Which moral exemplars inspire prosociality?

Abstract

Some stories of moral exemplars motivate us to emulate their admirable attitudes and behaviors, but why do some exemplars motivate us more than others? We systematically studied how motivation to emulate is influenced by the similarity between a reader and an exemplar in social or cultural background (Relatability) and how personally costly or demanding the exemplar’s actions are (Attainability). Study 1 found that university students reported more inspiration and related feelings after reading true stories about the good deeds of a recent fellow alum, compared to a famous moral exemplar from decades past. Study 2A developed a battery of short moral exemplar stories that more systematically varied Relatability and Attainability, along with a set of non-moral exemplar stories for comparison. Studies 2B and 2C examined the path from the story type to relatively low stakes altruism (donating to charity and intentions to volunteer) through perceived attainability and relatability, as well as elevation and pleasantness. Together, our studies suggest that it is primarily the relatability of the moral exemplars, not the attainability of their actions, that inspires more prosocial motivation, at least regarding acts that help others at a relatively low cost to oneself.

1. Introduction

Stories about exemplars are often used as sources for moral education and inspiration, but the features of stories that motivate emulation are poorly understood. Helping others often comes at a personal cost, so people sometimes require increased motivation to engage in prosocial behavior, such as donating money to charity or volunteering at a homeless shelter. Navigating the conflict between morality and self-interest is especially important in moral development, as adolescents acquire habits and expectations regarding what levels of personal sacrifice are the social norm or characteristic of virtuous people (Batson, 2016; Bicchieri, 2017; Tankard & Paluck, 2016).

Stories about the virtuous actions of moral exemplars are widely used for moral inspiration, education, and conflict resolution, with some stories being particularly effective at inspiring the emulation of exemplary moral behavior (Bandura, 1969; Kristjánsson, 2006; Sanderse, 2012; Čehajić-Clancy & Bilewicz, 2021). Several psychological mechanisms may explain why and how the presentation of moral stories facilitates the emulation of exemplary behavior. Candidates include vicarious social learning (Bandura, 1969; Chudek & Henrich, 2011; Tomasello et al., 1993), moral elevation (Haidt, 2000; Pohling & Diessner, 2016), and upward social comparison (Blanton et al., 1999). These theories might predict that any virtuous act is inspiring, regardless of who performs it, and that extraordinary acts are more inspiring than mundane deeds.

As philosophers have noted, however, unrealistically high moral standards can be problematic and even backfire (Carbonell, 2012; Wolf, 1982). Psychological studies support this worry, since people seem to be more moved by exemplars similar to themselves and whose good deeds are not out of the ordinary (Cialdini, 1980; Han et al., 2017; Lockwood & Kunda, 1997). Stories tend to induce more negative responses the more participants think the exemplar is not relatable to their own lives and engages in superhuman deeds that involve great personal sacrifice (Monin, 2007; Monin et al., 2008). For instance, the average resident in Japan might not feel particularly inspired by the story of Zell Kravinsky, who donated a kidney to a stranger and gave away nearly all of his $45 million real-estate fortunate to support public health initiatives in America. Although stories of moral exemplars often depict moral saints who engage in heroic deeds that risk their lives or livelihoods, inspirational stories can describe acts of altruism that are less demanding while nevertheless serving as examples of morally admirable behavior.

1.1. Current Study

These aspects of moral exemplar stories—relatability and attainability—can influence emulation. Attainability in this context refers to how difficult it is to emulate exemplary moral behavior given the amount of sacrifice it requires. (“Costly” or “demandingness” would be more precise but neither easily takes on both noun and adjective forms, so we will use “attainability” despite considering it a quasi-technical term.) Relatability refers to the degree to which the exemplar shares similar socio-economic and cultural backgrounds with participants. (Previous research has used the term relevance; see Han et al., 2017.) More relatable exemplars, such as peers, have been shown to promote volunteering more effectively than unrelatable exemplars who make or have made extraordinary sacrifices, such as historical figures like Martin Luther King Jr. (Cialdini, 1980; Lockwood & Kunda, 1997). However, especially demanding good deeds, such as donating large sums of money to charity, can elicit moral elevation and increase charitable donations among participants (Pohling & Diessner, 2016; Thomson & Siegel, 2013). So, it is unclear whether attainability always enhances emulation of moral exemplars.

It also remains unclear whether attainability and relatability are independently effective in the context of psychological interventions. A recent study suggested that the combination of these factors in moral exemplar stories, accessibility, increased voluntary service activity (Han et al., 2017). Peer exemplars presented to primary and secondary students in Korea better promoted prosocial behavior over a 6-week follow up period than did inaccessible historical figures. However, these studies did not distinguish between relatability and attainability, so it remains to be determined whether effects on emulation are driven by one or both factors. Such research can help us understand the motivation to be moral and how to enhance it, which has important implications for designing interventions and educational programs targeting moral development and/or improvement.

In two large studies conducted across 4 independent samples, our research examined how and why people are motivated by the actions of some moral exemplars but not others. We aimed to 1) develop sets of improved moral exemplar stories in English, 2) extend previous findings to some other forms of prosociality and to adults and university students, and 3) distinguish the effects of attainability and relatability on altruistic motivation. We examined factors that determine the effectiveness of moral exemplar interventions in making participants feel inspired (Study 1) and increasing prosociality (i.e., intention to donate, donation behavior, and prosocial emotions; Study 2). We developed and normed two sets of moral exemplar stories—one comprised of 4 stories that maximize differences in accessibility, and another comprised of 130 vignettes that systematically vary relatability and attainability. These stimuli were then used to establish which features of exemplar stories increase emulation.

