Social Comparison in Narrative Persuasion: When Struggling Characters Serve as Motivation

The effects of gain and loss frames have been examined extensively, but there can be more nuance in health narratives. Experimental research with narratives has not yet thoroughly investigated all combinations of protagonists’ health recommendation compliance and story outcomes. People engaging in healthy behaviors may experience negative outcomes. The current study utilized social cognitive theory to investigate protagonist self-efficacy (low vs. high) and social comparison self-evaluation as moderators of story outcome (positive vs. negative) with sleep and alcohol topics. A three-way interaction indicated that, for the alcohol narratives, the more participants compared themselves to low self-efficacy, negative ending protagonists (vs. low self-efficacy, positive ending protagonists), the greater their positive change in alcohol self-efficacy.

. Experimental research with narratives has not yet thoroughly investigated all possible combinations of protagonists' compliance with health recommendations and story outcomes (Lillie et al., 2022), often focusing on gain and loss frames as traditionally operationalized.
The current study incorporated message framing research and social cognitive theory (Bandura, 1977) to test the effects of story outcome (positive vs. negative) and protagonist self-efficacy (low vs. high) in narratives involving alcohol or sleep behaviors.Alcohol and sleep were target behaviors given their impacts on health (Perry et al., 2013).Lack of sleep negatively influences moods, memory, quality of life, and can lead to excessive alcohol consumption.Together, this study examined the extent to which individuals engage in social comparison self-evaluation (SCSE), producing assimilation or contrast effects, to aid in understanding which story outcomes and protagonists are most persuasive.

Message Frames
Message framing literature has conceptualized gain frames as messages portraying the benefits of engaging in health behaviors and loss frames as those portraying the consequences of ignoring health behaviors (Jensen et al., 2018).Rooted in prospect theory (Kahneman & Tversky, 1984), message frames may include base-rate information individuals interpret to determine risk (Jensen et al., 2018).Prospect theory hypothesizes that individuals will take risks when outcomes are presented as losses, but avoid risks when the same outcomes are presented as gains (Kahneman & Tversky, 1984).Narrative persuasion research (e.g., de Graaf et al., 2015;Robbins & Niederdeppe, 2019) has operationalized message frames with exemplars who experience the benefits or costs of engaging in or ignoring healthy behaviors.
Limited research has examined message framing within the health behaviors of interest in the current study (i.e., sleep, alcohol).Two studies demonstrated that message frame had no main effect on sleep behavior intentions (Li et al., 2022;Robbins & Niederdeppe, 2019).Robbins and Niederdeppe (2019) used narratives, whereas Li et al. (2022) examined cartoon messages and anthropomorphic voices.Two studies investigated message frames with alcohol (de Graaf et al., 2015;Yu et al., 2010).One study reported interaction effects between message frame and issue involvement in narratives, demonstrating that gain frames were more effective in influencing alcohol-related attitudes and intentions among individuals with low issue involvement (de Graaf et al., 2015).Loss frames were more effective for participants with high issue involvement.Another study examined how message frame and appeal (base-rate vs. exemplar) influenced fetal alcohol syndrome disorder (FASD) behavioral intentions (Yu et al., 2010).There was no significant difference between gain and loss frames in intentions to know more about FASD; however, gain frames influenced intentions to prevent FASD.Exemplar, loss frame messages increased perceived severity of alcohol consumption, but base-rate, gain frames had a greater impact on promoting efficacy to avoid FASD.
Studies have attempted to determine which message frame is most effective.Meta-analyses (e.g., O'Keefe & Jensen, 2009) and several studies reviewed in this section (Li et al., 2022;Robbins & Niederdeppe, 2019) have demonstrated no robust differences between gain and loss frames in influencing health-related outcomes.Subsequently, scholars have highlighted the need to identify other theoretical frameworks and moderating variables to understand message framing effects (Jensen et al., 2018).Because health narratives can be more complex than a gain or loss frame (e.g., characters attempting healthy behaviors experience negative endings), the current study drew upon social cognitive theory (SCT; Bandura, 1977) by incorporating protagonist self-efficacy (low vs. high) to extend message framing research.

