike English, Chinese has no verb conjugations (no irregular verbs), no tenses (the verb form never changes in function of the tense), no plurals (there are clear quantifiers before each noun), and no declensions of adjectives (adjectives never change). Chinese grammar is relatively uncomplicated and easy to deal with in translation.

We recruited 18 Chinese participants through personal networks who received RMB$5 (US$0.78) in return for their participation to validate the Chinese Prosocial Words Dictionary. They were asked to describe three prosocial goals, three proself goals, and three future goals. Content analysis of the participants’ Chinese goal text indicated that prosocial goal prompts yielded much higher prosocial word density (M = 13.0%, SD = 11.9%) than proself goal prompts (M = 0.6%, SD = 2.6%) and future goal prompts did (M = 0.4%, SD = 1.5%), F(2, 34) = 21.87, p < .001, η² = .56.

The Chinese-English translation of the Prosocial Words Dictionary is illustrated in the Online Appendix A. Online supplemental Table S2 shows the Chinese Prosocial Words Dictionary. A prosocial word was coded as 1 if it was in a headline and 0 otherwise, and all of the values were summed. We created a dummy moderating variable, prosocial words, for the group of all headlines with prosocial words.

To validate our classification of prosocial versus nonprosocial headlines, we randomly selected approximately 2,000 headlines. Two coders then independently judged whether each headline was prosocial or not, and then all disagreements were resolved (n = 980 headlines with prosocial words, n = 970 headlines without prosocial words; percent agreement = .91; kappa value = .81). Our headlines coded prosocial were then compared with headlines with prosocial words based on the Prosocial Words Dictionary, and the results (good to very good agreements) indicated that our coding was justifiable (percent agreement = .88; kappa value = .77).

Human alarm words in headlines

Davis and McLeod (2003) found that, from 1700 to 2001, newspaper stories related to death, natural injury (i.e., disaster and disease), and accident were the most prevalent type of news; these sensational stories have human alarm cues regarding threats to survival. We used the Cambridge English Dictionary (http://dictionary.cambridge.org/dictionary/english-chinese-traditional/) to translate the following human alarm words from English to Traditional Chinese/Simplified Chinese: death, disaster/catastrophe, disease, accident (see online supplemental Table S3). We separately coded each human alarm word as 1 if it was in a headline and 0 otherwise, and we summed all of the values to create the independent variable alarm words.

We also validated our classification of sensational headlines in our sample of randomly selected headlines. Two independent coders decided whether the headlines were sensational or not, and then all disagreements were resolved (n = 633 for headlines with alarm words, n = 1,317 for headlines without alarm words; percent agreement = .90; kappa value = .74). A comparison of the headlines we coded as sensational with sensational headlines with alarm words showed that our coding was justifiable (good to very good agreements: percent agreement = .90; kappa value = .75).

Punctuation and news type in the headlines

We identified all of the punctuation marks that we observed in the online news headlines. Online supplemental Table S4 shows the frequency of all of the observed punctuation marks. We excluded punctuation marks that occurred fewer than 1,000 times from further analysis. Punctuation marks, including exclamation points, question marks, commas, Chinese commas, colons, parentheses, English quotation marks, guillemets, and hyphens were coded as 1 if they were present in a headline and 0 otherwise, and they were treated as control variables.

We created another control variable, news type, and coded soft news (i.e., regimen news, entertainment news, car news, sport news, fashion news, game news, travel news, baby news, stories, food, essay) as 0 and hard news (i.e., society news, world news, finance news, technology news, military news, discovery news, history news) as 1. The third control variable was title length.

Analysis

Because the data set is large compared with traditional data sets in social sciences studies, we adopted a split/analyze/meta-analyze approach for data analysis (Cheung & Jak, 2016). This approach is necessary because other approaches are inferior. The first approach is to manage and analyze big data as a typical data set. However, this approach rarely does well because of problems, including overfitting, picking up patterns by chance, creating noise in the fit data, and showing a lack of generalizability to unseen data. A second approach is to randomly select and analyze a subset of data. However, it is better to include as much relevant data as possible when conducting scientific research. Also, the results based on the full data may be different from those based on a subset of the data. A third approach is to combine the data set based on characteristics such as geographic location. Researchers analyze the aggregated means of the data set rather than the raw data set. This approach has the limitation that the results based on the aggregated data set may not be the same as results based on the real data (Cheung & Jak, 2016).

As our data set was too big to be analyzed directly, we broke the data into smaller pieces for parallel analyses. We first split our data set into independent data sets. We applied a random split to the data set because no special sample characteristics were identified. We randomly and arbitrarily split the data set into four subsamples. The four data sets can be treated as four separate pseudostudies. Because each pseudostudy was similar to traditional data sets, we could conduct typical analyses such as Pearson’s correlation and multiple regressions on each study. Because each pseudostudy with similar characteristics was split randomly, we were able to conduct similar analyses on each sub-data set without affecting the random errors of our results. As a result, we conducted the same Pearson correlation and multiple ordinary least squares regression analyses for these four subsamples.

After getting the effect sizes such as Pearson’s r from each pseudostudy, we integrated the results of the four studies together using the meta-analytic method. The advantage of meta-analysis is that compared with the above approaches, the results of analyzing raw data are approximately equal to those of analyzing meta-analyzed data (Olkin & Sampson, 1998). We thus meta-analyzed the four subsamples. Microsoft Excel (Neyeloff, Fuchs, & Moreira, 2012) and JASP version 0.8.4 (JASP Team, 2017) were used to conduct the simple meta-analyses. The descriptive statistics for all of the variables are presented in Table 1.

Table 1.

Descriptive Statistics Across Samples.

