Previous Article in Journal
Human-Centred Design (HCD) in Enhancing Dementia Care Through Assistive Technologies: A Scoping Review
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Digital
Volume 5
Issue 4
10.3390/digital5040052
digital-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

When Fear Backfires: How Emotionality Reduces the Online Sharing of Threatening Messages

by
Violet Cheung-Blunden
1,* and
Emily Ann Zhou
2
1
Department of Psychology, University of San Francisco, San Francisco, CA 94117, USA
2
School Counseling, Steinhardt School of Culture, Education, and Human Development, New York University, NYU, New York, NY 10003, USA
*
Author to whom correspondence should be addressed.
Digital 2025, 5(4), 52; https://doi.org/10.3390/digital5040052
Submission received: 16 August 2025 / Revised: 30 September 2025 / Accepted: 3 October 2025 / Published: 6 October 2025
Browse Figures
Versions Notes

Abstract

The present study utilized two prominent emotion theories to investigate intention and behavior involved in propagating threatening social media messages. Participants were randomly assigned to different blocks of tweets/Xs with the same word count but different topics/sentiments. The topics in Study 1 (N = 619) were neutral and illegal border crossing, whereas the topics in Study 2 (N = 577) were the virulent risk of COVID-19 and the potential risks of newly developed vaccines. Dissemination intention was gauged by the number of tweets that participants wanted to share. Participants were also asked to summarize the messages to observe their behavioral engagement with the information, specifically through time spent on the task and the number of words written. An intention–behavior disjoint was found under all threatening topics and on both sides of the political divide. Fearful participants showed engaging intentions (wanted to share more tweets) but disengaging behaviors (wrote fewer words and submitted their summaries sooner). The necessary and sufficient conditions for the intention–behavior disjoint seemed to be the presence of threatening contents and subjective fear. Communicating risks can spark interest, but it is important not to burden the audience with too much fear, or they may stop spreading the word.

1. Introduction

Sensationalism has long been a strategic tool in journalism to attract readers and influence public opinion dating back to the era of Randolph Hearst and the rise in yellow journalism [1]. The question in the current media environment is whether the emotional appeals that once captivated newspaper readers retain their allure on social media. With over five billion users worldwide and growing [2], social media has enabled ordinary people, celebrities, and political leaders to partake in the discourse on prevalent topics such as social movements, current events, and more recently, the COVID-19 pandemic [3,4,5]. The popularity of a social media post is often gauged by the number of people who pay attention to the post, as well as the number of users who actively like, share, or comment on it [6,7,8,9]. Common strategies to attain popularity on social media appear to be sensationalism 2.0, which includes creating captivating content and using powerful emotions to frame the message [10].
However, research to date is mixed on whether sensationalism or emotional content can successfully drive virality [5,11,12,13,14,15,16,17,18]. This paper adopts a process-oriented approach to deconstruct the process of message sharing by distinguishing between two critical user responses: initial attention paid to a message and the subsequent decision to share it [18,19,20]. The psychological mechanisms of the two stages are different, such that highly strung users do not always disseminate the information they read. The first stage—attention—often involves bottom-up processing, where certain types of features are uniquely capable of capturing psychological resources. For example, marketing research has long recognized that visually salient packaging automatically captures consumers’ attention, biasing their otherwise more conscious choices [21]. The second stage—sharing—requires top-down processing, involving deliberation about whether one wants to carry out the sharing behaviors. Adding to the complexity of the cost analysis in the second stage are the various behavioral demands from different platforms, ranging from effortless sharing through a single click to more involved actions such as retelling, rephrasing, or commenting on the message. A similar development can be found in marketing research, which used eye tracking data to show that, despite being attracted by visually salient features, consumers’ ultimate product selection was arbitrated by considerations of goals and information value [22]. Beyond marketing research, the underlying neural mechanisms involved in the bottom-up and top-down processes are being unpacked in terms of neural firing rate as well as their variability and correlation across neurons [23].
The deconstruction of the information-sharing process into two stages may bring a refined framework for understanding users’ responses to digital content. Prior research on the dissemination of social media messages often glosses over these stages [9,10,11,12,24], leading to a premature assumption that attention-garnering content (indicated by “eyeballs” or visits/views) is inherently more widely shared (indicated by retweets, comments, or likes). The present paper challenges this assumption by drawing on two prominent emotion theories to identify one instance where a single emotion can either amplify or inhibit sharing at different stages of the dissemination process [20,25,26,27,28,29,30,31,32,33]. We show that fearful users may attend to threatening messages and even find them worthy to share, only to become inhibited when carrying out the necessary actions to spread the message. This potential disjunction between sharing intentions and behaviors provides an interpretive lens for understanding the mixed literature on sensationalism and social media. Whether a past study used intention or virality as an engagement metric may very well derive different findings, as each captures distinct aspects of content interaction.

1.1. Negative Bias Principle and Functionalist Approach to Fear

The negative bias principle is a prominent theory aimed at explaining the captivating power of negative content. Under this theoretical framework, adverse events and negatively framed information are disproportionately influential compared to positive events and news [34,35]. Adverse events are said to contain a higher informational value that often requires more urgent responses [30]. The salience of negative information has been validated in several domains of functioning, including prioritized cognitive processing, decision-making, and attention-capturing [20,25,27,30]. Previous studies have successfully validated this negativity bias by demonstrating the efficacy of using negatively valenced messages to attract attention [17]. Smith and Petty’s study on message processing suggested that negatively framed messages elicit greater cognitive involvement and attention from the recipient [16]. Specifically, when participants were presented with arguments for the importance of recycling, the negatively framed arguments resulted in greater message scrutiny than the positively framed messages. Pratto and John’s study [20] also found that negative undesirable trait adjectives printed on a word list captured more attention among participants when compared to positive desirable traits.
The ability of negative messages to capture attentional and cognitive bandwidths suggests greater engagement with negatively framed social media messages and even amplification along the message-sharing chain. Vosoughi [10] found that false news spreads faster online and attributed this effect to the novelty of the information, which inspires fear, disgust, and surprise in the audience. Similarly, Jagiello and Hills [12] found an amplification effect where high dread topics (e.g., nuclear energy) were transmitted with increasingly negative sentiments. However, negativity is not always a reliable driver on social media. In an analysis of millions of tweets from Russian and Iranian troll farms, Cheung-Blunden [11] found that negative tweets were less liked and retweeted than the average tweet. Adding to the mixed literature, Pivecka [15] found that both positive and negative political tweets were less likely to be retweeted than neutral messages. In a similar vein, negative messages regarding the harmfulness of an antibacterial agent in Moussaïd [13] diminished along the transmission chain, and so did their positive or neutral counterparts.
The negative bias principle may not fully explain the mixed performance of negative messages, perhaps due to its theoretical focus on the early stages of securing attentional and cognitive resources. Information sharing is a multi-stage process that requires behavioral output from users to actively share the message in the end. The functionalist approach to emotions is well situated to explain behaviors because each emotion is seen as a functional unit, with the goal to drive a set of actions to solve a particular evolutionarily recurrent problem [26,28,29,31,32]. A large body of literature following this approach has linked distinct negative emotions with unique behaviors and physiological responses. For example, anger was found to result in feelings of blood rushing through the body and thoughts of lashing out at others through a verbal or physical attack [31,36,37]. On the other hand, sadness was associated with tiredness and a desire to cry, perhaps to elicit help from others [31].
The behavioral outcomes of anger and sadness exemplify the tenet of the functionalist approach that negative emotions have different purposes and therefore are not functionally equivalent [8,31,38]. In this framework, fear, the emotion of focus, has a function of generating avoidance actions, so the protagonist would quickly distance themselves from threatening stimuli [7,31,33,39,40,41,42,43,44]. Other physiological and behavioral signatures of fear include blood draining away from the face, temperature lowering in fingertips and extremities, and experiencing an impulse to escape [26,31,45,46]. These bodily sensations enable avoidance behaviors, which seem to translate into disengagement rather than engagement in the social media space [26,41,47].