Protagonist Self-Efficacy
Self-efficacy is the belief that one has the ability to perform behaviors, and plays a critical role in behavior change, per SCT (Bandura, 1977).Without self-efficacy, individuals will likely not attempt behaviors.One way to enhance self-efficacy is through efficacy appeals in narratives where readers access characters' minds, including beliefs in engaging in new behaviors (Robinson & Knobloch-Westerwick, 2022).Characters' motivations should equip readers with the self-efficacy to attempt new behaviors.
Message framing research has investigated self-efficacy as a message feature and moderator (Jensen et al., 2018;Werrij et al., 2011).One study operationalized efficacy appeals as how to engage in physical activity, demonstrating that the selfefficacy message (vs.no self-efficacy) did not interact with message frame (Jensen et al., 2018).Another study operationalized efficacy as how easy it was to find ecological meat in a city (high or low; Werrij et al., 2011).Gain frames were more effective for high self-efficacy participants, who reported consuming ecological meat two weeks later; however, there were no impacts on attitudes or intentions.Low-self efficacy participants reported more defensive message processing when exposed to loss frames.These studies featured general messages of efficacy rather than specific protagonists or exemplars who believed they could engage in the behavior-encompassing the definition of self-efficacy per SCT (Bandura, 1977).
Narrative persuasion research has developed self-efficacy by emphasizing protagonists' confidence when engaging in health behaviors.In one study using Disney movies, protagonists' general self-efficacy (i.e., not toward a particular behavior) was manipulated (low vs. high) in overcoming adversities (Isberner et al., 2019).High self-efficacy protagonists increased participants' general self-efficacy when participants were more transported.Another study manipulated protagonist sleep selfefficacy (low vs. high) in mysteries (Robinson & Knobloch-Westerwick, 2022).Low and high self-efficacy protagonists enhanced participants' sleep self-efficacy immediately postexposure, but only participants in high self-efficacy conditions indicated greater self-efficacy three days later.For the current study, high self-efficacy, positive ending protagonists that solve mysteries by practicing healthy behaviors should increase participants' own self-efficacy, per SCT and message framing research.

Communication Reports 47
Low self-efficacy, negative ending protagonists should also motivate participants by depicting the consequences of protagonists' lack of confidence in practicing healthy behaviors and subsequent failure to solve mysteries.These conditions represent traditional operationalizations of gain and loss frames.
Limited experimental research has investigated how protagonists that adhere to health recommendations, but experience failure (Lillie et al., 2022), or ignore health behaviors and experience success (Nabi & Clark, 2008), influence individuals' health-related outcomes.For example, one study manipulated story outcome, protagonist morality, and adherence to safe sun recommendations, demonstrating that adherence (or lack thereof) did not influence participants' health intentions (Lupfer & Gingrich, 1999).Despite limited research, SCT would predict that a high self-efficacy protagonist who experiences a negative ending should be discouraging and reduce participants' self-efficacy.
When characters ignore health behaviors and experience success, correlational research has demonstrated a positive relationship with viewers' intent to engage in the same negative behaviors (Dal Cin et al., 2007, 2009).One experimental study demonstrated that participants who had never had a one-night stand reported a greater likelihood to engage in this behavior after viewing a serial TV show where one character had a one-night stand with no negative consequences (Nabi & Clark, 2008).In the current study, low self-efficacy protagonists expressed thoughts of uncertainty related to engaging in recommended behaviors, ignored them, and predicted that they would engage in unhealthy behaviors later.Participants who read about a low self-efficacy protagonist who experienced a positive ending may report lower self-efficacy themselves, since the protagonist had success without practicing healthy behaviors.
H1: High self-efficacy, positive ending and low self-efficacy, negative ending narratives will be more effective in influencing participants' health-related self-efficacy than high self-efficacy, negative ending and low self-efficacy, positive ending narratives.