Variable	Sample 1	Sample 2	Sample 3	Sample 4
Sample size (N)	42,404	41,887	43,138	43,074
Alarm words
Yes	2.9%	3.3%	3.0%	2.9%
No	97.1%	96.7%	97.0%	97.1%
Prosocial words dictionary
Yes	3.0%	2.7%	2.9%	2.8%
No	97.0%	97.3%	97.1%	97.2%
Punctuations^a
Yes	56.5%	57.2%	57%	56.7%
No	43.5%	42.8%	43%	43.3%
News type
Soft news	58.9%	58.9%	58.7%	59.6%
Hard news	41.1%	41.1%	41.3%	40.4%
Title length
M_length	18.12	18.16	18.21	18.15
SD_length	5.67	5.66	5.66	5.66
Clicks
M_clicks	8,665.81	8,356.41	8,150.06	8,550.95
SD_clicks	35,404.36	33,943.21	31,932.78	34,890.82
Likes
M_likes	20.17	19.19	19.05	20.28
SD_likes	104.07	102.20	98.97	111.06
Dislikes
M_dislikes	14.18	13.18	12.77	13.35
SD_dislikes	85.15	81.35	77.86	81.99
Shares
M_shares	106.47	106.57	107.52	111.33
SD_shares	412.11	413.18	423.34	433.81
Comments
M_comments	20.91	17.13	17.76	20.92
SD_comments	224.78	182.41	185.73	221.21

For punctuation marks among samples, see supplemental Table S4 for detail.

To test H1a and H1b, we conducted a Pearson’s correlation to find the relationship between headlines with alarm words and the two digital footprints (i.e., clicks and likes) on each of the four subsamples. We then meta-analyzed the Pearson’s rs. To examine H1c, we used Hayes’s (2013) PROCESS Model 4 to test the indirect effect of alarm words on likes through clicks. To examine H2, we used Hayes’s (2013) PROCESS Model 14 to test our hypothesized conditional indirect effect of alarm words on likes indirectly through clicks at different levels of prosocial words (i.e., headlines with vs. without prosocial words). All of the samples were repeatedly resampled 50,000 times.

To examine the magnitude of the indirect effects (regarding H1c) and to compare the magnitude of the indirect effects of headlines with prosocial words with those of headlines without prosocial words (regarding H2), the mediation effect size should be reported (Preacher & Kelley, 2011). We followed Kenny’s (2018) suggestions to determine the small, medium, and large effect sizes of the indirect effect. Cohen’s (1992) recommendations of effect size were 0.1 for small effect, 0.3 for medium effect, and 0.5 for large effect. As an indirect effect is a product of two effects (i.e., a × b), Kenny suggested that the effect sizes of the indirect effect should be squared: 0.01 for small effect, 0.09 for medium effect, and 0.25 for large effect.

In this study, we reported a partially standardized indirect effect (c’ ps) as the effect size for indirect effects. A partially standardized indirect effect is the ratio of the indirect effect to the standard deviation of Y. This index indicates the size of the indirect effect in terms of standard deviation units in Y (MacKinnon, 2008).

We decided to report this index because the circumstances of other effect size measures for mediation effect are inconclusive. Preacher and Kelley (2011) suggested κ² is the mediation effect size but Wen and Fan (2015) argued that κ² may not be an appropriate mediation effect size measure because of its nonmonotonicity; they suggested using the ratio of the indirect effect to the total effect (proportion mediated [P_M]) to represent the mediation effect size. However, because of its unboundedness, they acknowledged that P_M may not be a suitable mediation effect size measure for inconsistent mediation models in which the direct effect and the indirect effect have opposite signs (Wen & Fan, 2015). As our results showed competitive mediation models in which the mediation effect and the direct effect pointed in different directions (Zhao, Lynch, & Chen, 2010), we decided not to choose P_M in this study. Due to the inconclusive circumstances, we instead meta-analyzed the partially standardized indirect effect as the mediation effect size and the index of moderated mediation to assess the conditional indirect effects using the Hedges method. Furthermore, as each subsample was still very large, we adjusted the threshold p value (p < .001 for acceptance) and reported the effect sizes and confidence intervals (CIs) to enhance the credibility of the large samples (Lin, Lucas, & Shmueli, 2013). Materials, data, and analyses syntaxes are available online at https://osf.io/w9y37/?view_only=21941ec0b7a742c590ae90f245012517

Results

H1

To test H1a, we examined the relationship between headlines with alarm words and the number of clicks. The results showed that people were more likely to click on the headlines to read a news article if the headline contained alarm words, r _{meta-analyzed} = .0231, SE = .0024, p < .001, 95% CI = [.0183, .0278].

To test H1b, we tested the relationship between headlines with alarm words and the number of likes. The results indicated that people were more likely to “like” a news article if the headline contained alarm words, r _{meta-analyzed} = .0114, SE = .0032, p < .001, 95% CI = [.0051, .0177].

For H1c, we examined the indirect effect of alarm words on likes via clicks, controlling for the nine punctuation marks, news type (soft news vs. hard news), and title length. We estimated the mediation models through bootstrapping by resampling Samples 1 to 4 50,000 times each. The results indicated that people were more likely to read and then “like” a news article if the headline contained alarm words, c’ ps _{meta-analyzed} = .0500, SE = .0140, p < .001, 95% CI = [.0226, .0775].

We found that headlines with prosocial words were not correlated with the number of clicks, r _{meta-analyzed} = –.0076, SE = .0024, p = .002, 95% CI = [–.0124, –.0029], nor the number of likes, r _{meta-analyzed} = .0077, SE = .0024, p = .002, 95% CI = [.0029, .0124].

H2

To examine H2, we statistically tested whether prosocial words simultaneously moderate the path from alarm words to clicks, the path from alarm words to likes, and the path from clicks to likes. We found significant Clicks × Prosocial words interactions on likes for all four samples (ps < .001). We did not find other significant interactions: Alarm words × Prosocial words on clicks (p = .34/.90/.21/.11; Samples 1 to 4) and likes (p = .34/.40/.40/.005).

To ensure a parsimonious model, we removed the insignificant Alarm words × Prosocial words on clicks and likes for further analysis. We used PROCESS Model 14 to examine this condensed moderated mediation model (see Table 2). Prosocial words were found to strengthen the relationship between clicks and likes. The positive relationship between clicks and likes was stronger for headlines with prosocial words, Unstandardized B = .0027/.0028/.0033/.0032, ps <.001, 95% CI = [.0025, .0028]/[.0026, .0030]/[.0031, .0034]/[.0030, .0034], than for headlines without prosocial words, Unstandardized B = .0019/.0020/.0020/.0021, ps <.001, 95% CI = [.0019, .0020]/[.0019, .0020]/[.0020, .0020]/[.0021, .0021].