1.2. The Influence of Political Ideology on Risk Aversion

Since fear responses are known to differ across the political divide, participants’ political orientation must not be neglected in a study of political issues [48]. Political orientation is a known factor in the intensity of fear responses, which can further impart a difference in attention and behaviors. Decades of research have established that conservatives tend to be more sensitive to threats than liberals [48,49,50,51,52]. Their heightened fearful responses have been attributed to various aspects of conservatism, such as tendencies toward protectionism or adherence to traditions [34,52,53,54,55], and even biological differences. Brain imaging studies reveal that conservatives have a more prominent and active right amygdala—an area associated with expressing and processing fear [56]. This processing difference leaves conservatives with an overall smaller risk appetite and a greater need to seek safety [48,53].
Conservatives’ heightened fear responses are not uniform; rather, they vary depending on the topic. This sensitivity to context means that their risk responses can be either amplified or diminished based on the issue at hand. Immunization is a topic that seems to exacerbate fear responses in conservatives [57,58]. Speed and Mannion pointed to two levels of fear—moral and physical—in their study of the anti-vaccination movement that has flooded social media in recent years [59]. Activist groups and populist leaders “create and sustain a moral panic which plays on deep-seated atavistic fears around vaccination and public safety” (p. 1976). Likely contributing factors include perceived physical dangers and the key right-wing values that vaccination affronts, e.g., adherence to tradition, individual liberty, and limited government [54,55,60,61].
Unlike immunization, the virulent risk of COVID-19 seems to draw muted responses from conservatives as compared to liberals [62,63,64]. The common reasons cited include distrust in government-issued public health warnings and examples set by conservative party leaders [65]. As a result, COVID-19 joins climate change, environmental pollution, healthcare access, and corporate misconduct in a category of risks that are exempt from conservatives’ risk sensitivity [66,67].

1.3. Present Studies

The present studies postulate an intention–behavior disjoint in the way fearful participants approach threatening social media messages. According to the negative bias principle, fear-laden messages attract attention and thus make users say they would spread the information. However, once fear is induced in a participant, they have a limited behavioral capacity to make their promises a reality according to the functionalist approach to emotions. To detect behavioral disengagement, the present study chose the most demanding sharing behavior of retelling (instead of less demanding clicks, for example). In sum, fearful participants would show a larger intention–behavior disjoint than their less-fearful counterparts; they are more likely to say they intend to share threatening information but less likely to engage with it during the summary task.
Study 1 tested the hypothesis on the intention–behavior disjoint (H1), and Study 2 included two additional hypotheses to test whether fear (H2) and the intention–behavior disjoint (H3) are sensitive to the topic and political orientation.
H1. 
Fearful participants demonstrate an intention–behavior disjoint in the face of threatening information. This means that despite their more expressed intention to disseminate messages, they would fall short in behavioral follow-through by writing fewer words and submitting their summaries sooner.
H2. 
Politically conservative participants would report more fear than liberal participants. Conservatives’ threat response is heightened toward COVID-19 vaccines but muted toward the COVID-19 virus.
H3. 
The intention–behavior disjoint is moderated by political ideology. Specifically, the disjoint is expected to be more severe for conservatives in the vaccine condition and for liberals in the COVID-19 condition.

2. Study 1

This study was designed to test the intention–behavior disjoint (H1) under two types of topics—neutral messages and threatening messages. Deportation fears under the first Trump administration drove tens of thousands of refugees into Eastern Canada [68]. Canada’s ill-preparedness for its unprecedented migrant crisis on the US–Canada border became a high-profile issue at the time of the present study. Therefore, illegal immigration and alleged lone-wolf terrorist attacks were presented to the experimental group, while neutral topics were given to the control group.

2.1. Materials and Methods

2.1.1. Participants

The participants were 593 voting-age Canadian citizens (M = 51.2 years, SD = 16.4 years, 34.3% male and 65.7% female). The common party affiliations were the Liberal Party (31.9%), the Conservative Party (23.2%), and the New Democratic Party (14.9%), with almost one-fifth of the participants self-identifying as politically independent (18.3%). A quarter of the participants were Roman Catholic (25.0%), followed by Protestant (23.4%), and Atheist (15.7%). A great majority of the participants self-identified as White (78.1%).

2.1.2. Procedure

Qualtrics Professional Services recruited a cross-sectional convenience sample of voting-age Canadian citizens. Participants completed a consent form and responded to demographic questions. They were then randomly assigned to one of two groups to read either ten tweets on neutral topics about everyday life or ten tweets about border insecurity and terrorist attacks. The two tweet sets were drafted to (a) reflect the trending social media posts in Canada at the time, (b) match the word counts across the two sets of tweets with 100 words in each set, and (c) contain predetermined amounts of neutral and negative/fearful sentiments, respectively.
To craft the sentiments in the tweets, the content was checked by the word-emotion association analytic tool (also known as NRC emotion lexicon). Mohammad and Turney [69] used crowdsourcing to develop a dictionary to score the sentiments of texts based on their word contents. The NRC emotion lexicon is a widely used and validated tool for text-based sentiment analysis. It has been successfully applied in previous studies, revealing public opinion and emotions of tweets on the COVID-19 pandemic and the vaccines [11,70,71]. This tool showed that our tweet set for the control group was emotionally neutral (contained no words that were scored as positive, negative, fear, joy, surprise, anger, disgust, or sadness). Our tweet set for the experimental group featured negative sentiments (18 negative words and three positive words) and fear (18 fearful words and five or fewer words on other discrete emotions). A sample neutral tweet is, “I have to remind myself that there are seasons in other parts of the world.” A sample fear-laden tweet is “Gunman concealed weapon during a lethal shootout that killed a legislative assistant.”
The 10 tweets were presented on the same page and were said to come from ostensibly “well-established sources.” Participants reported the total number of tweets they intended to retweet/share and summarized the tweets. Behavioral indicators from the summarizing exercise (word count and time to completion) were recorded and calculated. Participants also reported their levels of fear immediately after the summary task. Outliers of the behavioral indicators were replaced with values capped at two standard deviations around the mean. This Winsorizing technique helps minimize the undue impact of outliers on statistical estimates and improve inference [72].
The first phase of data cleaning was conducted by Professional Services at Qualtrics, based on a set of criteria for closed-ended responses and survey metrics. Participants’ data were delivered to the research team only if they spent at least 10 min on the survey (about half of the median response time) and did not give straight-lined responses to items in matrix tables [73]. The rest of the data cleaning was conducted at the lab based on open-ended responses. The data from 19 participants were discarded due to poor-quality summaries. Their text entries showed clear signs of either straight-lining (i.e., similar summaries to all tweets) or irrelevance (e.g., entered “fake news” or “this is not true” in many of their summaries).