1.2. Hypotheses and Predictions

Theories of vicarious social learning and moral elevation seem to suggest that observing any socially desirable behavior will increase the emulation of such behavior. Recent findings (Han et al., 2017), however, suggest that the inspiration of prosociality depends on characteristics of the moral exemplars and of their actions. In line with the theories of social learning and social comparison, we hypothesized that more relatable and attainable exemplar stories are more effective at inspiring prosociality than relatively unrelatable exemplars who model demanding acts of altruism.

We thus predicted that more relatable and attainable exemplar stories would increase prosocial emotions and behavior. In Studies 1 and 2, we predicted that more relatable and attainable heroes would be more inspirational than less relatable and less attainable ones. Self-reported perceived relatability and attainability were further predicted to partly mediate the effect of condition on inspiration. The main objective of Study 2 was to establish more firmly the effects of attainability and relatability on prosociality, or whether they share sufficient variance to warrant their collapse into a single factor (accessibility).

2. Study 1

2.1. Background

Study 1 tested the hypothesis that university students find exemplars more inspiring when the exemplars are more relatable and their actions less demanding. Participants read about either similar but everyday exemplars (specifically, recent alumnae of their school) or extraordinary but unrelatable historical exemplars. We hypothesized that a shared alma mater and gender would make exemplars especially relatable and that their deeds of helping others in their communities would be seen as not especially demanding. The extraordinary deeds of historical figures, on the other hand, are less likely to be seen as relatable and attainable, even if they are objectively more admirable (Frimer et al., 2012).

2.2. Method

2.2.1. Participants

One hundred and one (57 female) Calvin University undergraduates aged 18 or older participated in Study 1. Of these participants, twenty-six participated on a strictly voluntary basis and 75 earned credit for an introductory psychology course. Participation took around 5 minutes and included a question about gender to inform the assignment of participants to the appropriate experimental conditions. No participants failed the attention check (“If you are reading this, choose the ‘2’ option below.”), but data were excluded from one participant who circled multiple answers in response to several survey items.¹

The original recruitment target as described in the pre-registration (https://osf.io/5ck7m/) was to enroll 200 participants during the first two weeks of November 2019, or as many participants as possible in that time window.² Because we were only able to collect data from 100 participants, we conducted a sensitivity power analysis that indicated a sample of this size (N = 50 per group) would provide power to detect medium-sized effects (d = 0.50, or around 6% of the variance; Faul et al., 2009). Of note, this effect size is comparable to the medium-sized effect reported in related studies (e.g. Han et al., 2017; f = 0.22, approaching 6% of the variance).

2.2.2. Procedure

The hypothesis, predictions, and materials for Study 1 were pre-registered: https://osf.io/5ck7m/. Participants received printed copies of packets that included a consent form, one moral exemplar story, and a questionnaire to assess their feelings after reading the story. Participants were randomly assigned to one of two conditions: an accessible moral exemplar story (Calvin alum Alivia Hibbler or Tyler Smies) or an inaccessible exemplar story (Rosa Parks or Nelson Mandela). The stories of Mandela and Parks described their well-known sacrifices for social justice. The stories of the Calvin graduates described how each pursued careers that would help others but yield smaller salaries—namely, starting a non-profit organization and taking a job in poorer school district (see Supplementary Materials for story texts). To further manipulate perceived similarity, each participant read about someone of the same or different gender. In the accessible condition, women read about Alivia and men read about Tyler. In the inaccessible condition, participants were assigned to read about opposite-gendered exemplars (i.e., women read about Nelson Mandela and men read about Rosa Parks).

After each story, participants rated their feelings along the following dimensions: “moved”, “uplifted”, “optimistic about humanity,” “warm feeling in chest,” “want to help others,” and “want to become a better person” (9-point Likert scale; 1 = “Didn’t feel anything”, 9 = “Felt very strongly”). Individuals’ ratings were summed and, as in previous studies, the resulting scores were interpreted as measuring inspiration (e.g. Schnall et al., 2010). On another 9-point scale, participants indicated their perceptions of attainability (“How attainable for you is [exemplar name]’s goodness?”) and relatability (“How similar do you think you and [exemplar name] are?”). Two additional questions, not examined here, assessed participants’ agreement with statements about having a sense of purpose and of having a sense of meaning in one’s life.

2.3. Results and Discussion

We tested for an effect of condition by using an independent-samples t-test (one-tailed, since the hypothesis that accessible exemplars are more inspiring than inaccessible exemplars was directional and pre-registered). We also conducted a mediational analysis according to published recommendations (Preacher & Hayes, 2008; 1-tailed). Specifically, we simultaneously tested whether relatability and attainability together mediate the effect of condition (accessible vs. inaccessible).

Consistent with our pre-registered hypothesis, Calvin University undergraduate students who read stories about the exemplary actions of Calvin alum reported higher levels of inspiration (5.88 ± 1.52) than students who read about Rosa Parks or Nelson Mandela (5.20 ± 1.70; Table 1), t(98) = 2.10, p = 0.019, d = 0.42. The effect of condition on inspiration was mediated by the perceived relatability (95% CI = [0.23, 1.08]) and attainability (95% CI = [.11, 0.79]) of each exemplar, with a regression analysis suggesting these variables fully mediated the relationship. When inspiration scores were simultaneously regressed on condition, attainability, and relatability, there was no relationship between inspiration and condition (β(97) = −0.09, p = 0.351).

Table 1 |.

Mean ratings for inspiration, attainability, and relevance for each moral exemplar story from Study 1.