Table 2.

Regression Analysis of Clicks and Likes After Controlling for the Covariates Across Samples.

Predictor	Clicks				Likes
Predictor	β	B	t	95% CI	β	B	t	95% CI
Sample 1
Alarm words	.03	1756.95	5.29***	[1105.94, 2407.97]	−.01	−1.19	−1.64	[−2.61, 0.23]
Clicks					.66	0.002	180.84***	[0.002, 0.002]
Prosocial words					−.0003	−0.21	−0.09	[−4.63, 4.21]
Clicks × Prosocial words					.04	0.001	11.41***	[0.001, 0.001]
R²	.004				.45
Sample 2
Alarm words	.02	983.64	3.35***	[408.32, 1558.95]	−.01	−1.34	−2.01*	[−2.65, −0.03]
Clicks					.65	0.002	174.13***	[0.002, 0.002]
Prosocial words					.003	1.64	0.67	[−3.12, 6.40]
Clicks × Prosocial words					.04	0.001	9.05***	[0.001, 0.001]
R²	.003				.43
Sample 3
Alarm words	.02	1227.36	4.24***	[660.56, 1794.16]	−.01	−1.58	−2.34*	[−2.91, −0.26]
Clicks					.64	0.002	175.15***	[0.002, 0.002]
Prosocial words					−.01	−3.34	−1.51	[−7.68, 1.00]
Clicks × Prosocial words					.06	0.001	17.01***	[0.001, 0.001]
R²	.004				.43
Sample 4
Alarm words	.02	1570.99	5.02***	[957.67, 2184.31]	.0005	0.10	0.13	[−1.36, 1.56]
Clicks					.66	0.002	180.21***	[0.002, 0.002]
Prosocial words					.001	0.54	0.22	[−4.37, 5.46]
Clicks × Prosocial words					.05	0.001	12.20***	[0.001, 0.001]
R²	.003				.44

Note. See online supplemental Table S5 to Table S8 showing all covariates. CI = confidence interval.

p < .05. ***p < .001.

We estimated the moderated mediation models through bootstrapping by resampling Samples 1 to 4 50,000 times each (see Table 3). The indexes of moderated mediation in two out of the four samples were significant within a 95% CI (three out of four within a 90% CI).

Table 3.

Conditional Indirect Effects of Alarm Words on Like Through Clicks at Two Conditions of Prosocial Words After Controlling for the Covariates Across Samples.

Sample	Index of moderated mediation	Boot SE	Boot 95% CI	Boot 90% CI	Indirect effect
Sample	Index of moderated mediation	Boot SE	Boot 95% CI	Boot 90% CI	With prosocial words	Boot 95% CI	Without prosocial words	Boot 95% CI
1	1.25	0.98	[−0.07, 4.03]	[0.12, 3.47]	4.67	[2.14, 8.79]	3.42	[1.63, 5.62]
2	0.86	0.90	[−0.21, 3.47]	[−0.06, 2.95]	2.78	[1.04, 6.13]	1.92	[0.78, 3.29]
3	1.56	0.94	[0.21, 4.14]	[0.40, 3.70]	4.01	[1.67, 7.84]	2.45	[1.00, 4.19]
4	1.75	1.01	[0.35, 4.54]	[0.55, 4.03]	5.03	[2.34, 9.24]	3.28	[1.56, 5.51]

Note. Each conditional indirect effect was repeatedly resampled 50,000 times. CI = confidence interval.

We meta-analyzed the index of moderated mediation to assess the indirect effect of alarm words on likes via clicks for the two prosocial word conditions across the four samples. We then compared the mediation effect size (i.e., partially standardized indirect effect, c’ ps) and the meta-analyzed indirect effect of headlines with prosocial words with headlines without prosocial words.

The meta-analyzed index of moderated mediation was significant. Headlines with alarm words generated more likes than headlines without alarm words indirectly by increasing the number of clicks when headlines also had prosocial words (c’ ps _{meta-analyzed} = .0613; indirect effect _{meta-analyzed} = 3.90, SE = .75, 95% CI = [2.43, 5.38]) compared with headlines without prosocial words (c’ ps _{meta-analyzed} = .0392; indirect effect _{meta-analyzed} = 2.52, SE = .41, 95% CI = [1.72, 3.32]), index of moderated mediation _{meta-analyzed} = 1.33, SE = .48, 95% CI = [.40, 2.26].

Because the distribution of the generic digital footprints were positively skewed, we also log-transformed clicks and likes without removing the outliers that were more than 6 SD away from the mean. Log-transformation and exclusion of outliers yielded similar results. The meta-analyzed index of moderated mediation was also significant. Headlines with alarm words attracted more likes than headlines without alarm words indirectly by enhancing the number of clicks when headlines also had prosocial words (indirect effect _{meta-analyzed} = .08, SE = .01, 95% CI = [.06, .09]) compared with headlines without prosocial words (indirect effect _{meta-analyzed} = .07, SE = .01, 95% CI = [.06, .09]), index of moderated mediation _{meta-analyzed} = .004, SE = .001, 95% CI = [.002, .01].

Other Criterion Variables

We also tested the same moderated mediation models to predict other generic digital footprints, including dislikes, shares, and comments. We found insignificant Alarm words × Prosocial words on dislikes (p = .31/.27/.89/.80; Sample 1 to Sample 4), shares (p = .61/.37/.79/.91), and comments (p = .91/.89/.96/.58), and inconsistent Clicks × Prosocial words on dislikes (p < .001/ < .001/ <.001/ = .74), shares (p = .47/ < .001/.17/.001), and comments (p < .001/ = .39/.001/.94).

Based on the preliminary results, we kept Clicks × Prosocial words interactions and removed Alarm words × Prosocial words on clicks, dislikes, shares, and comments to further meta-analyze the index of moderated mediation. The meta-analyzed indexes of moderated mediation were not significant (except one): Prosocial words did not moderate the indirect effects of alarm words on dislikes (index _{meta-analyzed} = –.40, SE = .31, 95% CI = [−1.01, .20]), shares (marginally significant: index _{meta-analyzed} = −1.44, SE = .69, 95% CI = [–2.79, –.09]), and comments (index _{meta-analyzed} = –.58, SE = .32, 95% CI = [−1.20, .04]) via clicks.