2.1.3. Instruments

Dissemination intent. The number of tweets, out of the set of 10, that the participant intended to share.
Words written. The total number of words that the participant typed to summarize the tweet set, used as an indicator of the participant’s cognitive engagement with the tweets.
Task completion time. The total amount of time the participant spent writing the summaries, primarily used as an indicator of their behavioral output/efforts.
State fear. A 9-item measure by Cheung-Blunden and Ju [19] was used to assess fearful feelings towards a topic on a 4-point scale, ranging from 1 (not at all) to 4 (very much). The scale was chosen as it targets fear and showed discriminant validity from measures of state anxiety [74]. Sample items include “I feel terrified” and “I feel chills down my spine.” The measure demonstrated high internal consistency (α = 0.97).

2.2. Results

Prior to hypothesis testing, an overview of participants’ performance in the summarizing task was sought through descriptive statistics. Each participant was exposed to a tweet set of 100 words and the average summary was about half of the original length (M = 43.5 words, SD = 27.1 words). The emotional contents of the summaries were analyzed by the NRC emotion lexicon [69] and calculated by a method similar to Jagiello and Hills [12] which focuses on the proportion of an emotion relative to the total of all emotions in a given taxonomy (e.g., positive relative to overall negative and positive emotions; fear relative to overall fear, joy, surprise, anger, disgust, and sadness). Participants’ summaries of the neutral tweet set were more emotional than the original tweets. Note that the neutral tweet set was designed with negligible emotional content; however, the summaries contained 29.0% negative words and 8.1% fearful words. In contrast, summaries of the threatening tweets had similar emotional content as the original. The threatening tweet set contained 85.7% negative words and 52.9% fearful words. The summaries of these tweets contained 89.9% negative words and 41.4% fearful words.
Another preliminary analysis was to determine if the exposure to fear-charged messages necessarily made participants more fearful. An independent sample t-test revealed that the experimental and control groups experienced similar levels of fear, t(591) = −0.59, p = 0.552, 95% CI [−0.13, 0.07], d = 0.05. Therefore, the group assignment did not directly translate into subjective fear; thus, the two factors were entered as independent predictors in two-way ANOVAs for the upcoming hypothesis testing.
Two-way ANOVAs were conducted on three indicators—the intent to disseminate the tweets, the number of words written, and the time to completion in the summary task. The analysis of the first indicator showed a significant main effect of fear, F(1, 589) = 25.68, p < 0.001, ηp2 = 0.041, but no other significant main or interaction effects on the intent to disseminate. The left panel of Figure 1 shows participants’ dissemination intentions as a function of their fear. The regression lines (or the main effect of fear) showed a greater interest in disseminating messages of both topics as fear level increased.
The analysis of the second indicator, the total number of words written, found a significant interaction effect, F(1, 589) = 3.90, p = 0.049, ηp2 = 0.006, but no main effects of fear or group assignment. This significant interaction effect can be found in the middle panel of Figure 1. Fearful participants’ summaries were particularly short if they happened to be assigned to work on threatening content.
The analysis of the third indicator, time taken to complete the summary task, found a significant main effect of group assignment, F(1, 589) = 17.76, p < 0.001, ηp2 = 0.029, a significant interaction effect F(1, 589) = 4.20, p = 0.041, ηp2 = 0.007, but no significant main effect of fear. These significant effects are shown in the right panel of Figure 1. The main effect of the group assignment was that the threatening tweets took longer or seemed more daunting to summarize than neutral tweets. The interaction effect was that fearful participants bucked this trend as they uniquely spent less time on the threatening tweets than their less fearful counterparts.

2.3. Discussion

The intention–behavior gap among fearful participants, especially when processing threatening messages (H1), found some support in the data. On the intention front, fearful participants indeed reported a stronger intention to share the messages they encountered, and their heightened sharing intention extended to both neutral and threatening messages. Their broadly engaging mindset, regardless of the emotional contents of the message, went beyond the scope of the negativity bias principle, where negative contents should be the key to attract attention.
In contrast to intentions, behaviors showed a clearer differentiation between message types. Fearful participants were only behaviorally inhibited when engaging with threatening messages. This threat-specific avoidance aligns with the functional perspective on fear, which holds that fear serves an adaptive purpose by promoting avoidance of threatening stimuli. Normative fear—typical fear responses in non-pathological populations—should be stimulus-specific, leading to withdrawal from the threat at hand. Generalized behavioral withdrawal is more characteristic of pathological fear, but even then, their avoidance tends to generalize to a category of stimuli (e.g., claustrophobia and acrophobia) and rarely to all stimuli.

3. Study 2

Since the intention–behavior disjoint was found in Study 1 under a threatening topic, Study 2 aimed to replicate the disjoint in two new threatening topics. Canadian participants were recruited in the early stage of COVID-19 vaccination. At the time of the study, the COVID-19 positivity rate in Canada was 3.7% [75] and the vaccination rate stood at just 1.5% [76]. These low rates highlight the considerable uncertainty surrounding both the spread of the virus and the rollout of its vaccines. Together, they reflect a period in which COVID-19 and its immunization efforts were perceived as emerging risks, marked by limited information and widespread public uncertainty. Fearful participants were expected to show intention–behavior disjoint in both threat scenarios (H1). Furthermore, two additional hypotheses (H2 and H3) were tested to accommodate partisan differences in fear responses and in the intention–behavior disjoint.

3.1. Materials and Methods

3.1.1. Participants

The participants were 577 voting-age Canadian citizens (M = 49.9 years, SD = 16.3 years, 45.4% males and 54.6% females). Their party affiliations were primarily the Liberal Party (32.9%), the Conservative Party (22.2%), and the New Democratic Party (12.7%), with close to one-fifth of the participants self-classifying as politically independent (20.5%). A quarter of the participants were Roman Catholic (24.8%), followed by Protestant (20.3%), and Atheist (15.8%). A great majority of the participants self-identified as White (73.0%).

3.1.2. Procedure

The procedure is similar to Study 1, except that participants reported their political ideologies before they were randomly given messages on COVID-19 or its vaccines. The two sets of tweets were matched in terms of word count (135 words) and also in terms of negativity/fear sentiments. According to the NRC emotion lexicon [69], each set contained eight negative words (and one positive word); eight fear-laden words (and no more than three words of each of the other discrete emotions). A sample COVID-19 tweet is “WHO warned that the new COVID-19 variants have already infiltrated North America.” A sample vaccine tweet is, “Two Canadian health workers showed mysterious side effects after taking the first dose of vaccination.”
The messages were presented one at a time, such that the minimum response required for each message was set at three seconds and a minimum of five characters on Qualtrics. Data from 23 participants were excluded (13 due to recent positive COVID-19 test results, and 10 others were dropped due to recent COVID-19 vaccination), as their experiences might shape distinct attitudes toward the threats. Additional data exclusion strategies are the same as those in Study 1, such that the data from 22 participants were discarded due to straight-lined or irrelevant summaries.