Exemplar Story	N	Inspiration	Perceived Attainability	Perceived Relatability
Rosa Parks	19	5.22 ± 1.98	6.26 ± 1.73	4.32 ± 1.60
Nelson Mandela	31	5.19 ± 1.54	4.97 ± 1.74	3.65 ± 1.85
Unrelatable & Unattainable	50	5.20 ± 1.70	5.46 ± 1.83	3.90 ± 1.78
Alivia Hibbler	25	6.03 ± 1.44	6.88 ± 1.59	5.48 ± 1.56
Tyler Smies	25	5.73 ± 1.60	6.84 ± 1.40	6.08 ± 1.93
Relatable & Attainable	50	5.88 ± 1.52	6.86 ± 1.48	5.78 ± 1.76

3. Study 2A

3.1. Background

The previous study did not systematically vary the relatability and attainability of the exemplar stories, and it measured prosocial emotions but not behavior. The aim of Study 2A was to generate a large, standardized battery of moral exemplar vignettes in which each vignette’s features were carefully modified to manipulate their relatability (e.g., fellow American adults) and attainability (e.g., donating a modest amount of money to charity vs. a large amount). These stimuli could then be used in our subsequent studies to measure their effects on prosocial behavior among adults online (Study 2B) and university students (Study 2C). We also tested whether two story features—attainability and relatability—could be collapsed into one factor (“accessibility”).

3.2. Method

3.2.1. Participants

A total of 401 participants were enrolled into Study 2A in November 2017 (214 male, 186 female, 1 sex unknown; 35.81 ± 9.76 years of age). An a priori power analysis indicated that a sample size of around N = 400 participants would enable at least 80% power to detect effects of magnitudes similar to those reported previously (f = .22; Han et al., 2017). Participants were recruited through Amazon’s Mechanical Turk (MTurk) to complete an online survey administered with the Qualtrics platform. They were compensated at a rate of $7.25 for about 1 hour of their time in accordance with federal minimum wage. Only people aged 19 or older could enroll, which is considered the age of majority in the state of Alabama, where the study was conducted. None of the participants’ data were excluded.³

3.2.2. Procedure

We adapted our stimuli from a well-characterized set of ecologically valid true moral stories described by participants in an independent study (Knutson et al., 2010). We selected a subset of 26 vignettes that described individuals who made personal sacrifices to do something good for themselves or others. It was important to use scenarios that describe familiar situations since recent evidence suggests moral judgments elicited by extreme sacrificial moral scenarios are poor predicators of real-world behavior (Bostyn et al., 2018; FeldmanHall et al., 2012). These vignettes were modified to vary independently along dimensions of attainability and relatability. Since attainability involves the perceived difficulty of emulating an exemplary behavior, we manipulated the level of sacrifice made by each moral exemplar (e.g., volunteering under 3 hours per week vs. over 9; giving a homeless person $5 USD vs. $1,000 USD; comforting a survivor of a car crash vs. entering the blazing wreckage to save her). Since relatability refers to the perceived correspondence of characteristics between oneself and a moral exemplar, we manipulated background information used to characterize each exemplar (e.g., German vs. American nationality; modern day vs. historical setting).

A total of 130 brief vignettes were constructed, each of which belonged to one of five subgroups (26 stories per subgroup: relatable and attainable, relatable and unattainable, unrelatable and attainable, and unrelatable and unattainable). We also constructed 26 novel vignettes about non-moral exemplars, which described individuals who carried out exemplary actions that primarily benefited themselves (e.g., winning a skill-based contest after months of arduous training). All the vignettes were similar in length and structure (Table S1). The complete set of vignettes is provided in the Supplementary Materials, and a thorough characterization of each vignette’s features can be downloaded from the OSF archive corresponding to this research (https://osf.io/v5nk7/).

In Study 2A, participants rated a subset of the 130 moral exemplar vignettes (104 moral stories, 26 non-moral; see Table 2 for examples). Participants each read and rated 52 stories (26 randomly selected moral stories and all 26 non-moral stories). In addition to attainability and relatability, participants rated the stories in terms of elevation, praiseworthiness, emotional intensity, pleasantness, moral relatability, benefit to others, and benefit to self. At the end of the survey, participants also completed a demographic questionnaire (see Supplementary Materials).

Table 2 |.

Example vignettes from Study 2 corresponding to each stimulus category (relatable or unrelatable, attainable or unattainable, and non-moral).

	Relatable	Unrelatable
Attainable	Albert is a 21-year old recent US college graduate who had been looking for work unsuccessfully for a few weeks. Albert started working in a restaurant and had been there only 2 weeks when a fellow co-worker’s brother died. Albert’s co-worker couldn’t come into work the day her brother died and none of her fellow co-workers would cover the shift. So, Albert worked a double shift that day and covered for his co-worker.	Anya is 65 and retired a few years back from her job at a bottling plant in Russia but was recently looking for work to stay active. Anya started working in a restaurant and had been there only 2 weeks when a fellow co-worker’s brother died. Anya’s co-worker couldn’t come into work the day her brother died and none of her fellow co-workers would cover the shift. So, Anya worked a double shift that day and covered for her co-worker.
Unattainable	Albert is a 21-year old recent US college graduate who had been looking for work unsuccessfully for a few weeks. Albert started working in a restaurant and had been there only 2 weeks when a fellow co-worker’s brother died. Albert’s co-worker couldn’t come into work the day her brother died and none of her fellow co-workers would cover the shift. So, Albert worked a double shift that day and covered the rest of his shifts for the following 2 weeks.	Anya is 65 and retired a few years back from her job at a bottling plant in Russia but was recently looking for work to stay active. Anya started working in a restaurant and had been there only 2 weeks when a fellow co-worker’s brother died. Anya’s co-worker couldn’t come into work the day her brother died and none of her fellow co-workers would cover the shift. So, Anya worked a double shift that day and covered her shifts for the following 2 weeks.
Non-moral exemplar
Michael is a 32-year old head chef at a restaurant in Tulsa, Oklahoma. While Michael enjoys the praise his restaurant is getting, his primary concern is making good food. Michael was recently profiled by a local newspaper because of his success in the culinary industry despite his young age. Several prominent critics in the area have eaten at Michael’s restaurant and said his was some of the best food they’d eaten in years.
Original story from Knutson et al., 2010
I started working in a restaurant and had been there only 2 weeks when a fellow co-worker’s brother died. She couldn’t come into work the day he died and none of her fellow co-workers would cover her shift. So, I worked a double shift that day and covered for her.