Alarm Words Dictionary

For headlines with alarm words that we operationalized above, we only included words that were generated from the most prevalent type of sensational news from 1700 to 2001 (Davis & McLeod, 2003). To improve its validity as an independent variable, we developed the completed Alarm Words Dictionary to test the same conditional indirect effect model.

Similar to the development of the Prosocial Words Dictionary, we used Davis and McLeod’s (2003) study and other expert opinions as a starting point to generate the dictionary development: accident, casualty, catastrophe, death, die, disaster, disease, kill, murder, tragedy, victim, war. Two individuals independently looked up these 12 alarm words in the Cambridge English Dictionary, Merriam-Webster Dictionary, and Oxford English Dictionary. The two individuals discussed and compared the results for all of the alarm words. Disagreements were resolved after discussion. Table S9 shows the completed English Alarm Words Dictionary.

The same 16 MTurk workers were recruited to validate the English Alarm Words Dictionary and were asked to describe three life-threatening experiences, three safe experiences, and three memorable experiences. Content analysis of the English experience text showed that life-threatening experience prompts had about 10 times higher alarm word density (M = 5.2%, SD = 4.5%) than safe experience prompts (M = 0.7%, SD = 1.4%) and memorable experience prompts did (M = 0.4%, SD = 1.0%), F(2, 30) = 14.98, p < .001, η² = .50.

The English alarm words were translated into Chinese. We adopted the Cambridge English Dictionary and Google Translate to translate the English terms into Chinese. After the translation, we consulted a good online Chinese dictionary (http://www.zdic.net/) to generate a culturally valid Chinese Alarm Words Dictionary.

To validate the Chinese Alarm Words Dictionary, the same 18 Chinese participants were asked to describe three life-threatening experiences, three safe experiences, and three memorable experiences. Content analysis of their Chinese experience text showed that life-threatening experience prompts yielded higher alarm word density (M = 5.5%, SD = 7.7%) than safe experience prompts (M = 0.0%, SD = 0.0%) and memorable experience prompts did (M = 1.2%, SD = 2.5%), F(2, 34) = 7.36, p = .002, η² = .30.

The Chinese-English translation of the Alarm Words Dictionary is displayed in Online Appendix B. Table S10 shows the Chinese Alarm Words Dictionary. A completed alarm word was coded as 1 if it was in a headline and 0 otherwise, and all were summed. We created a new dummy independent variable, completed alarm words, for the group of all headlines with alarm words (10.8/11.3/11.1/11.1%; Samples 1 to 4).

We then used the completed alarm words to investigate the same moderated mediation models through bootstrapping by resampling Samples 1 to 4 50,000 times each. We meta-analyzed the index of moderated mediation to test the indirect effect of completed alarm words on likes via clicks for the two prosocial word conditions across the four samples. The meta-analyzed index of moderated mediation with the completed alarm words was also significant, index of moderated mediation _{meta-analyzed} = 2.68, SE = .88, 95% CI = [.96, 4.40]. Headlines with completed alarm words indirectly yielded more likes than headlines without completed alarm words by enhancing the number of clicks when headlines also had prosocial words (c’ ps _{meta-analyzed} = .0549; indirect effect _{meta-analyzed} = 8.10, SE = 1.23, 95% CI = [5.69, 10.52]) compared with headlines without prosocial words (c’ ps _{meta-analyzed} = .0518; indirect effect _{meta-analyzed} = 5.31, SE = .63, 95% CI = [4.08, 6.53]).

Discussion

This study used a uses and gratifications approach to understand two evolutionary needs—the environmental surveillance need and social involvement need—in online news media. The two fundamental evolutionary needs, the need to detect potential threats surrounding individuals and the need to connect with others by showing prosociality, can be gratified via clicks and likes on online news platforms. Although the uses and gratifications framework has been suggested to explain the biological, psychological, and sociological motivations behind active and goal-oriented media users (Ruggiero, 2000), studies have mainly used this approach to explain motivations and needs related to sociocultural and personality trait factors. To the best of our knowledge, this is the first study to investigate evolutionary needs and motivations by using the uses and gratifications approach.

This study empirically examined sensational online news headlines with prosocial acts and their associated generic digital footprints. Our findings supported the concept of a human alarm system for sensational news by showing that individuals were more likely to click news headlines with alarm words (H1a). They were also more likely to “like” these articles (H1b). Furthermore, after reading news headlines with alarm words, individuals expended cognitive effort reading the story about the threatening event and then “liked” the article, signaling a survival concern (H1c). The results of the significant mediation effect indicated a causal model in which clicking and liking were causally related: The effect of alarm words on likes was mediated by clicks.

In addition, we found that although headlines with prosocial words did not generate either clicking or liking, prosociality moderated the human alarm system for sensational news, the psychological mechanism that signals threatening news to enhance the survival and reproduction of genes (Davis & McLeod, 2003; Eisenberger & Lieberman, 2004; Eisenberger et al., 2003; Hendriks Vettehen et al., 2008; van den Bos et al., 2008). Our findings indicated that prosocial words in news headlines did not moderate the association between alarm words in headlines and the survival selection device (i.e., clicks), but they did moderate the path from click to “like.” There is thus a conditional indirect effect: Prosociality does not affect the likelihood of getting clicked, but once clicked, prosociality increases the likelihood of being “liked” (H2).

This study enriches the theoretical understanding of sensational news research. Previous research has investigated the direct and indirect effects of sensational news on viewer attitudes, attention, beliefs, trust, and satisfaction with the news (Burgers & de Graaf, 2013; Hendriks Vettehen et al., 2008). Consistent with a previous study that found that sensational content enhanced selective exposure (Stanca et al., 2013), our findings showed that online news platform users were more likely to click sensational news headlines containing alarm-related words. This offers empirical support for the selectivity paradigm.