3.1.3. Instruments

Dissemination intent, words written, and task completion time were measured in a similar manner to Study 1. The fear scale was the same as in Study 1 and again demonstrated high internal consistency (α = 0.98). A composite score for political ideology was created by aggregating the componential beliefs of conservatism—right-wing authoritarianism (RWA), nationalism, and neoliberalism—such that high scores indicate conservative leaning and low scores indicate liberal leaning.
RWA. Similarly to Cheung-Blunden [77], 11 items were randomly drawn from Altemeyer’s [44] 30 original items. A 6-point Likert scale was used, and the internal consistency was satisfactory (α = 0.78).
Nationalism. The construct was operationalized by an 8-item measure from Cheung-Blunden [68]. The items asked participants to rate, on a scale from 1 (not important) to 4 (very important), whether a list of nativist criteria should be used to judge a person as genuinely Canadian. Sample criteria include “born in Canada” and “master the English language.” The measure demonstrated satisfactory internal consistency (α = 0.77), comparable to its previously reported range, which was 0.76 to 0.82 [68].
Neoliberalism. The construct was operationalized using 16 items related to dirigisme [55]. The items were taken verbatim from the German Socio-Economic Panel, and sample items included “fighting unemployment” and “taking care of the old.” A neoliberal participant would have a high score on the scale that ranged from 1 (only the state) to 5 (only private forces). The measure demonstrated satisfactory internal consistency (α = 0.90), comparable to its previous range between 0.88 and 0.94 [68].

3.2. Results

Descriptive statistics showed that the average summary of the 135-word tweet set was about half the original length (M = 78.5 words, SD = 36.8 words). The emotional contents of the summaries were calculated in a similar fashion to Study 1. The original messages were designed with 88.9% negative words per set. In response, participants wrote more muted summaries using 65.2% and 45.7% negative words in the two topics, respectively. In terms of fear contents, each of the original tweet sets contained 47.1% fear words. Participants’ summaries were again more modulated than the originals, using 35.4% and 30.3% fear words on the two topics, respectively.

3.2.1. Intention–Behavior Disjoint

H1 was tested in the same manner as Study 1. Two-way ANOVAs were conducted on the intention indicator (intention to disseminate) and the behavioral indicators (number of words written and time used). In terms of the first indicator, participants’ intention to propagate tweets showed signs of engagement as fear increased. There was a significant main effect of fear F(1, 573) = 12.62, p < 0.001, ηp2 = 0.029, a significant main effect of group assignment F(1, 573) = 19.12, p < 0.001, ηp2 = 0.024, but no significant interaction. As shown in the left panel of Figure 2, fearful participants were eager to disseminate the information on both threatening topics (main effect of fear). The COVID-19 messages were considered more worthy to disseminate than vaccine tweets (main effect of group assignment). The lack of interaction suggests that fear’s engaging effect on participants’ intention was generalizable across the two threat types.
The analysis of behavioral indicators found signs of disengagement among fearful participants. In terms of the first behavioral indicator, total number of words written, there was a significant main effect of fear, F(1, 573) = 6.23, p = 0.013, ηp2 = 0.014, a significant main effect of group assignment, F(1, 573) = 8.26, p = 0.004, ηp2 = 0.010, but no significant interaction. The analysis of the second behavioral indicator, time to complete the summary task, found a significant main effect of fear, F(1, 573) = 7.36, p = 0.007, ηp2 = 0.016, a significant main effect of group assignment, F(1, 573) = 6.00, p = 0.015, ηp2 = 0.007, but no significant interaction. These behavioral lags in fearful participants can be found in the middle and right panels of Figure 2. The negative slopes of the trend lines illustrate that the summaries become shorter and quicker as fear increases.

3.2.2. Generalizability of Intention–Behavior Disjoint Across Political Divide

The political ideology composite score was used to explore whether participants on the two sides of the political divide approached the threats of COVID-19 and vaccination differently. H2 pertains to the politically divided risk appetite of the two threats. H3 goes beyond threat response and examines if the intention–behavior disjoint is reshaped by political orientation.
To test H2, two-way ANOVAs were conducted with group assignment and the right-wing ideology composite as predictors of fear. Results showed a significant main effect of group assignment F(1, 573) = 57.51, p < 0.001, ηp2 = 0.088, a significant main effect of right-wing ideology F(1, 573) = 16.84, p < 0.001, ηp2 = 0.027, but no significant interaction. To follow up on the significant main effects, mean-level comparisons showed that the vaccine group reported significantly higher levels of fear than the COVID-19 group (Mvaccine = 2.31, SDvaccine = 1.07; MCOVID-19 = 1.69, SDCOVID-19 = 0.90); conservatives were significantly more fearful than liberals (Mconservative = 2.15, SDconservative = 1.08; Mliberal = 1.83, SDliberal = 0.96). Without a significant interaction effect, H2 was not supported. Meaning, no opposite risk appetites to the two threats were found across the political divide. Instead, conservatives showed more fear under both threats than liberals, a finding that attests to their general risk aversion. To illustrate the findings of H2, 2D density plots were superimposed onto the three panels in Figure 2. Note that conservatives’ risk aversion can be seen in the rightward placement of red/dark gray contour lines toward the high fear zone, whereas their liberal counterparts’ blue/light gray contour lines can be found toward the low fear zone.
H3 was tested by conducting three-way ANOVAs on the intention and behavioral indicators. Special attention was paid to the main effects or interaction effects that involve political ideology. Political ideology showed several significant main effects. It was a significant predictor for the intent to disseminate information, F(1, 569) = 8.84, p = 0.003, ηp2 = 0.015, and for the number of words written, F(1, 569) = 10.58, p = 0.001, ηp2 = 0.016, but not for task completion time. Mean level differences showed that conservatives intended to spread significantly more tweets (Mconservative = 2.4, SDconservative = 3.2; Mliberal = 1.7, SDliberal = 2.8) but wrote significantly fewer words to summarize the tweets (Mconservative = 72.6, SDconservative = 33.0; Mliberal = 82.0, SDliberal = 31.0). Political ideology did not show any significant two-way or three-way interactions. The lack of interactions means that political ideology did not reshape the earlier results, leaving conservatives to mostly follow the patterns of fearful individuals and liberals to mostly follow the patterns of less fearful individuals. Therefore, the results on the intention–behavior disjoint were more straightforward than hypothesized in H3; the disjoint is generalizable across the political spectrum.

3.3. Discussion

Study 2 successfully replicated the intention–behavior disjoint (H1) using two new threat scenarios. In support of this hypothesis, a disjoint was found in both the COVID-19 and the COVID-19 vaccines conditions. Fearful participants were engaged intention-wise to disseminate the messages, but disengaged behaviorally, as indicated by their shorter and quicker summaries. The intention indicator is consistent with the negative bias principle, and the behavioral indicators are in concert with the functionalist approach.
Political ideology was hypothesized to change/moderate risk response (H2) and the intention–behavior disjoint (H3). H2 was partially supported in the present study. Conservatives were more fearful than liberals, but this politically divided risk appetite turned out to apply to both the COVID-19 condition and the COVID-19 vaccine condition. The downstream effect of conservatives’ fear sensitivity was a degree of intention–behavior disjoint that is commensurate with their heightened fear. Meaning, conservatives’ intention engagement and behavioral disengagement were both more pronounced (i.e., a larger intention–behavior disjoint), which is typical of fearful participants. Conversely, liberals showed relatively smaller intention–behavior disjoints as expected of less fearful participants. Overall, the results were more straightforward than anticipated in H2 and H3. The only significant finding was the conservatives’ heightened risk response. This heightened response did not differ across the threat types and incurred no moderating downstream effects on the intention–behavior disjoint outside the typical intention–behavior signature of a fearful individual.