3.2.3. Statistical Analysis

We examined the following: first, whether stories in the attainable and relatable conditions were seen as more attainable and relatable (respectively) than those in the unattainable and unrelatable conditions; second, whether moral vignettes evoked stronger emotional responses than non-moral vignettes. We calculated mean attainability and relatability scores for each exemplary story from Study 2A. If detected, the final moral exemplar intervention used in Studies 2B and 2C would exclude vignettes that received attainability and relatability ratings in unexpected directions.

We next performed a mixed-effects analysis to examine the effect of the moral exemplar stories (attainable/relatable | attainable/unrelatable | unattainable/relatable | unattainable/unrelatable) on participants’ responses (using lmer in lmerTest). For dependent variables, eight dimensions were assessed: perceived attainability (two items), perceived relatability (two items), elevation, praiseworthiness, emotional intensity, and pleasantness. The story condition was entered into the model as a fixed effect and participant ID was entered as a random effect. To further inform our inferences, we performed a Bayesian mixed-effects analysis to examine whether the data supported the presence of a non-zero main effect (using anovaBF in BayesFactor). For Bayesian inference, we employed 2log(Bayes Factor (BF)) ≥ 2 as a threshold (Han et al., 2018; Wagenmakers et al., 2018). The effect size representing the main effect was estimated in ${\Omega }_0^2$, which was introduced by Xu (2003) as an indicator for effect sizes in mixed-effects analyses. Additionally, we performed a Bonferroni-corrected post hoc test (using glht in multcomp).

We also examined whether two story features—attainability and relatability—could be collapsed into one feature—accessibility—with the collected dataset. We used multinomial logistic regressions to address this question. We compared two multinomial logistic regression models: one predicted the story type with attainability and relatability as two separate features, and the other predicted story type with accessibility, which was calculated by averaging attainability and relatability scores. In this comparison, we examined which model better predicted the story type. We compared the models’ Akaike information criterion (AIC) and Bayesian information criterion (BIC) values and interpreted lower AIC and BIC values as indicating the better model.

3.3. Results and Discussion

Table 3 presents descriptive statistics and the results of a mixed-effects analysis. Both frequentist and Bayesian mixed-effects analyses detected a significant main effect of condition (p < .001 and 2log(BF) > 10), the magnitude of which was large in all cases. We found that our stimuli have the intended features—that is, they appropriately manipulated attainability and relatability in the expected directions. With respect to relatability, however, the results were mixed. For the first relatability item (“How similar do you think your cultural and social background is to the person described in the story?”), relatability scores were significantly higher for stories categorized as relatable compared with stories categorized as unrelatable. For the second relatability item (“I know people who have done similar things in the past to the person described in the story.”), the pattern of responses was similar to that observed for attainability. Furthermore, as predicted, participants had stronger emotional responses to moral compared to non-moral exemplar stories.

Table 3 |.

Descriptive statistics and results from the mixed-effects analysis in Study 2A.

Variable	Mixed-effects analysis					Descriptive statistics M (SD)
	F	df	p	Ω2	logBF	Relatable, Attainable	Relatable, Unattainable	Unrelatable, Attainable	Unrelatable, Unattainable	Non-moral
Attainability 1	853.30	(4, 20484.63)	< .001	.32	1556.20	4.61 (1.90)a	3.46 (1.87)b	4.58 (1.89)a	3.52 (1.88)b	3.04 (1.67)c
Attainability 2	386.99	(4, 20476.20)	< .001	.32	735.10	3.92 (2.10)a	3.06 (1.97)b	3.85 (2.10)a	3.10 (2.00)b	2.76 (1.88)c
Relatability 1	281.91	(4, 20464.47)	< .001	.41	533.79	4.32 (1.52)a	4.15 (1.55)b	3.52 (1.75)c	3.31 (1.76)d	3.77 (1.62)c
Relatability 2	197.11	(4, 20476.79)	< .001	.29	372.41	4.76 (1.96)a	3.90 (2.06)b	4.61 (1.98)a	3.92 (2.08)b	3.85 (2.07)b
Elevation	1016.07	(4, 20467.49)	< .001	.43	1844.57	5.31 (1.53)a	5.68 (1.43)b	5.39 (1.47)a	5.72 (1.38)b	4.39 (1.78)c
Praiseworthy	830.14	(4, 20465.40)	< .001	.44	1525.38	5.58 (1.45)a	5.94 (1.36)b	5.66 (1.37)a	5.96 (1.33)b	4.84 (1.69)c
Intensity	981.08	(4, 20464.99)	< .001	.45	1772.24	4.00 (1.84)a	4.24 (1.81)b	4.02 (1.77)a	4.29 (1.80)b	2.90 (1.86)c
Pleasantness	606.75	(4, 20466.02)	< .001	.42	1130.03	5.50 (1.47)a	5.73 (1.40)b	5.53 (1.38)a	5.74 (1.40)b	4.88 (1.38)c

Note. logBF: log(Bayes Factor).

^a-d:averages in a row without a common superscript letter differ at p < .05 (Bonferroni corrected).

Except for the second relatability item, the between-group differences in participants’ responses (including perceived attainability, perceived relatability, and emotions) were consistent with our predictions. Based on these results, we concluded that the moral exemplar stories designed for Studies 2B and 2C appropriately differentiate relatability from attainability. Since the second relatability item did not conform to our expectations about what constitutes relatability, this item was excluded from all following studies. Similarly, since both attainability items had the predicted effects, we opted to use only the first item in future studies (“How difficult do you think it would be to do the same things as the person described in the story?”).