Traditionally, communication scientists have used a sociocultural perspective to explain the proximate psychological antecedents and consequences of media selection and influence, but have neglected the adaptive function by not investigating communication phenomena from an evolutionary perspective (Hennighausen & Schwab, 2015; Ng, 2016; Piazza & Ingram, 2009, 2015; Sherry, 2004). Following the recommendation of Sherry (2004) to consider media use and effect as the outcome of both nature and nurture, the present study is among the first to empirically examine why sensational news was selected over evolutionary time. This study provides empirical support for the notion that humans have psychological mechanisms designed to increase the likelihood of selecting and reading sensational news that can increase their chance of survival and reproductive success (Shoemaker, 1996).

Consistent with Dor’s (2003) argument that newspaper headlines function as relevance optimizers for readers, our results showed that alarm words in headlines have indirect effects on “likes” via clicks. The findings supported the relevance optimizer function in that the human alarm activated the survival concern after readers made a cognitive investment to read the article related to increased reproductive fitness (Davis & McLeod, 2003). The alarm words in the online news headlines directed readers to click them to gain additional information regarding the threat through cognitive investment after which readers then “liked” the news to show survival concern.

There are no previous studies in the sensational news literature on the importance of human alarm words in online news headlines as cues for sensational news that attract clicks as a survival selection device and “likes” as a proxy for survival concern. Furthermore, the possible impact of prosocial words in news headlines as a moderator has not been considered before. This represents an important new contribution to the literature on sensational news in general and on disasters news and related donations in particular.

Previous research on sensational news with prosocial acts has focused on the direct media effects of disaster news on donations of money by the audience and on the indirect effects through charitable thoughts and emotions (e.g., Bennett & Kottasz, 2000; Martin, 2013; Oosterhof et al., 2009; Seo et al., 2012). However, our findings provide empirical support for a selective media effect, in that sensational news attracts people to read the news and, when such news involves prosocial acts, individuals are more likely to show concern by “liking” them. The combination of the selective media effect of sensational news in this study and the direct and indirect media effect of sensational news influence shown in previous studies contributes theoretical insights into the SESAM model (Knobloch-Westerwick, 2015), indicating that selective exposure and responses to sensational news may dynamically affect people’s prosocial behavior. In other words, people show concern for news about selected disasters or diseases involving prosocial acts, and this selective exposure dynamically influences their own prosocial behavior. This transactional media effect is worth investigating in future.

Our findings may provide implications because they furnish the possibility that not only donations but also other charitable acts may be related to sensational news. Furthermore, the category of sensational news is not limited to disasters but includes diseases and accidents where prosocial acts are involved. Future sensational news research may identify the effect of sensational news other than disasters (e.g., car accidents) and prosocial acts other than money donations (e.g., organ donation). An important theoretical insight from our findings is that prosocial words in headlines enhance the chance of “liking” the news, given that individuals read the sensational news after clicking the headline.

This study also has practical implications for journalists. Over the past few decades, journalists have realized the importance and news value of sensational news (Galtung & Ruge, 1965; Harcup & Neill, 2001; Shoemaker & Reese, 2014). Audiences are attracted by sensational news due to their evolutionary and fundamental need to pay attention to threatening information (Davis & McLeod, 2003; Shoemaker, 1996). However, sensational news selection has been perceived as a negative act (Kilgo et al., 2016), which has created a barrier to satisfying a basic human instinct and a need for orientation from news consumption. This study found that sensational news did attract individuals’ attention and that when such news mentioned prosocial acts, readers were more likely to show concern. The implication of the findings for journalists is that while the selection of sensational news may have a negative stigma, the selection of sensational news depicting prosocial acts demonstrates a positive side of human nature. Sensational news about prosocial acts can satisfy both the evolutionary need for orientation and the need for social involvement.

This study also contributes methodological advances. Traditionally, in studies of the influence of sensational news, scholars have used surveys or experimental designs with small sample sizes. The availability of big data online offers a flexible alternative for social scientists to conduct ecologically valid research on an unprecedented scale to gain insights from online media language and word use at the macro level (Kern et al., 2016). We have introduced the Alarm Words Dictionary to reflect sensational news and also developed the Prosocial Words Dictionary, two categories not found in LIWC (Pennebaker et al., 2015), for use by future researchers to study the functions of alarm and prosocial words online.

One may argue that the current study cannot directly consider the two fundamental evolutionary needs that can be gratified through media engagement (i.e., clicks and likes). In fact, humans rarely make decisions based explicitly on accurate evaluations of fitness enhancement. Instead, they do so using proxies of fitness, which are psychological and behavioral antecedents and consequences that are indirectly associated with biological fitness (Mishra, 2014; Mishra, Barclay, & Sparks, 2017). We propose that clicks and likes that indicate a proxy for survival selection and concern increase positive proxies of fitness: humans gain threat-related information by paying cognitive effort to these news headlines.

Furthermore, while gratifying the two fundamental evolutionary needs through media engagement does not actually resolve issues related to survival, need fulfillment through news media use can enhance fitness. The “must-solve” category represents distinct adaptive problems that cause reproductive and survival failure if they are not solved immediately, whereas the “beneficial” category represents adaptive problems that are not urgent to solve but increase fitness if they are solved (Lewis, Al-Shawaf, Conroy-Beam, Asao, & Buss, 2017). The environmental surveillance need and social involvement need gratified via clicks and likes on online news platforms increase positive proxies of fitness although humans do not experience immediate reproductive and survival failure if these two needs are not gratified.

Our study has several limitations. We investigated only one online news platform in one country, the People’s Republic of China, where the media are under government control. However, sensational news such as natural disasters and disease propagation, which imply evolutionary adaptedness, attracts all human beings regardless of cultural and political influence. Future research should consider the functions of prosocial words on other popular social media such as Facebook and Twitter.

Although we ensured the validity of the Prosocial Words Dictionary and Alarm Words Dictionary in this study, the validation of alarm and prosocial words was not our main purpose. However, the findings of this study provide insights for the strict validation of the English, Traditional Chinese, and Simplified Chinese versions of the dictionaries for future studies.