Social Comparison Self-Evaluation
To date, one study has explored social comparison in message framing effects (Hoffner & Ye, 2009).Social comparison self-evaluation (SCSE) is examined because SCT posits that learning from others occurs through self-evaluation processes (see Robinson & Knobloch-Westerwick, 2022 for a review of how SCSE differs from common concepts in narrative persuasion, including transportation and identification).SCSE occurs when individuals compare their behaviors or opinions with those of others to assess how they are doing (Festinger, 1954).Originally, it was proposed that self-evaluation comparison targets must be similar (Festinger, 1954); however, later theorization (e.g., Wood, 1989) indicated that targets perceived as better or worse can be informative.
When individuals compare their behaviors to targets' behaviors, assimilation or contrast effects may occur from engaging in upward or downward comparisons (Wood, 1989).Assimilation occurs when one feels similar to a target, and contrast occurs when one feels different from a target.Comparisons can produce positive and negative contrastive or assimilation effects.
For upward comparisons, role models may be protagonists succeeding at behaviors (Helgeson & Mickelson, 1995), and comparison helps individuals understand how their behaviors measure up, serving as inspiration (Wood, 1989).For example, parents engaging in upward comparisons while viewing social media posts of children demonstrating healthy habits (vs.downward comparison to targets engaging in unhealthy habits) were more likely to experience assimilation, resulting in increased intentions to create healthy habits for their children (Rheu et al., 2023).Upward comparisons can also produce negative effects.For instance, participants with greater social comparison tendencies reported greater perceived risk of getting skin cancer, but lower intentions to use sunscreen after reading gain frame messages on safe sun practices (Hoffner & Ye, 2009).A contrast effect may have occurred where participants reduced risk perceptions by convincing themselves they were better off than the comparison targets, and thus, did not need to change their behaviors.
Downward comparison targets who experience negative outcomes teach individuals what they should not do (Helgeson & Mickelson, 1995).Contrasting with protagonists who experience negative endings could create positive effects.For example, smokers who experience negative health outcomes (e.g., lung cancer) may produce contrast effects, reducing intentions to smoke.Assimilating to negative characters could produce hopeless feelings or reduce the likelihood of change.For example, participants with greater social comparison tendencies reported lower intentions to engage in safe sun behaviors after reading loss frames where exemplars were diagnosed with precancerous lesions (Hoffner & Ye, 2009).Individuals may have assimilated to the loss frame exemplars, producing hopeless feelings.
Narrative persuasion studies (Krakowiak & Tsay-Vogel, 2014;Richter et al., 2015;Robinson & Knobloch-Westerwick, 2022) have examined social comparison as a moderator of effects.Robinson and Knobloch-Westerwick (2022) demonstrated that greater SCSE with high sleep self-efficacy protagonists (vs.low) led to increased self-efficacy.Building on previous research examining selfefficacy (Werrij et al., 2011) and social comparison (Hoffner & Ye, 2009) as moderators of message frame, the current study investigated a three-way interaction.
RQ1: How, if at all, will SCSE moderate the relationship between story outcome, protagonist self-efficacy, and participant self-efficacy to produce assimilation or contrast effects?
The experiment involved a 2 (story outcome: positive vs. negative) × 2 (protagonist self-efficacy: low vs. high) between-subjects design with health topic (alcohol or sleep) as a within-subjects factor.Participants were randomly assigned to one of two stories per health topic.Self-efficacy and story outcome manipulations were the same in both stories.For example, if a participant was randomly assigned to first read an alcohol narrative with a high self-efficacy, positive ending protagonist, then the sleep narrative they read next also featured a high self-efficacy, positive ending protagonist. 3  The survey took about 22 minutes 4 (M = 22.67, SD = 9.25).Participants completed demographics, health behavior items, including alcohol and sleep with distractor behaviors, and social comparison orientation.Baseline self-efficacy for alcohol and sleep were captured with distractor behaviors.After reading each narrative, participants completed manipulation checks and social comparison items.They responded to health-related self-efficacy items and received $3.50 in compensation.Participants passed a single-item attention check.