We also performed multinomial logistic regression analyses to examine whether attainability and relatability could be collapsed into one feature: accessibility. In the first regression, we used one variable for each feature as explained previously. For the second regression, accessibility scores were calculated by averaging the two items. The first regression model that included both attainability and relatability as separate independent variables yielded an AIC = 54,785.06 and BIC = 54,880.40. The second regression model that only included accessibility yielded an AIC = 55,985.18 and BIC = 56,048.74. Given that both AIC and BIC values were smaller when modeling attainability and relatability independently, we conclude that it is better to treat attainability and relatability as independent features instead of collapsing them into one feature (accessibility).

4. Study 2B

4.1. Background

The aim of Study 2B was to investigate whether moral exemplars inspire emulation of prosociality because they are more relatable, more attainable, or both. Specifically, we examined whether acute changes in prosociality following exposure to moral exemplars is more strongly associated with the relatability of the exemplars, with the attainability of their actions, or alternatively with a single factor representing both (accessibility).

4.2. Method

4.2.1. Participants

401 participants recruited through MTurk that were aged 19 or older were enrolled into Study 2B (222 male, 179 female; 34.99 ± 10.18 years of age). As in Study 2A, a sample of N = 400 participants was expected to provide at least 80% power to detect the effects of interest. Participants completed an online survey on Qualtrics and were compensated $7.25 for around an hour of their time. After enrollment, participants were randomly assigned to one of five exemplar intervention conditions, which involved reading 26 stories of only one type: relatable and attainable (n = 97), relatable and unattainable (n = 71), unrelatable and attainable (n = 78), unrelatable and unattainable (n = 78), or non-moral (n = 77).⁴

4.2.2. Procedure

The same exemplary stories used in Study 2A comprised the moral exemplar intervention in Study 2B. While reading the 26 stories, participants rated each one along four dimensions: perceived attainability, perceived relatability, elevation, and pleasantness. Specifically, participants answered the following questions: “How difficult do you think it would be to do the same things as the person described in the story?” (1 = “Not difficult at all”, 7 = “Extremely difficult”), “How similar do you think your cultural and social background is to the person described in the story?” (1 = “Not at all similar”, 7 = “Extremely similar”), “The story made me feel morally elevated” (warm, uplifted – like when seeing unexpected acts of human goodness, kindness, or compassion)” (1 = “Strongly disagree”, 7 = “Strongly agree”), and “How pleasant do you find the actions of the person described in the story?” (1 = “Extremely unpleasant”, 7 = “Extremely pleasant”). After reading all 26 stories, participants completed the Narrative Transportation Scale to assess how absorbed into the stories they felt (data not examined here; Green & Brock, 2000).

Our previous work suggests the moral exemplar intervention may selectively increase some but not all aspects of prosociality (Han et al., 2017), although the nature of this relationship has not been systematically investigated. To accurately assess the efficacy of the moral exemplar intervention used in Study 2B, then, it was necessary to include several complementary measures to capture multiple aspects of prosociality. (See Supplementary Materials for the exact text presented to participants.) First, before reading any stories, participants indicated how many hours they planned to spend volunteering in the following month. Second, before reading the stories, participants were informed they would receive a $2 bonus they could either keep for themselves or donate to one of 6 charities, and that they could further decide not only whether but how much of their bonus to donate.

After the intervention, participants were again presented with opportunities to indicate how many hours they intended to volunteer over the next month and how much of their $2 bonus they wished to donate to charity. Next, participants were informed the donation could not be made on their behalf and that they would instead receive both $5.25 for participating and the entire $2 bonus ($7.25 in total). Demographic information (e.g., age, sex, ethnicity, and socioeconomic status) was also recorded (see the Supplementary Materials).

4.2.3. Statistical Analysis

We examined the descriptive statistics characterizing participants’ responses as well as correlations between variables of interest (perceived attainability, perceived relatability, elevation, pleasantness, change in donation, and change in intention to volunteer). We performed both frequentist and Bayesian correlation analyses, with the latter examining if the data positively supported the presence of non-zero correlations based on the resultant Bayes Factor (2log(BF) ≥ 2). We also conducted both frequentist and Bayesian ANOVAs to test for a main effect of story type on the variables of interest. Where appropriate, we performed Bonferroni corrected post hoc testing to characterize differences between the categories of exemplar stories.

We performed a path analysis to characterize the influence of the attainability and relatability of the moral exemplars on prosociality. Specifically, we examined paths from the story type to behavioral responses: “story type (attainable, relatability) → perceived attainability and relatability → emotional responses → behavioral responses.” We set the final path model based on the full path model including all possible paths (see Figure 1). The final model was selected using Bayesian Model Averaging (BMA), which identifies the best regression model from the possible candidate regressions comprising each path analysis (Hoeting et al., 1999). Instead of excluding predictors that failed to achieve significance—which is likely to inflate false positives or estimated coefficients and does not properly address the issue of multicollinearity—BMA selects the most probable regression model from possible candidates by examining the posterior probability of each candidate with Bayesian inference (Han & Dawson, 2021). For instance, when BMA was applied to identify the best regression model for predicting changes in donation behavior, we entered all candidate predictors (story type, perceived attainability, perceived relatability, elevation, pleasantness) and examined which model yielded the highest posterior probability. Once optimal regression models were identified for all paths, the path model was estimated with sem in lavaan. We also performed Bayesian path analysis with bsem in blavaan to examine whether the 95% Bayesian credible interval (CI) of each estimated coefficient included zero. Since the non-moral stories did not vary categorically in terms of attainability and relatability, data collected from participants assigned to read the non-moral stories was excluded from the path analysis.

Figure 1 |.