As no experimental variation was carried out in this study, the exact causal relationship is unclear. The cross-sectional and correlational nature of our data does not allow for causal inferences between prosocial and alarm words and survival selection mechanisms and concern. We also assessed word use and digital footprints in the same time period. Future longitudinal research should test the traditional prosocial media effects over time.

Supplemental Material

online_supplemental_materials – Supplemental material for The Human Alarm System for Sensational News, Online News Headlines, and Associated Generic Digital Footprints: A Uses and Gratifications Approach

Supplemental material, online_supplemental_materials for The Human Alarm System for Sensational News, Online News Headlines, and Associated Generic Digital Footprints: A Uses and Gratifications Approach by Yu-Leung Ng and Xinshu Zhao in Communication Research

Footnotes

Acknowledgements

We thank Horace Lu for data crawling, Angela Wang for duplicate data deletion, and June Yeung and Ke Zhang for data coding.

Authors’ Note

YLN analyzed the data and wrote the article; XSZ crawled the data and provided comments.

Declaration of Conflicting Interests

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

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by a donation from Dr. Lee Shau Kee, GBM, through Hong Kong Baptist University Research Grant Scheme (2015-2017, Zhao PI), and a China Ministry of Education Major Grant for Social Sciences, through Fudan University Center for Information and Communication Studies (11JJD860007, Zhao PI) to XSZ for data crawling.

ORCID iD

Yu-Leung Ng

Supplemental Material

Supplemental material for this article is available online.

Author Biographies

Yu-Leung Ng is a postdoctoral fellow at the Journalism and Media Studies Center, The University of Hong Kong. His research interests include media effects, media psychology, and impacts of new media technologies.

Xinshu Zhao is a chair professor at the School of Communication, Hong Kong Baptist University. His research interests include quantitative social research and communication and media.

References

Arbaoui

De Swert

van der Brug

(2016). Sensationalism in news coverage: A comparative study in 14 television systems. Communication Research, 47, 299-320. doi:10.1177/0093650216663364

Baumeister

R. F.

Bratslavsky

Finkenauer

Vohs

K. D.

(2001). Bad is stronger than good. Review of General Psychology, 5, 323-370. doi:10.1037//1089-2680.5.4.323

Bennett

Kottasz

(2000). Emergency fund-raising for disaster relief. Disaster Prevention and Management: An International Journal, 9, 352-360. doi:10.1108/09653560010361393

Blumler

J. G.

(1979). The role of theory in uses and gratifications studies. Communication Research, 6, 9-36. doi:10.1177/009365027900600102

Boyd

Richerson

P. J.

(1989). The evolution of indirect reciprocity. Social Networks, 11, 213-236. doi:10.1016/0378-8733(89)90003-8

Burgers

de Graaf

(2013). Language intensity as a sensationalistic news feature: The influence of style on sensationalism perceptions and effects. Communications: The European Journal of Communication Research, 38, 167-188. doi:10.1515/commun-2013-0010

Cheung

W.-L.

Jak

(2016). Analyzing big data in psychology: A split/analyze/meta-analyze approach. Frontiers in Psychology, 7, 1-13. doi:10.3389/fpsyg.2016.00738

Cohen

(1992). A power primer. Psychological Bulletin, 112, 155-159. doi:10.1037/0033-2909.112.1.155

Davis

McLeod

S. L.

(2003). Why humans value sensational news: An evolutionary perspective. Evolution and Human Behavior, 24, 208-216. doi:10.1016/S1090-5138(03)00012-6

10.

Dehghani

Johnson

Hoover

Sagi

Garten

Parmar

N. J.

. . . Graham

(2016). Purity homophily in social networks. Journal of Experimental Psychology: General, 145, 366-375. doi:10.1037/xge0000139

11.

Dor

(2003). On newspaper headlines as relevance optimizers. Journal of Pragmatics, 35, 695-721. doi:10.1016/S0378-2166(02)00134-0

12.

Ecker

U. K. H.

Lewandowsky

Chang

E. P.

Pillai

(2014). The effects of subtle misinformation in news headlines. Journal of Experimental Psychology: Applied, 20, 323-335. doi:10.1037/xap0000028

13.

Eichstaedt

J. C.

Schwartz

H. A.

Kern

M. L.

Park

Labarthe

D. R.

Merchant

R. M.

. . . Seligman

M. E. P.

(2015). Psychological language on Twitter predicts county-level heart disease mortality. Psychological Science, 26, 159-169. doi:10.1177/0956797614557867

14.

Eisenberger

N. I.

Lieberman

M. D.

(2004). Why rejection hurts: A common neural alarm system for physical and social pain. Trends in Cognitive Sciences, 8, 294-300. doi:10.1016/j.tics.2004.05.010

15.

Eisenberger

N. I.

Lieberman

M. D.

Williams

K. D.

(2003). Does rejection hurt? An FMRI study of social exclusion. Science, 302, 290-292. doi:10.1126/science.1089134

16.

Frimer

J. A.

Schaefer

N. K.

Oakes

(2014). Moral actor, selfish agent. Journal of Personality and Social Psychology, 106, 790-802. doi:10.1037/a0036040

17.

Galtung

Ruge

M. H.

(1965). The structure of foreign news. Journal of Peace Research, 2, 64-90. doi:10.1177/002234336500200104

18.

Gan

(2017). Understanding WeChat users’ liking behavior: An empirical study in China. Computers in Human Behavior, 68, 30-39. doi:10.1016/j.chb.2016.11.002

19.

Goetz

J. L.

Keltner

Simon-Thomas

(2010). Compassion: An evolutionary analysis and empirical review. Psychological Bulletin, 136, 351-374. doi:10.1037/a0018807

20.

Graham

Haidt

Nosek

B. A.

(2009). Liberals and conservatives rely on different sets of moral foundations. Journal of Personality and Social Psychology, 96, 1029-1046. doi:10.1037/a0015141

21.

Hamilton

W. D.

(1964a). The genetical evolution of social behaviour. I. Journal of Theoretical Biology, 7, 1-16. doi:10.1016/0022-5193(64)90038-4

22.

Hamilton

W. D.

(1964b). The genetical evolution of social behaviour. II. Journal of Theoretical Biology, 7, 17-52. doi:10.1016/0022-5193(64)90039-6

23.