Stimuli
Four mysteries around 1,400 words each were modified from Robinson and Knobloch-Westerwick (2022; see online Appendix A).Narratives had a protagonist with the same gender as participants.Two stories featured four positive alcohol behaviors (e.g., removing all alcohol from the household, seeking counseling for a drinking problem, etc.) based on recommendations from the National Institute on Alcohol Abuse and Alcoholism (2001).Two stories featured four positive sleep behaviors (e.g., avoiding activities such as using screen devices in bed and consuming alcohol before bed) based on guidelines from the American Academy of Sleep Medicine (2005).It took about six minutes to read the alcohol (M = 6.15,SD = 3.49) and sleep (M = 6.12,SD = 3.57) narratives.Protagonists' health-related self-efficacy (low vs. high) was linked to the story outcome.In the positive endings, protagonists solved murders or caught thieves.In the negative endings, protagonists were unable to identify murderers or thieves escaped.See manipulation checks in the online Appendix B.

Control Variables
The following demographic variables were controlled for due to their association with alcohol dependence (Swendsen et al., 2009) and sleep problems (Grandner et al., 2010): age, gender, education, race and ethnicity, employment and relationship status. 5 Alcohol and sleep behaviors, including the average number of drinks consumed on a typical weekend day/night (M = 2.34, SD = 2.14) and how often alcohol was consumed during the last month, were also controlled for.Responses varied, with 44.6% of participants indicating that they never consumed alcohol in the last month and the remaining participants reporting consuming alcohol at least once a month (11.3%), two to three times a month (11.3%), once a week (5.4%), twice a week (11.7%),three to four times a week (5.9%), five to six times a week (4.1%), and every day (4.1%).For sleep, participants completed five items from the Pittsburgh Insomnia Rating Scale (Moule et al., 2002; α = .92,M = 11.67,SD = 7.61).Social comparison orientation was controlled for with six items (Gibbons & Buunk, 1999; α = .95,M = 20.51,SD = 10.43).

Social comparison self-evaluation
Two items adapted from Robinson and Knobloch-Westerwick (2022) captured alcohol and sleep SCSE (e.g., "I compared my own drinking behaviors to the main character's behaviors to see how I'm doing," "The main character's drinking behaviors in the story provided insight into my own situation").Items were evaluated on Communication Reports 51 a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree), with higher scores representing greater alcohol (r = .91,p < .005;M = 4.42, SD = 3.40) and sleep (r = .86,p < .005;M = 6.03,SD = 3.72) SCSE.

Results
Moderation analyses using Process SPSS macro (version 3.5) Model 3 (Hayes, 2018) were conducted.For alcohol, story outcome was entered as the independent variable, SCSE as the primary moderator, protagonist self-efficacy as the secondary moderator, and participant change in alcohol self-efficacy as the dependent variable.In addition to the control variables mentioned above, story order 7 and average reading time were controlled for (Robinson & Knobloch-Westerwick, 2022).Table 1 contains the full model results.The interaction of story outcome and protagonist selfefficacy was not significant (p = .19),providing no support for H1.
For alcohol, the three-way interaction between story outcome, protagonist selfefficacy, and SCSE was significant (RQ1).High protagonist self-efficacy did not appear to make a difference among participants regardless of story outcome, producing no interaction (p = .87).A significant test of conditional interaction emerged among participants in the low self-efficacy conditions, F(1, 193) = 8.11, p = .005.The more participants engaged in SCSE with low self-efficacy, negative ending protagonists (vs.low self-efficacy, positive ending protagonists), the more they experienced a positive change in alcohol self-efficacy (see Table 1 and Figure 1).
For sleep, message frame was entered as the independent variable, SCSE as the primary moderator, protagonist self-efficacy as the secondary moderator, and participant change in sleep self-efficacy as the dependent variable.The interaction of message frame and protagonist self-efficacy was not significant (p = .50),providing no support for H1.The three-way interaction was also not significant (p = .36),clarifying RQ1.