4. General Discussion

In the present studies, two prominent emotion theories were applied to understand fear as an affective engagement tool on social media. The negative bias principle suggests that negative messages can capture users’ attention, likely extending the transmission of the messages. On the other hand, the functionalist framework suggests that fear motivates avoidance action tendencies, potentially truncating the propagation of fear-laden messages. The seemingly opposite predictions of the two theories were reconciled by deconstructing the dissemination process into its separate stages. Both theories found support in the present studies. Following the negative bias principle, fearful participants showed a greater intention to spread threatening messages. As anticipated by the functionalist approach, the same participants became more disengaged when summarizing the threatening information. Our finding of an intention–behavior disjoint is a result of embracing opposing theoretical predictions to explain the human factors in the dissemination process.
An alternative explanation for why fearful participants were behaviorally disengaged with threatening messages during the summary task is that they may have quickly produced more black-and-white reductionist summaries, which is seen as a sign of engagement in the studies of message amplification. Jagiello and Hills [12] observed that messages often became more emotionally charged and polarized during transmission, thereby adding to the momentum of the messages’ propagation. However, ad hoc analyses in our study found similar results to Moussaïd et al. [13], where participants tended to generate summaries with a more muted emotional tone than the original messages. Thus, our participants did not show the alternative form of engagement via poignant summaries.
The intention–behavior disjoint was generalizable across three threat scenarios presented in the two studies (border insecurity, COVID-19, and its vaccines). Fearful participants showed the anticipated engaging intentions and disengaging behaviors in all three threatening topics. However, they also unexpectedly responded to the neutral topic in Study 1 with engaging intentions. Fearful participants’ indiscriminate attention to both threatening and neutral messages is inconsistent with the predictions of the negative bias principle. The negative bias principle underscores the objective emotional content as the key to users’ attention. Since our neutral messages were designed with little affect, we believe the key to their attention is the subjective fear they experience. When users feel afraid—possibly due to external factors unrelated to the neutral messages they read—their attention heightens, making the content they read seem more significant. The difference between the objective feature of a message and the subjective experience of the user is one theoretical nuance to ponder in the future.
The intention–behavior disjoint was applicable across the political spectrum. Study 2 directly compared conservatives’ reactions to two distinct threats: COVID-19 and vaccination, and expected the opposite/topic-specific fear responses from conservatives and liberals [56]. The results were far more straightforward than hypothesized, with conservatives exhibiting a generalized heightened risk aversion to both topics. The generalized risk aversion of conservatives is in concert with Altemeyer’s [44] theory on RWA. Moreover, political ideology did not moderate/distort the intention–behavior disjoint beyond what was expected of their fear levels. Conservatives showed a larger intention–behavior disjoint because their fear responses were elevated. Conversely, liberals, who reported lower levels of fear, exhibited a smaller gap between their sharing intentions and behaviors.

4.1. Limitations

The results of the present studies may be limited to chat rooms and microblogs, where messages are often conveyed through the retelling of stories or commenting on posts. Popular social media platforms such as Twitter/X, Instagram, and Facebook differ in the behavioral demands of sharing. While one platform may have greater behavioral demands, such as writing and posting summaries of the message from another source, other platforms may require only a single click to disseminate a published post (e.g., a single click to reshare a post or keyboard shortcuts to copy and paste a URL). The effect of intention–behavior dissonance may be smaller on platforms with low behavioral demands; however, all platforms are worthy of future research due to their sheer size and reach. A small effect on a large user base may amount to significant differences in political outreach and election results.

4.2. Future Directions

Intentions can fail to translate into actions in everyday life, as evidenced by empty promises or unfulfilled New Year’s resolutions. Under this light, the narrative in fear communication that messages with the power to pique users’ interests would necessarily go viral seems counterintuitive. The intention–behavior disjoint not only challenged this narrative, but also alludes to other points of disconnect from the word embeddings of a message to its actual dissemination. The present studies controlled for the emotionality of the language used, but participants did not respond with the same level of emotion—even when exposed to messages on the same topic, let alone across different topics. These results call for the re-examination of such assumptions as emotional word embeddings in posts would evoke matching emotional responses, and messages with similar emotional intensity would elicit similar emotional reactions regardless of topic.

5. Conclusions

The intention–behavior disjoint found in the present research has important practical and theoretical implications. First, in the context of risk communication, our results highlight the critical need to craft messages that engage audiences emotionally without fomenting too much fear. While fear can capture attention and even motivate initial intent, it may also suppress follow-through behavior, particularly when individuals face more dauting sharing demands of commenting or re-telling. This dynamic is especially relevant in the design of viral campaigns, where emotional engagement must be calibrated to encourage meaningful action rather than passive consumption or withdrawal. Second, the study underscores a key gap in the literature: the tendency to conflate sharing intention with actual behavior. This conflation can set up a false dichotomy between the negative bias principle and the functionalist approach to emotions when the two theories apply to distinct stages of content interaction. The negative bias principle explains readership through the captivating quality of negative content, whereas the functionalist approach is better positioned to predict the sharing behaviors from a particular emotional state. The conflation can also obscure important differences in past findings, as studies relying solely on intent or behavior as outcome measures can arrive at divergent conclusions. By focusing on fear as a case in point, our work contributes to a clearer distinction between affective engagement indicators, helping to reconcile some of the inconsistencies in social media research.

Author Contributions

V.C.-B. was the principal investigator of this project, overseeing and partaking in all phases from conceptualization to manuscript preparation. E.A.Z. contributed significantly to a previous project that inspired the current project and also to the literature review and writing—original draft preparation, review, and editing. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by Division 48 of the American Psychological Association under its small grants program and by the University of San Francisco under its Faculty Development Grant.

Institutional Review Board Statement

The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of the University of San Francisco (protocol IRB000555 and revisions approved on 7 May 2018 and 1 February 2021).

Informed Consent Statement

Informed consent was obtained from all subjects involved in the study.

Data Availability Statement

Research materials and data were accessed on 16 August 2025 at https://osf.io/aqfge/?view_only=051de7588c754ad08d0a3afb171572e3.