Harcup

T. O.

Neill

(2001). What is news? Galtung and Ruge revisited. Journalism Studies, 2, 261-280. doi:10.1080/14616700118449

24.

Harcup

T. O.

Neill

(2017). What is news? News values revisited (again). Journalism Studies, 18, 1470-1488. doi:10.1080/1461670X.2016.1150193

25.

Hayes

A. F.

(2013). An introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York, NY: Guilford Press.

26.

Hendriks Vettehen

Nuijten

Beentjes

. (2005). News in an age of competition: The case of sensationalism in Dutch television news, 1995-2001. Journal of Broadcasting & Electronic Media, 49, 282-295. doi:10.1207/s15506878jobem4903

27.

Hendriks Vettehen

Nuijten

Peeters

. (2008). Explaining effects of sensationalism on liking of television news stories: The role of emotional arousal. Communication Research, 35, 319-338. doi:10.1177/0093650208315960

28.

Hennighausen

Schwab

(2015). Evolutionary media psychology and its epistemological foundation. In T. Breyer (Ed.), Epistemological dimensions of evolutionary psychology (pp. 131-158). New York, NY: Springer. doi:10.1007/978-1-4939-1387-9_7

29.

Hong

Chen

(2017). “Liking” and being “liked”: How are personality traits and demographics associated with giving and receiving “likes” on Facebook? Computers in Human Behavior, 68, 292-299. doi:10.1016/j.chb.2016.11.048

30.

Ifantidou

(2009). Newspaper headlines and relevance: ad hoc concepts in ad hoc contexts. Journal of Pragmatics, 41, 699-720. doi:10.1016/j.pragma.2008.10.016

31.

Ireland

M. E.

Schwartz

H. A.

Chen

Ungar

L. H.

Albarracín

(2015). Future-oriented tweets predict lower county-level HIV prevalence in the United States. Health Psychology, 34, 1252-1260. doi:10.1037/hea0000279

32.

Ito

T. A.

Larsen

J. T.

Smith

N. K.

Cacioppo

J. T.

(1998). Negative information weighs more heavily on the brain: The negativity bias in evaluative categorizations. Journal of Personality and Social Psychology, 75, 887-900. doi:10.1037/0022-3514.75.4.887

33.

JASP Team. (2017). JASP (Version 0.8.4) [Computer software]. Amsterdam, The Netherlands: University of Amsterdam.

34.

Katz

Blumler

J. G.

Gurevitch

(1973). Uses and gratifications research. Public Opinion Quarterly, 37, 509-523. doi:10.1086/268109

35.

Katz

Blumler

J. G.

Gurevitch

(1974). Utilization of mass communication by the individual. In Blumler

J. G.

Katz

(Eds.), The uses of mass communication: Current perspectives on gratifications research (pp. 19-32). Beverly Hills, CA: Sage.

36.

Katz

Haas

Gurevitch

(1973). On the use of the mass media for important things. American Sociological Review, 38, 164-181. doi:10.2307/2094393

37.

Keltner

Kogan

Piff

P. K.

Saturn

S. R.

(2014). The sociocultural appraisals, values, and emotions (SAVE) framework of prosociality: Core processes from gene to meme. Annual Review of Psychology, 65, 425-460. doi:10.1146/annurev-psych-010213-115054

38.

Kenny

D. A.

(2018). Mediation. Retrieved from http://davidakenny.net/cm/mediate.htm

39.

Kern

M. L.

Park

Eichstaedt

J. C.

Schwartz

H. A.

Sap

Smith

L. K.

Ungar

L. H.

(2016). Gaining insights from social media language: Methodologies and challenges. Psychological Methods, 21, 507-525. doi:10.1037/met0000091

40.

Kilgo

D. K.

Harlow

García-Perdomo

Salaverría

(2016). A new sensation? An international exploration of sensationalism and social media recommendations in online news publications. Journalism, 19, 1497-1516. doi:10.1177/1464884916683549

41.

Knobloch-Westerwick

(2015). The selective exposure self- and affect-management (SESAM) model: Applications in the realms of race, politics, and health. Communication Research, 42, 959-985. doi:10.1177/0093650214539173

42.

Kosinski

Stillwell

Graepel

(2013). Private traits and attributes are predictable from digital records of human behavior. Proceedings of the National Academy of Sciences, 110, 5802-5805. doi:10.1073/pnas.1218772110

43.

Kurzban

Burton-Chellew

M. N.

West

S. A.

(2015). The evolution of altruism in humans. Annual Review of Psychology, 66, 575-599. doi:10.1146/annurev-psych-010814-015355

44.

Lewis

D. M. G.

Al-Shawaf

Conroy-Beam

Asao

Buss

D. M.

(2017). Evolutionary psychology: A how-to guide. American Psychologist, 72, 353-373. doi:10.1037/a0040409

45.

Lin

Lucas

H. C.

Shmueli

(2013). Too big to fail: Large samples and the p-value problem. Information Systems Research, 24, 906-917. doi:10.1287/isre.2013.0480

46.

MacKinnon

D. P.

(2008). Introduction to statistical mediation analysis. Mahwah, NJ: Lawrence Erlbaum.

47.

Martin

J. A.

(2013). Disasters and donations: The conditional effects of news attention on charitable giving. International Journal of Public Opinion Research, 25, 547-560. doi:10.1093/ijpor/eds044

48.

Matthes

(2005). The need for orientation towards news media: Revising and validating a classic concept. International Journal of Public Opinion Research, 18, 422-444. doi:10.1093/ijpor/edh118

49.

McAndrew

F. T.

(2012). New evolutionary perspectives on altruism: Multilevel-selection and costly-signaling theories. Current Directions in Psychological Science, 11, 79-82. doi:10.1111/1467-8721.00173

50.

Mishra

(2014). Decision-making under risk: Integrating perspectives from biology, economics, and psychology. Personality and Social Psychology Review, 18, 280-307. doi:10.1177/1088868314530517

51.

Mishra

Barclay

Sparks

(2017). The relative state model: Integrating need-based and ability-based pathways to risk-taking. Personality and Social Psychology Review, 21, 176-198. doi:10.1177/1088868316644094

52.