The hypothetical full path model.

4.3. Results and Discussion

The descriptive statistics characterizing the variables of interest are given in Table 4. The results of the correlational analyses are shown in Figure 2. Both frequentist and Bayesian ANOVAs detected a significant main effect of story type on perceived attainability and relatability when elevation and pleasantness were included in the model. The main effect was not significant, however, for either of the behavioral variables. As shown in Table 4, perceived attainability was higher among participants assigned to attainable story groups. Although relatable stories were perceived to be more relatable in general, there was no significant difference in the perceived relatability of stories in the relatable & attainable and unrelatable & attainable conditions. Both elevation and pleasantness were significantly increased for the four kinds of moral stories compared to the non-moral stories. The results of the correlation analyses indicate that perceived relatability was significantly associated with perceived attainability and emotional responses. Of the two behavioral variables, only change in intentions to volunteer was positively associated with perceived relatability and emotional responses to the exemplar stories according to the frequentist correlation analysis; using Bayesian correlation analysis, however, only the relation between change in intention to volunteer and elevation survived.

Table 4 |.

Descriptive statistics and results from the mixed-effects analysis in Study 2B.

Variable	Mixed-effects analysis					Descriptive statistics M (SD)
	F	df	p	Ω2	logBF	Relatable, Attainable	Relatable, Unattainable	Unrelatable, Attainable	Unrelatable, Unattainable	Non-moral
Attainability	71.79	(4, 324)	< .001	.47	92.36	4.69 (.92)a	3.37 (.89)b	4.98 (.93)a	3.36 (.86)b	2.84 (.87)c
Readability	11.75	(4, 324)	< .001	.13	14.52	4.50 (1.11)a	4.11 (1.02)ab	3.51 (1.46)bc	3.23 (1.40)c	3.83 (1.03)bc
Elevation	14.95	(4, 324)	< 001	.13	19.33	5.37 (.94)a	5.58 (.89)a	5.06 (1.11)a	5.57 (1.14)a	4.32 (1.15)b
Pleasantness	9.85	(4, 324)	< .001	.11	11.17	5.69 (.83)ab	5.91 (.73)a	5.43 (0.80)b	5.82 (.88)ab	5.09 (.82)c
Change in donation	1.52	(4, 324)	.20	.02	−2.51	.07 (.28)a	.11 (.46)a	.05 (.24)a	.09 (.33)a	−.02 (.23)a
Change in volunteering	2.33	(4, 324)	.06	.03	−1.17	.47 (1.22)a	.27 (.65)a	.51 (1.85)a	.88 (1.60)a	.17 (1.30)a

Note. logBF: log(Bayes Factor).

^a-c:averages in a row without a common superscript letter differ at p < .05 (Bonferroni corrected).

Figure 2 |.

Correlations from Study 2B. Elements below the diagonal represent correlations significant at p < .05. Elements above the diagonal represent correlations significant at 2logBF ≥ 3. Attn: perceived attainability. Rel: perceived relatability. Elev: elevation. Pleas: pleasantness. Donate: change in donation. Volunteer: change in volunteering intention.

Figure 3 shows the best path model (BMA) as well as the estimated coefficients (using sem). The identified model fit the data well given the reported model fit indicators, RMSEA = .06 (< .08), SRMR = .05 (< .08), CFI = .98 (≥ .90), TLI = .96 (≥ .95), posterior predictive p-value = .15 (≥ .05). All examined coefficients were significantly different from zero (p < .05; did not include zero in the 95% Bayesian CI). The results of the path analysis suggest that participants’ volunteering intentions were significantly predicted by a path via perceived relatability and elevation. Interestingly, perceived attainability did not play a significant role in modulating the effects of moral exemplar stories on prosociality. In fact, stories describing attainable exemplary actions were negatively associated with elevation.

Figure 3 |.

The resultant path model from Study 2B. Standardized path coefficients are presented. All paths were significant at p < .05 and did not include zero in their 95% Bayesian CI.

These results extend the finding from Study 1 that perceived attainability and relatability mediated the relationship between the story type and emotional response by disentangling the respective contributions of attainability and relatability. Although story type did not have a significant effect on the prosocial behavior of participants, relatability and attainability are relative notions, so it was imperative to examine the behavioral effects of perceived attainability and relatability. The path analysis accordingly uncovered a significant change in intentions to volunteer, mediated in particular by perceived relatability. These results suggest that the relatability of stories, as opposed to attainability, more strongly influences emotional and motivational responses to reading moral exemplar stories, even if not prosocial behavior. One reason similar effects were not detected for donation behavior may be that, for many MTurk workers, their compensation is a source of already meager income, so the prospect of donating a part of their income to charities may seem less feasible. This underscores the need for additional research to clarify the behavioral consequences of moral exemplar interventions, especially in the kinds of educational settings where such interventions might eventually be used.

5. Study 2C

5.1. Background

The aim of Study 2C was to replicate the findings from Studies 2A and 2B in a population of students recruited from an undergraduate educational setting, the University of Alabama. We reasoned that the average university student would be more likely to volunteer and donate to charities than the average worker on MTurk.

5.2. Method

5.2.1. Participants

A total of 218 participants were enrolled in Study 2B (190 female; 21.49 ± 6.67 years of age). Participants were undergraduate students from the University of Alabama aged 18 years or older who were recruited through one of two undergraduate psychology volunteering pools. Once enrolled, student volunteers came for an in-person visit to our computer lab where they completed the study procedures, starting with providing written informed consent. Participants received course credit for participating in the study, which lasted around an hour. There were no exclusions of data. A sample of N = 200 participants or greater was expected to provide at least 80% power to detect the effects of interest.⁵

5.2.2. Procedure

The procedure was identical to Study 2B, except for the compensation. Study 2C was originally designed to span two sessions, with participants receiving an additional $2.50 if they completed a second session several weeks later. During the first session, participants reported their intentions to donate a portion of the $2.50 (from $0.00 to $2.50) after completing the second session. Participants were asked this question before and after the first session to characterize changes in intentions to donate. Ultimately, owing to administrative constraints outside the control of the research team, it was not possible to invite participants back for a second session.