Discussion
The current study incorporated message framing research and SCT (Bandura, 1977) to test the effects of story outcome (positive vs. negative) and protagonist selfefficacy (low vs. high) in narratives with alcohol or sleep behaviors.Experimental research has not yet thoroughly investigated all possible combinations of protagonists' compliance with health recommendations and story outcomes (Lillie et al., 2022), often focusing on gain and loss frames as traditionally operationalized.A three-way interaction indicated that, for the alcohol narratives, the more participants compared themselves to low self-efficacy, negative ending protagonists (vs.low self-efficacy, positive ending protagonists), the greater their positive change in alcohol self-efficacy, but this interaction did not replicate with sleep narratives.The theoretical implications of these findings are detailed below.
The interaction of story outcome and protagonist self-efficacy on participant selfefficacy was not significant.The lack of a significant interaction effect aligns with Jensen et al. (2018) and partially with Werrij et al. (2011), who detected behavior change, but no attitudinal change.The current study did not include a behavior measure and thus, was unable to determine if increased self-efficacy led to behavioral change, per SCT.Future research should consider including a behavior measure alongside assessing changes in attitudes, such as participant self-efficacy.Identifying SCSE as a moderator of message frame and protagonist self-efficacy addressed calls to examine other theoretical frameworks and moderating variables in order to better understand message framing effects (Jensen et al., 2018).A significant three-way interaction emerged in the alcohol narratives, demonstrating that those who engaged in greater SCSE with low self-efficacy, negative ending protagonists (i.e., loss frames as traditionally operationalized) reported greater positive change in alcohol self-efficacy compared to low self-efficacy, positive ending protagonists, producing contrast effects.These results differ from Hoffner and Ye (2009), who found that greater social comparison tendencies led to lower intentions to engage in safe sun behaviors after reading loss frames where exemplars were diagnosed with precancerous lesions.Thoughts of cancer may have inadvertently led to a mortality salience induction that produced hopeless feelings (Lee & Loiselle, 2012).In the current study, contrast effects may have occurred with alcohol due to the stigma associated with alcoholism (Room, 2005).Alcohol disorders can produce more negative emotions and social rejection than other mental health disorders (Schomerus et al., 2011).Individuals may have disassociated themselves from potentially similar protagonists to protect their own identities (Kim, 2019).Contrasting may have motivated participants to improve their alcohol self-efficacy to avoid experiencing negative consequences like the protagonists.
Narrative persuasion research has examined situations where protagonists' negative health behaviors, including dislikable protagonists (Robinson & Knobloch-Westerwick, 2017) and negative consequences of protagonists' unhealthy actions  (Ma & Nan, 2019), enhance persuasion.The current study contributed to this work by demonstrating that alcohol narratives depicting struggling protagonists with negative endings (i.e., loss frames) increased participants' alcohol-related selfefficacy, particularly among those with greater SCSE.Future research should continue to examine how individuals compare themselves with protagonists demonstrating negative characteristics by varying the severity of their health behaviors and their consequences, especially within stigmatized contexts.
Comparisons to low self-efficacy, positive ending protagonists seemingly resulted in reduced participant alcohol-related self-efficacy with assimilation effects.This finding highlights an area of concern, in line with research that has also demonstrated an increased likelihood to engage in unhealthy behaviors after viewing characters who engage in the same behaviors and experience positive outcomes (Dal Cin et al., 2009;Nabi & Clark, 2008).The current study did not measure behavior, and future research should investigate if viewing negative protagonists who succeed reduces individuals' compliance with health recommendations, especially because this type of protagonist is often represented in the media (Nabi & Clark, 2008).Examining the incongruency between positive or negative story outcomes and protagonists engaging in (or ignoring) health recommendations is an area in need of further investigation.
The three-way interaction between story outcome, protagonist self-efficacy, and SCSE did not replicate with sleep.Perhaps manipulating story outcomes (positive vs. negative) and contrasting with negative protagonists is more relevant or important for behaviors associated with stigma or cessation behaviors (e.g., smoking), as previous framing research has demonstrated gain and loss frames are more effective for certain types of behaviors (Jensen et al., 2018).A post hoc moderation analysis with sleep self-efficacy as the independent variable, SCSE as the moderator, and change in sleep self-efficacy as the dependent variable (controlling for story outcome) replicated results from Robinson and Knobloch-Westerwick (2022; see online Appendix D).The more participants engaged in SCSE with high self-efficacy protagonists, the greater change in sleep self-efficacy reported.Perhaps story ending is not an important factor for health promotion behaviors like sleep.This finding aligns with Robbins and Niederdeppe (2019), who demonstrated that narratives with gain or loss frames had no direct impact on sleep intentions.Future research should continue to examine the incongruency between protagonist self-efficacy and story outcomes, expanding beyond typical understandings of gain and loss frames.Future work may also examine when story outcomes and protagonist self-efficacy are effective by grouping behaviors by cessation (e.g., smoking, alcohol) or health promotion (e.g., sleep, eating healthy).
This study offers recommendations for alcohol-related public health messaging.Alcohol narratives featuring negative endings with low self-efficacy protagonists were motivating, possibly due to the need to disassociate oneself with protagonists given the stigma surrounding alcohol dependence (Room, 2005).One recommendation is to consider using protagonists with negative characteristics (e.g., low self-efficacy or loss Communication Reports 55 frames) that individuals can contrast with.When considering individual differences in social comparison motivations, and assimilation and contrast effects, public health messaging could use interactive narratives where individuals select protagonists with characteristics that they find important, allowing participants to make decisions for protagonists as the narrative progresses (Green & Jenkins, 2014).