Acknowledgments

The authors would like to thank Chonhei Liao, Vivienne Minaglia, and Madelyn Wu for their valuable contributions to completing this work, particularly in their meticulous proofreading and careful preparation of citations.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Office of the Historian. U.S. Diplomacy and Yellow Journalism, 1895–1898. Available online: https://history.state.gov/milestones/1866-1898/yellow-journalism (accessed on 12 July 2025).
  2. Statista. Number of Internet and Social Media Users Worldwide as of February 2025. Available online: https://www.statista.com/statistics/617136/digital-population-worldwide/ (accessed on 12 July 2025).
  3. Palmer, A. Twitter Misses Earnings Expectations, Partially Blames Revenue Drop on Elon Musk Takeover Bid. Available online: https://www.cnbc.com/2022/07/22/twitter-twtr-earnings-q2-2022.html (accessed on 24 February 2023).
  4. Shehata, W.M.; Abdeldaim, D.E. Social media and spreading panic among adults during the COVID-19 Pandemic, Egypt. Environ. Sci. Pollut. Res. 2022, 29, 23374–23382. [Google Scholar] [CrossRef] [PubMed]
  5. Yao, L.; Ngai, C.S.B.; Singh, R.G.; Chen, F. Social media users’ engagement with fear appeal elements in government’s health crisis communication via state-owned media. J. Health Commun. 2024, 29, 524–537. [Google Scholar] [CrossRef] [PubMed]
  6. Marcus, G.E. Emotions in politics. Annu. Rev. Political Sci. 2000, 3, 221–250. [Google Scholar] [CrossRef]
  7. Ruiter, R.A.C.; Abraham, C.; Kok, G. Scary warnings and rational precautions: A review of the psychology of fear appeals. Psychol. Health 2001, 16, 613–630. [Google Scholar] [CrossRef]
  8. Vaish, A.; Grossmann, T.; Woodward, A. Not all emotions are created equal: The negativity bias in social-emotional development. Psychol. Bull. 2008, 134, 383–403. [Google Scholar] [CrossRef]
  9. Voorveld, H.A.M.; van Noort, G.; Muntinga, D.G.; Bronner, F. Engagement with social media and social media advertising: The differentiating role of platform type. J. Advert. 2018, 47, 38–54. [Google Scholar] [CrossRef]
  10. Vosoughi, S.; Roy, D.; Aral, S. The spread of true and false news online. Science 2018, 359, 1146–1151. [Google Scholar] [CrossRef]
  11. Cheung-Blunden, V.; Sonar, K.U.; Zhou, E.A.; Tan, C. Foreign disinformation operation’s affective engagement: Valence versus discrete emotions as drivers of tweet popularity. Anal. Soc. Issues Public Policy 2021, 21, 980–997. [Google Scholar] [CrossRef]
  12. Jagiello, R.D.; Hills, T.T. Bad news has wings: Dread risk mediates social amplification in risk Communication. Risk Anal. 2018, 38, 2193–2207. [Google Scholar] [CrossRef]
  13. Moussaïd, M.; Brighton, H.; Gaissmaier, W. The amplification of risk in experimental diffusion chains. Proc. Natl. Acad. Sci. USA 2015, 112, 5631–5636. [Google Scholar] [CrossRef]
  14. Oh, H.; Goh, K.-Y.; Phan, T.Q. Are you what you tweet? The impact of sentiment on digital news consumption behaviors. Inf. Syst. Res. 2022, 33, 1007–1028. [Google Scholar] [CrossRef]
  15. Pivecka, N.; Ratzinger, R.A.; Florack, A. Emotions and virality: Social transmission of political messages on Twitter. Front. Psychol. 2022, 13, 931921. [Google Scholar] [CrossRef] [PubMed]
  16. Smith, S.M.; Petty, R.E. Message framing and persuasion: A message processing analysis. Pers. Soc. Psychol. Bull. 1996, 22, 257–268. [Google Scholar] [CrossRef]
  17. Stieglitz, S.; Dang-Xuan, L. Political communication and influence through microblogging–An empirical analysis of sentiment in Twitter messages and retweet behavior. In Proceedings of the 2012 45th Hawaii International Conference on System Sciences, Maui, HI, USA, 4–7 January 2012; pp. 3500–3509. [Google Scholar] [CrossRef]
  18. Tellis, G.J.; MacInnis, D.J.; Tirunillai, S.; Zhang, Y. What drives virality (sharing) of online digital content? The critical role of information, emotion, and brand prominence. J. Mark. 2019, 83, 1–20. [Google Scholar] [CrossRef]
  19. Cheung-Blunden, V.; Ju, J. Anxiety as a barrier to information processing in the event of a cyberattack. Political Psychol. 2016, 37, 387–400. [Google Scholar] [CrossRef]
  20. Pratto, F.; John, O.P. Automatic vigilance: The attention-grabbing power of negative social information. J. Pers. Soc. Psychol. 1991, 61, 380–391. [Google Scholar] [CrossRef]
  21. Milosavljevic, M.; Navalpakkam, V.; Koch, C.; Rangel, A. Relative Visual Saliency Differences Induce Sizable Bias in Consumer Choice. J Consum. Psychol. 2012, 22, 67–74. [Google Scholar] [CrossRef]
  22. Vanunu, Y.; Hotaling, J.M.; Le Pelley, M.E.; Newell, B.R. How Top-down and Bottom-up Attention Modulate Risky Choice. Proc. Natl. Acad. Sci. USA 2012, 118, e2025646118. [Google Scholar] [CrossRef]
  23. Katsuki, F.; Constantinidis, C. Bottom-up and Top-down Attention: Different Processes and Overlapping Neural Systems. Neuroscientist 2014, 20, 509–521. [Google Scholar] [CrossRef]
  24. Wang, X.R.; Nan, X.; Stanley, S.; Wang, Y.; Waks, L.; Broniatowski, D. Emotion and Virality of Food Safety Risk Communication Messages on Social Media. J. Appl. Commun. 2021, 105, 3. [Google Scholar] [CrossRef]
  25. Fiske, S.T. Attention and weight in person perception: The impact of negative and extreme behavior. J. Pers. Soc. Psychol. 1980, 38, 889–906. [Google Scholar] [CrossRef]
  26. Frijda, N.H.; Parrott, W.G. Basic emotions or ur-emotions? Emot. Rev. 2011, 3, 406–415. [Google Scholar] [CrossRef]
  27. Hansen, C.H.; Hansen, R.D. Finding the face in the crowd: An anger superiority effect. J. Pers. Soc. Psychol. 1988, 54, 917–924. [Google Scholar] [CrossRef] [PubMed]
  28. Lazarus, R.S. Stress and Emotion: A New Synthesis; Springer Publishing: New York, NY, USA, 1999. [Google Scholar]
  29. Levenson, R.W. Basic emotion questions. Emot. Rev. 2011, 3, 379–386. [Google Scholar] [CrossRef]
  30. Peeters, G.; Czapinski, J. Positive-negative asymmetry in evaluations: The distinction between affective and informational negativity effects. Eur. Rev. Soc. Psychol. 1990, 1, 33–60. [Google Scholar] [CrossRef]
  31. Roseman, I.J.; Wiest, C.; Swartz, T.S. Phenomenology, behaviors, and goals differentiate discrete emotions. J. Pers. Soc. Psychol. 1994, 67, 206–221. [Google Scholar] [CrossRef]
  32. Scherer, K.R.; Ceschi, G. Lost luggage: A field study of emotion–antecedent appraisal. Motiv. Emot. 1997, 21, 211–235. [Google Scholar] [CrossRef]
  33. Tooby, J.; Cosmides, L. The past explains the present: Emotional adaptations and the structure of ancestral environments. Ethol. Sociobiol. 1990, 11, 375–424. [Google Scholar] [CrossRef]
  34. Baumeister, R.F.; Bratslavsky, E.; Finkenauer, C.; Vohs, K.D. Bad is stronger than good. Rev. Gen. Psychol. 2001, 5, 323–370. [Google Scholar] [CrossRef]
  35. Rozin, P.; Royzman, E.B. Negativity bias, negativity dominance, and contagion. Personal. Soc. Psychol. Rev. 2001, 5, 296–320. [Google Scholar] [CrossRef]