5.3. Results and Discussion

The descriptive statistics characterizing the variables of interest are given in Table 5. The results of the correlational analyses are shown in Figure 4. The overall findings from the correlational analysis and ANOVAs were similar to those from Study 2B, including the lack of a significant effect of story type on changes in intentions to donate or volunteer. However, we found additional significant frequentist associations between the change in donating intent, perceived relatability, elevation, and pleasantness. In addition, a Bayesian correlational analysis revealed one additional significant correlation between change in volunteering intention and pleasantness.

Table 5 |.

Descriptive statistics and results from the mixed-effects analysis in Study 2C.

Variable	Mixed-effects analysis					Descriptive statistics M (SD)
	F	df	p	Ω2	logBF	Relatable, Attainable	Relatable, Unattainable	Unrelatable, Attainable	Unrelatable, Unattainable	Non-moral
Attainability	43.62	(4, 213)	< .001	.45	54.18	4.93 (.94)a	3.85 (.93)b	5.24 (.84)a	3.87 (.93)b	3.06 (.76)c
Readability	8.51	(4, 213)	< .001	.14	8.97	4.33 (1.21)a	4.16 (1.09)ab	3.54 (1.41)bc	3.19 (1.04)c	3.38 (.96)c
Elevation	4.74	(4, 213)	< .001	.14	3.10	5.59 (1.13)a	5.36 (1.10)a	5.24 (1.15)ab	5.45 (1.26)a	4.66 (.83)b
Pleasantness	4.63	(4, 213)	< .001	.08	2.93	5.76 (1.09)a	5.50 (1.05)ab	5.46 (1.06)ab	5.62 (1.11)a	4.89 (.85)b
Change in donation intent	2.21	(4, 213)	.07	.04	−.90	.05 (.22)a	.05 (.21)a	.06 (.23)a	−.07 (.33)a	−.01 (.25)a
Change in volunteering	1.22	(4, 213)	.30	.02	−2.44	.40 (1.96)a	.73 (1.63)a	.40 (2.05)a	1.00 (2.76)a	.09 (1.95)a

Note. logBF: log(Bayes Factor).

^a-c:averages in a row without a common superscript letter differ at p < .05 (Bonferroni corrected).

Figure 4 |.

Correlations from Study 2C. Elements below the diagonal represent correlations significant at p < .05. Elements above the diagonal represent correlations significant at 2logBF ≥ 3. Rel: perceived relatability. Elev: elevation. Pleas: pleasantness. Donate: change in donating intention. Volunteer: change in volunteering intention.

Figure 5 shows the result from the path analysis (using sem). Similar to Study 2B, the identified model fit the data well given the reported model fit indicators, RMSEA = .03 (< .08), SRMR = .05 (< .08), CFI = .99 (≥ .90), TLI = .99 (≥ .95), posterior predictive p-value = .39 (≥ .05). All examined coefficients were significantly different from zero (p < .05; did not include zero in the 95% Bayesian CI). We found two significant paths: the relatability of the presented story → perceived relatability → elevation → change in donating intention, and → change in volunteering intention.

Figure 5 |.

The resultant path model from Study 2C. Standardized path coefficients are presented. All paths were significant at p < .05 and did not include zero in their 95% Bayesian CI.

The overall findings from Study 2C were similar to those from Study 2B. Again, while a priori categorical differences in story type did not significantly increase prosocial behavior among students, path analysis may nevertheless reveal behavioral consequences linked to perceived attainability and perceived relatability. Similar to the previous study, the present path analysis suggests that relatability, not attainability, significantly increased prosocial emotions and behavior. However, we found several interesting differences between Studies 2B and 2C. First, only relatability was significant in the current path analysis. Second, in contrast to Study 2B, we found that the change in donating intent was significantly predicted by relatability and elevation in the current study. These differences might originate from the selection of a participant pool comprised of students rather than workers. Unlike MTurk workers who might consider the provided compensation as a source of income, college students in Study 2C might see the compensation as a luxury that would be inessential for sustaining their lives, and thus consider donation more favorably.

Attainability may also require further scrutiny. In both Studies 2B and 2C, the attainability condition as well as perceived attainability did not demonstrate the hypothesized effects on the motivational and behavioral measures. Perhaps our “unattainable” stories didn’t involve acts that were sufficiently costly or demanding. Most of the stories described moderate demands (e.g., donating $1,000 USD), not life-threatening risks (Čehajić-Clancy & Bilewicz, 2021). Unfortunately, our story set did not include stories of exemplars taking such great risks, partly because it used true stories collected in previous research. Further studies might include more diverse exemplar stories to understand the potential effect of attainability better.

6. General Discussion

Stories can describe moral exemplars who are more or less similar to the reader (relatability) and who engage in acts that are more or less difficult to emulate (attainability). The overarching aim of this research was to address whether prosocial motivation is increased by greater attainability, relatability, or both. Overall, as predicted, more relatable and attainable exemplar stories generate greater inspiration (Study 1) and emulation of prosociality on some measures (Study 2), with perceived relatability being most influential. We developed a battery of ecologically valid exemplar stories that systematically varied attainability and relatability. Although differences in our story types did not produce detectable changes in prosocial behavior, perceived attainability and relatability are highly relative to the individual and thus difficult to systematically manipulate for all or even most participants. For instance, the average American might relate little to a Russian retiree, while others in our studies might do so easily (e.g., if their parents grew up in the Soviet Union). Similarly, donating $50 USD to charity is a major sacrifice for some Americans but not others. So, it was important for us to directly examine the effects of perceived attainability and relatability on prosociality.