Limitations and Future Directions
This study is not without limitations.One social comparison type (i.e., selfevaluation) was examined.Future research should examine other comparison types, including self-improvement and self-enhancement.Different social comparison types can account for positive and negative effects of health-related outcomes.Future research should also consider how common processes of narrative persuasion (e.g., transportation, identification) relate to social comparison with characters.Despite its limitations, the current study demonstrated the importance of investigating SCSE and subsequent contrast or assimilation effects in narratives manipulating story outcomes and protagonist self-efficacy.

3.
Fifty-two participants were randomly assigned to the high self-efficacy, positive ending condition; 56 to the high self-efficacy, negative ending condition; 57 to the low selfefficacy, positive ending condition; and 57 to the low self-efficacy, negative ending condition.

4.
Per Trauzettel-Klosinski and Dietz (2012) and Frazer et al. (2021), participants who spent less than 7 minutes (2 SD from the mean; n = 91) or over 2 hours (n = 2) were considered outliers, suggesting insufficient attention.Those who spent 100 seconds or less reading (n = 16) were removed from sleep analyses, representing the lowest 7% of reading times.5.
Sleep item reliability was low, possibly because items captured behaviors, not perceptions, which are less likely to be correlated (Robinson & Knobloch-Westerwick, 2022).Test-retest reliability demonstrated that the scale was consistent over time with a high correlation between T1 and T2 (r = .81,p < .001;Boateng et al., 2018).7.
A repeated measures ANOVA was conducted with participant alcohol and sleep self-efficacy measures and story topic (alcohol and sleep) as within-subjects factors.Story outcome and protagonist self-efficacy were between-subjects factors.The interaction of time, story topic, and protagonist self-efficacy was not significant, F(1, 198) = .00,p = .996.The interaction of

Figure 1
Figure 1 Three-way interaction of SCSE, story outcome, and protagonist alcohol self-efficacy on participants' change in self-efficacy

Table 1
Three-Way Interaction of Story Outcome, Protagonist Self-Efficacy, and Social Comparison Self-Evaluation on Participants' Change in Alcohol Self-Efficacy