Neyeloff

J. L.

Fuchs

S. C.

Moreira

L. B.

(2012). Meta-analyses and forest plots using a microsoft excel spreadsheet: Step-by-step guide focusing on descriptive data analysis. BMC Research Notes, 5, Article 52. doi:10.1186/1756-0500-5-52

53.

Y.-L.

(2016). More than social–cultural influences: A research agenda for evolutionary perspectives on prosocial media effects. Review of General Psychology, 20, 317-335. doi:10.1037/gpr0000084

54.

Olkin

Sampson

(1998). Comparison of meta-analysis versus analysis of variance of individual patient data. Biometrics, 54, 317-322. doi:10.2307/2534018

55.

Oosterhof

Heuvelman

Peters

(2009). Donation to disaster relief campaigns: Underlying social cognitive factors exposed. Evaluation and Program Planning, 32, 148-157. doi:10.1016/j.evalprogplan.2008.10.006

56.

Pennebaker

J. W.

Boyd

R. L.

Jordan

Blackburn

(2015). The development and psychometric properties of LIWC2015. Austin, TX: University of Texas at Austin.

57.

Penner

L. A.

Dovidio

J. F.

Piliavin

J. A.

Schroeder

D. A.

(2005). Prosocial behavior: Multilevel perspectives. Annual Review of Psychology, 56, 365-392. doi:10.1146/annurev.psych.56.091103.070141

58.

Piazza

J. R.

Ingram

G. P. D.

(2009). Evolutionary cyber-psychology: Applying an evolutionary framework to Internet behavior. Computers in Human Behavior, 25, 1258-1269. doi:10.1016/j.chb.2009.07.002

59.

Piazza

J. R.

Ingram

G. P. D.

(2015). Evolutionary cyberpsychology 2.0: Revisiting some old predictions and posting some new ones in the age of Facebook. In Zeigler-Hill

Welling

L. M.

Shackelford

T. K.

(Eds.), Evolutionary perspectives on social psychology (pp. 159-174). New York, NY: Springer. doi:10.1007/978-3-319-12697-5_13.

60.

Preacher

K. J.

Kelley

(2011). Effect size measures for mediation models: Quantitative strategies for communicating indirect effects. Psychological Methods, 16, 93-115. doi:10.1037/a0022658

61.

Ruggiero

T. E.

(2000). Uses and gratifications theory in the 21st century. Mass Communication and Society, 3, 3-37. doi:10.1207/S15327825MCS0301_02

62.

Schultz

(2007). The journalistic gut feeling: Journalistic doxa, news habitus and orthodox news values. Journalism Practice, 1, 190-207. doi:10.1080/17512780701275507

63.

Seo

Sun

Merolla

A. J.

Zhang

(2012). Willingness to help following the Sichuan earthquake: Modeling the effects of media involvement, stress, trust, and relational resources. Communication Research, 39, 3-25. doi:10.1177/0093650210388407

64.

Sherman

L. E.

Payton

A. A.

Hernandez

L. M.

Greenfield

P. M.

Dapretto

(2016). The power of the like in adolescence: Effects of peer influence on neural and behavioral responses to social media. Psychological Science, 27, 1027-1035. doi:10.1177/0956797616645673

65.

Sherry

J. L.

(2004). Media effects theory and the nature/nurture debate: A historical overview and directions for future research. Media Psychology, 6, 83-109. doi:10.1207/s1532785xmep0601_4

66.

Shoemaker

(1996). Hardwired for news: Using biological and cultural evolution to explain the surveillance function. Journal of Communication, 46, 32-47. doi:10.1111/j.1460-2466.1996.tb01487.x

67.

Shoemaker

Reese

(2014). Mediating the message in the 21st century: A media sociology perspective (3rd ed.). New York, NY: Routledge.

68.

Stanca

Gui

Gallucci

(2013). Attracted but unsatisfied: The effects of sensational content on television consumption choices. Journal of Media Economics, 26, 82-97. doi:10.1080/08997764.2013.785552

69.

Tausczik

Y. R.

Pennebaker

J. W.

(2010). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology, 29, 24-54. doi:10.1177/0261927X09351676

70.

Tomasello

(2014). The ultra-social animal. European Journal of Social Psychology, 44, 187-194. doi:10.1002/ejsp.2015

71.

Trivers

R. L.

(1971). The evolution of reciprocal altruism. The Quarterly Review of Biology, 46, 35-57. doi:10.1086/406755

72.

Vaish

Grossmann

Woodward

(2008). Not all emotions are created equal: The negativity bias in social-emotional development. Psychological Bulletin, 134, 383-403. doi:10.1037/0033-2909.134.3.383

73.

Valenzuela

Bachmann

Aguilar

(2016). Socialized for news media use: How family communication, information-processing needs, and gratifications determine adolescents exposure to news. Communication Research, 46, 1095-1118. doi:10.1177/0093650215623833

74.

van den Bos

Ham

Lind

E. A.

Simonis

van Essen

W. J.

Rijpkema

. (2008). Justice and the human alarm system: The impact of exclamation points and flashing lights on the justice judgment process. Journal of Experimental Social Psychology, 44, 201-219. doi:10.1016/j.jesp.2007.03.001

75.

Weaver

D. H.

(1980). Audience need for orientation and media effects. Communication Research, 7, 361-373. doi:10.1177/009365028000700305

76.

Wen

Fan

(2015). Monotonicity of effect sizes: Questioning kappa-squared as mediation effect size measure. Psychological Methods, 20, 193-203. doi:10.1037/met0000029

77.

Youyou

Kosinski

Stillwell

(2015). Computer-based personality judgments are more accurate than those made by humans. Proceedings of the National Academy of Sciences, 112, 1036-1040. doi:10.1073/pnas.1418680112

78.

Zaki

(2014). Empathy: A motivated account. Psychological Bulletin, 140, 1608-1647. doi:10.1037/a0037679

79.

Zhao

Lynch

J. G.

Chen

(2010). Reconsidering Baron and Kenny: Myths and truths about mediation analysis. Journal of Consumer Research, 37, 197-206. doi:10.1086/651257

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.18 MB