  36. Roelofs, K. Freeze for action: Neurobiological mechanisms in animal and human freezing. Philos. Trans. R. Soc. B Biol. Sci. 2017, 372, 20160206. [Google Scholar] [CrossRef]
  37. Scarpa, A.; Raine, A. Psychophysiology of anger and violent behavior. Psychiatr. Clin. N. Am. 1997, 20, 375–394. [Google Scholar] [CrossRef] [PubMed]
  38. Cheung-Blunden, V.; Blunden, B. The emotional construal of war: Anger, fear, and other negative emotions. Peace Confl. J. Peace Psychol. 2008, 14, 123–149. [Google Scholar] [CrossRef]
  39. Amir, N.; Kuckertz, J.M.; Najmi, S. The effect of modifying automatic action tendencies on overt avoidance behaviors. Emotion 2013, 13, 478–484. [Google Scholar] [CrossRef] [PubMed]
  40. Glotzbach, E.; Ewald, H.; Andreatta, M.; Pauli, P.; Mühlberger, A. Contextual fear conditioning predicts subsequent avoidance behaviour in a virtual reality environment. Cogn. Emot. 2012, 26, 1256–1272. [Google Scholar] [CrossRef]
  41. Kreibig, S.D.; Wilhelm, F.H.; Roth, W.T.; Gross, J.J. Cardiovascular, electrodermal, and respiratory response patterns to fear- and sadness-inducing films. Psychophysiology 2007, 44, 787–806. [Google Scholar] [CrossRef]
  42. Perkins, A.M.; Cooper, A.; Abdelall, M.; Smillie, L.D.; Corr, P.J. Personality and defensive reactions: Fear, trait anxiety, and threat magnification. J. Pers. 2010, 78, 1071–1090. [Google Scholar] [CrossRef]
  43. Shen, L.; Bigsby, E. Behavioral activation/inhibition systems and emotions: A test of valence vs. action tendency hypotheses. Commun. Monogr. 2010, 77, 1–26. [Google Scholar] [CrossRef]
  44. Tovote, P.; Fadok, J.P.; Lüthi, A. Neuronal circuits for fear and anxiety. Nat. Rev. Neurosci. 2015, 16, 317–331. [Google Scholar] [CrossRef]
  45. Mauss, I.B.; Levenson, R.W.; McCarter, L.; Wilhelm, F.H.; Gross, J.J. The tie that binds? Coherence among emotion experience, behavior, and physiology. Emotion 2005, 5, 175–190. [Google Scholar] [CrossRef]
  46. Ridout, T.N.; Searles, K. It’s my campaign I’ll cry if I want to: How and when campaigns use emotional appeals. Political Psychol. 2011, 32, 439–458. [Google Scholar] [CrossRef]
  47. Song, Y.; Wang, X.; Jia, Y. Deep learning-based COVID-19 Twitter analysis. In Proceedings of the 2021 6th International Conference on Big Data and Computing, Shenzhen, China, 22–24 May 2021; ACM: New York, NY, USA, 2021; pp. 8–14. [Google Scholar] [CrossRef]
  48. Altemeyer, B. Enemies of Freedom: Understanding Right-Wing Authoritarianism, 1st ed.; Jossey-Bass: San Francisco, CA, USA, 1988. [Google Scholar]
  49. Butler, J.C. Authoritarianism and fear responses to pictures: The role of social differences. Int. J. Psychol. 2013, 48, 18–24. [Google Scholar] [CrossRef]
  50. Eigenberger, M.E. Fear as a correlate of authoritarianism. Psychol. Rep. 1998, 83, 1395–1409. [Google Scholar] [CrossRef]
  51. Jost, J.T.; Napier, J.L.; Thorisdottir, H.; Gosling, S.D.; Palfai, T.P.; Ostafin, B. Are needs to manage uncertainty and threat associated with political conservatism or ideological extremity? Pers. Soc. Psychol. Bull. 2007, 33, 989–1007. [Google Scholar] [CrossRef] [PubMed]
  52. Roccato, M.; Russo, S. Right-wing authoritarianism, societal threat to safety, and psychological distress. Eur. J. Soc. Psychol. 2017, 47, 600–610. [Google Scholar] [CrossRef]
  53. Kam, C.D.; Simas, E.N. Risk orientations and policy frames. J. Politics 2010, 72, 381–396. [Google Scholar] [CrossRef]
  54. Harvey, D. A Brief History of Neoliberalism; Oxford University Press: Oxford, UK, 2007. [Google Scholar]
  55. McDonald, M.; Gough, B.; Wearing, S.; Deville, A. Social psychology, consumer culture and neoliberal political economy. J. Theory Soc. Behav. 2017, 47, 363–379. [Google Scholar] [CrossRef]
  56. Schreiber, D.; Fonzo, G.; Simmons, A.N.; Dawes, C.T.; Flagan, T.; Fowler, J.H.; Paulus, M.P. Red brain, blue brain: Evaluative processes differ in Democrats and Republicans. PLoS ONE 2013, 8, e52970. [Google Scholar] [CrossRef]
  57. Baumgaertner, B.; Carlisle, J.E.; Justwan, F. The influence of political ideology and trust on willingness to vaccinate. PLoS ONE 2018, 13, e0191728. [Google Scholar] [CrossRef]
  58. Gugushvili, A.; Koltai, J.; Stuckler, D.; McKee, M. Votes, populism, and pandemics. Int. J. Public Health 2020, 65, 721–722. [Google Scholar] [CrossRef]
  59. Speed, E.; Mannion, R. Populism and health policy: Three international case studies of right-wing populist policy frames. Sociol. Health Illn. 2020, 42, 1967–1981. [Google Scholar] [CrossRef]
  60. Ware, A. Political Parties and Party Systems, 1st ed.; Oxford University Press: Oxford, UK; New York, NY, USA, 1995. [Google Scholar]
  61. Fridman, A.; Gershon, R.; Gneezy, A. COVID-19 and Vaccine Hesitancy: A Longitudinal Study. PLoS ONE 2021, 16, e0250123. [Google Scholar] [CrossRef] [PubMed]
  62. Agius, C.; Rosamond, A.B.; Kinnvall, C. Populism, ontological insecurity and gendered nationalism: Masculinity, climate denial and Covid-19. Politics Relig. Ideol. 2020, 21, 432–450. [Google Scholar] [CrossRef]
  63. Jones-Jang, S.M.; Chung, M. Can we blame social media for polarization? Counter-evidence against filter bubble claims during the COVID-19 pandemic. New Media Soc. 2022, 26, 3370–3389. [Google Scholar] [CrossRef]
  64. Leonhardt, D. Covid’s Partisan Errors. Available online: https://www.nytimes.com/2021/03/18/briefing/atlanta-shootings-kamala-harris-tax-deadline-2021.html (accessed on 1 March 2022).
  65. Winter, T.; Riordan, B.C.; Pakpour, A.H.; Griffiths, M.D.; Mason, A.; Poulgrain, J.W.; Scarf, D. Evaluation of the english version of the fear of COVID-19 scale and its relationship with behavior change and political beliefs. Int. J. Ment. Health Addict. 2023, 21, 372–382. [Google Scholar] [CrossRef]
  66. Clarke, E.J.R.; Ling, M.; Kothe, E.J.; Klas, A.; Richardson, B. Mitigation system threat partially mediates the effects of right-wing ideologies on climate change beliefs. J. Appl. Soc. Psychol. 2019, 49, 349–360. [Google Scholar] [CrossRef]
  67. Eadeh, F.R.; Chang, K.K. Can threat increase support for liberalism? New insights into the relationship between threat and political attitudes. Soc. Psychol. Personal. Sci. 2020, 11, 88–96. [Google Scholar] [CrossRef]
  68. Austen, I.; Bilefsky, D. Arrests in Terror Plot Raise Questions About Canada’s Refugee Embrace; The New York Times: New York, NY, USA, 2019. [Google Scholar]
  69. Mohammad, S.M.; Turney, P.D. Crowdsourcing a Word-Emotion Association Lexicon. Available online: https://arxiv.org/abs/1308.6297v1 (accessed on 25 February 2023).
  70. Xue, J.; Chen, J.; Hu, R.; Chen, C.; Su, Y.; Zhu, T. Twitter Discussions and Emotions About the COVID-19 Pandemic: Machine Learning Approach. J. Med. Internet Res. 2020, 22, e20550. [Google Scholar] [CrossRef]