The path analyses conducted in Studies 2B and 2C suggest in particular that the perceived relatability—not attainability—of a moral exemplar tends to increase emulation among readers. The more attainable stories and perceived attainability did not positively predict emotional and behavioral outcomes, but the more relatable stories and perceived relatability did. This suggests that the relatability of exemplars is more fundamental in motivating people compared with the attainability of their acts. Another possibility is that highly attainable moral actions require little personal sacrifice, such as donating $1 to a charity, which is not particularly inspiring and in some cases is perhaps even seen as insulting (compare Thomson and Siegel 2013). Further research could explore these possibilities.

Our studies have several limitations. One concerns generalizability since we did not measure all forms of prosocial attitudes and behaviors. With few exceptions, our stories focused on relatively familiar acts of altruism, not potentially life-threatening risks. Moreover, we primarily measured relatively low stakes prosociality among our participants, such as feelings of inspiration, donating small amounts of money to charity, and reporting intentions to volunteer. Can relatable moral exemplars motivate even greater sacrifices than these? Moreover, morality involves much more than helping others. What about stories of moral exemplars who bring loved ones to justice or who refuse to cheat when others are doing so with abandon? Further studies could explore whether relatable moral exemplars are especially likely to effect change in other moral domains as well.

Another limitation of our studies is that they do not uncover why the relatability of moral exemplars increases prosocial motivation. Nevertheless, psychological mechanisms posited by existing theories might explain our results. Reading about morally admirable behavior can motivate emulation by inducing warm and uplifting feelings that increase one’s desire to “affiliate and help others” (Haidt, 2000; Pohling & Diessner, 2016). This theory of moral elevation, however, might not easily explain the greater influence of relatable moral exemplars. Alternatively, moral exemplar stories might harness social or cultural learning by signaling to readers that such exemplary actions conform to expected norms, particularly when exemplars belong to one’s social group (Bandura, 1969; Chudek & Henrich, 2011; Tomasello et al., 1993). More specifically, exemplar stories might instigate a form of social comparison that makes salient disparities in perceived moral excellence between oneself and a member of one’s in-group, and readers believe that emulation will help close the gaps (Blanton et al., 1999). Although more research is needed to adjudicate among these and other theories, the importance of relatability in our studies does suggest that the mechanism involves a fundamentally social element that can explain why readers would be more inspired by a moral exemplar from their own social group.

Despite present limitations, these and future studies have the potential to generate both theoretical and practical implications for many fields, particularly applied ethics, moral psychology, moral development, and moral education. Consider first some theoretical models of moral psychology and development. Experimental evidence increasingly suggests that moral judgment is largely automatic and driven by unconscious processes that nevertheless involve complex computations (May, 2018; Mikhail, 2011; Railton, 2014). Models of cultural transmission and learning suggest that humans naturally learn by imitating others in their social group, particularly those perceived to have skill, success, confidence, experience, and prestige (Chudek & Henrich, 2011; Tomasello et al., 1993). Behavior that appears morally optional can turn morally necessary the more that members of one’s community engage in it (Bicchieri, 2017; Tankard & Paluck, 2016). Research on moral exemplars potentially connects such models of cultural norms with theories in moral psychology that aim to uncover the tacit processes driving moral cognition and learning. Given that our participants were more inclined to emulate personally relatable exemplars, relatability might be a tacit consideration that drives moral cognition, such that moral learning is not special but rather like other forms of cultural learning.

Understanding the effects of moral exemplars on prosociality might have practical implications as well. Findings like ours can aid in the development of educational interventions that are fine-tuned to promote character in students. Parents and educators know the importance of role models in transmitting moral beliefs and knowledge, and our studies suggest that role models will be more effective, and less likely to backfire, when relatable (and perhaps when their good deeds are not especially demanding). Our research might also support psychological interventions targeting the general public. Čehajić-Clancy and Bilewicz (2021) demonstrated that stories of moral exemplars can promote reconciliation among antagonistic groups living in societies recovering from violent conflicts, such as war and genocide.

The present study also has implications for moral philosophy. In applied ethics, there is a lively debate about whether we ought to enhance our moral capacities to increase ethical action beyond normal levels. Some even believe that such enhancement is necessary to prevent evil people from using powerful biotechnologies to devastate human societies (e.g. Persson & Savulescu, 2012). Other ethicists worry that enhancing moral behavior is unethical because it is manipulative, inauthentic, or paternalistic (e.g., Fukuyama, 2002; Sandel, 2004). Whether such enhancement is morally problematic, though, depends on how it is best carried out, which is an empirical question. Bioethicists have recently focused on new-fangled technologies that directly manipulate the brain, such as transcranial direct current stimulation and pharmaceutical drugs. Our studies might provide a useful corrective by emphasizing the power of traditional interventions, such as role models, in promoting morally desirable characteristics. Our studies indicate that more traditional forms of moral enhancement may be effective and at the same time less morally problematic than other strategies.

7. Conclusion

What types of moral exemplar narratives are most effective at promoting prosocial attitudes and behaviors? This research examined whether the psychological effects of moral exemplars are attributable to the costliness of helping, the relatability of the exemplars, or some combination of both factors. Study 1 found that college students felt more inspired by fellow alumni than historical figures known for their extraordinary moral actions. Study 2 found that stories seen as more relatable elicited more motivation to volunteer or donate to charity. Our research begins to provide both researchers and moral educators with insights about the types of moral exemplars that can effectively promote prosociality.

Data availability statement:

All data supporting the findings reported herein, along with the study materials and code for the statistical analyses, are available from the Open Science Framework repository: https://osf.io/v5nk7/