  71. Hu, T.; Wang, S.; Luo, W.; Zhang, M.; Huang, X.; Yan, Y.; Liu, R.; Ly, K.; Kacker, V.; She, B.; et al. Revealing Public Opinion Towards COVID-19 Vaccines with Twitter Data in the United States: Spatiotemporal Perspective. J. Med. Internet Res. 2021, 23, e30854. [Google Scholar] [CrossRef]
  72. Wilcox, R.R. Introduction to Robust Estimation and Hypothesis Testing; Academic Press: Amsterdam, NY, USA, 2012. [Google Scholar]
  73. Kim, Y.; Dykema, J.; Stevenson, J.; Black, P.; Moberg, D.P. Straightlining: Overview of Measurement, Comparison of Indicators, and Effects in Mail–Web Mixed-Mode Surveys. Soc. Sci. Comput. Rev. 2019, 37, 214–233. [Google Scholar] [CrossRef]
  74. Cheung-Blunden, V.; Cropper, K.; Panis, A.; Davis, K. Functional Divergence of Two Threat-Induced Emotions: Fear-Based versus Anxiety-Based Cybersecurity Preferences. Emotion 2019, 19, 1353–1365. [Google Scholar] [CrossRef]
  75. Government of Canada. COVID-19 Epidemiology Update: Summary. Available online: https://health-infobase.canada.ca/covid-19/ (accessed on 25 February 2023).
  76. CBC News. Track How Many People Have Been Given the COVID-19 Vaccine Across Canada. Available online: https://www.cbc.ca/news/canada/track-how-many-people-have-been-given-the-covid-19-vaccine-across-canada-1.5870573 (accessed on 25 February 2023).
  77. Cheung-Blunden, V. Situational Insecurity versus Entrenched Ideologies as the Source of Right-Wing Voters’ Anti-Migrant Sentiment on Both Sides of the Atlantic. J. Appl. Soc. Psychol. 2020, 50, 337–350. [Google Scholar] [CrossRef]
Digital 05 00052 g001
Figure 1. Engaging intentions and disengaging behaviors in Study 1. The intention–behavior disjoint is shown by the opposite effects of fear on intention (left panel) and behaviors (middle and right panels), which is particularly pronounced in the threatening topic of border insecurity. For the neutral topic, more fearful participants engaged in terms of intention, but did not behave differently from their less fearful counterparts. Taken together, the positive-sloped regression line for intentions (left panel) and the negative-sloped regression line for behaviors (middle and right panels) pointed to an intention–behavior disjoint as participants experienced more fear. The intention–behavior disjoint was the most evident (in support of H1) under the threatening condition.
Figure 1. Engaging intentions and disengaging behaviors in Study 1. The intention–behavior disjoint is shown by the opposite effects of fear on intention (left panel) and behaviors (middle and right panels), which is particularly pronounced in the threatening topic of border insecurity. For the neutral topic, more fearful participants engaged in terms of intention, but did not behave differently from their less fearful counterparts. Taken together, the positive-sloped regression line for intentions (left panel) and the negative-sloped regression line for behaviors (middle and right panels) pointed to an intention–behavior disjoint as participants experienced more fear. The intention–behavior disjoint was the most evident (in support of H1) under the threatening condition.
Digital 05 00052 g001
Digital 05 00052 g002
Figure 2. General and politically divided intention–behavior disjoint in Study 2. The disjoint is shown by the opposite directions of the regression lines, where the dissemination intention significantly increased as a function of fear, but the behaviors significantly decreased as a function of fear. The intention–behavior disjoint was generalizable across the two threat scenarios and also across the political divide. Political orientation is a continuous variable in the statistical tests, but is dichotomized here in order to show the visual placements of conservatives (red/dark color) relative to liberals (blue/light color). Across all panels, the rightward placement of conservatives over liberals attests to the greater fear among conservatives (Mconservative = 2.15, SDconservative = 1.08; Mliberal = 1.83, SDliberal = 0.96). Conservatives’ higher placement on the y-axis in the left panel shows their greater dissemination intention (Mconservative = 2.4, SDconservative = 3.2; Mliberal = 1.7, SDliberal = 2.8); and conservatives’ lower placement in the middle panel shows their lackluster word count (Mconservative = 72.6, SDconservative = 33.0; Mliberal = 82.0, SDliberal = 31.0). The two sets of density curves do not seem to differ in the vertical placement in the right panel because the main effect of political ideology was not significant in terms of time completion.
Figure 2. General and politically divided intention–behavior disjoint in Study 2. The disjoint is shown by the opposite directions of the regression lines, where the dissemination intention significantly increased as a function of fear, but the behaviors significantly decreased as a function of fear. The intention–behavior disjoint was generalizable across the two threat scenarios and also across the political divide. Political orientation is a continuous variable in the statistical tests, but is dichotomized here in order to show the visual placements of conservatives (red/dark color) relative to liberals (blue/light color). Across all panels, the rightward placement of conservatives over liberals attests to the greater fear among conservatives (Mconservative = 2.15, SDconservative = 1.08; Mliberal = 1.83, SDliberal = 0.96). Conservatives’ higher placement on the y-axis in the left panel shows their greater dissemination intention (Mconservative = 2.4, SDconservative = 3.2; Mliberal = 1.7, SDliberal = 2.8); and conservatives’ lower placement in the middle panel shows their lackluster word count (Mconservative = 72.6, SDconservative = 33.0; Mliberal = 82.0, SDliberal = 31.0). The two sets of density curves do not seem to differ in the vertical placement in the right panel because the main effect of political ideology was not significant in terms of time completion.
Digital 05 00052 g002
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Cheung-Blunden, V.; Zhou, E.A. When Fear Backfires: How Emotionality Reduces the Online Sharing of Threatening Messages. Digital 2025, 5, 52. https://doi.org/10.3390/digital5040052

AMA Style

Cheung-Blunden V, Zhou EA. When Fear Backfires: How Emotionality Reduces the Online Sharing of Threatening Messages. Digital. 2025; 5(4):52. https://doi.org/10.3390/digital5040052

Chicago/Turabian Style

Cheung-Blunden, Violet, and Emily Ann Zhou. 2025. "When Fear Backfires: How Emotionality Reduces the Online Sharing of Threatening Messages" Digital 5, no. 4: 52. https://doi.org/10.3390/digital5040052

APA Style

Cheung-Blunden, V., & Zhou, E. A. (2025). When Fear Backfires: How Emotionality Reduces the Online Sharing of Threatening Messages. Digital, 5(4), 52. https://doi.org/10.3390/digital5040052

Article Metrics

Back to TopTop