Children's moral reasoning about self- versus other-benefiting public health measures

Abstract

The COVID-19 pandemic has introduced novel public health measures such as masking and social distancing. In adults, framing these behaviors as benefiting others versus the self has been shown to affect people’s perceptions of public health measures and willingness to comply. Here we asked whether self- versus other-oriented frames of novel public health measures influence children’s endorsement and moral reasoning. Children aged 5 to 10 years viewed hypothetical dilemmas of aliens in which we manipulated the frame (other-oriented or self-oriented) of the prevention behavior and the severity (high or low) of the potential harm. Across two studies (Study 1: N = 48; Study 2: N = 61), results showed that across ages framing the behaviors as other-oriented, but not self-oriented, yielded more positive ratings of individuals who followed the public health measures and more negative ratings of those who did not. Across both frames, children generally endorsed these public health measures when the severity was high. Children used more moralizing concepts in other-oriented frames and were more critical of intentional transgressions over accidental transgressions, demonstrating further evidence that other-oriented frames induce moral reasoning. We discuss the theoretical and practical implications of these framing effects for sociomoral reasoning and action.

Introduction

The COVID-19 pandemic upended the lives of children around the world: schools closed, playdates were canceled, and visits with family members became Facetime calls. One particularly salient change for adults and children alike was the introduction of novel public health measures such as masking and social distancing. For children, masking became an increasingly important behavior in slowing the spread of COVID-19 so that schools could safely re-open (Donovan et al., 2022). Because behaviors such as masking are crucial in preventing the spread of COVID-19, it is imperative to identify how to best communicate these measures to children to maximize their compliance. Moreover, it raises important questions about children’s sociomoral reasoning that informs theories of moral development and in turn can have practical implications for what kinds of messaging is most persuasive. Preventative behaviors such as masking and social distancing have both self- and other-benefiting effects: wearing a mask can prevent one from getting the virus as well as prevent one from spreading the virus to others (Gandhi et al., 2020). It is not yet known whether children focus more on the selfish or prosocial aspect of mask wearing. This is a central question given that framing influences the attitudes and perceptions of public health measures in adults (e.g., Ceylan and Hayran, 2021, Grant and Hofmann, 2011, Jordan et al., 2021, Luttrell and Petty, 2021). For this reason, we first review findings from adults and then develop our rationale for testing for such effects in children.

Framing effects occur when the way information is presented influences how people think about that information (Chong & Druckman, 2007). Framing effects have been demonstrated in a variety of contexts and have been distinguished as emphasis framing effects and equivalency framing effects. Equivalency framing effects present logically equivalent information in different perspectives such as framing a disease or disaster in terms of lives saved versus lives lost (e.g., Tversky & Kahneman, 1981). On the other hand, we focus here on emphasis framing effects, which highlight different relevant aspects of an issue such as framing a policy in terms of economic versus social impact (see Chong & Druckman, 2007). Most relevant to the current studies, message framing effects that characterize a behavior as either self- or other-focused have been found to influence health-related decisions: One famous study investigated framing effects in hospitals and found that posting signage emphasizing the importance of hand washing for protecting others was more effective than signage framing it as protecting the self (Grant & Hofmann, 2011).

More recently, a few studies with adults have applied the effects of self versus other framing to COVID-19 preventative measures. Ceylan and Hayran (2021) manipulated the framing of messages regarding social distancing to prevent COVID-19 as either prosocial (“For all our health, stay home”) or self-interested (“For your own health, stay home”) and found that participants in the prosocial condition found the message to be more persuasive and were more motivated to help others. Jordan et al. (2021) conducted a similar study and found that, at least early in the pandemic, messaging emphasizing the benefits of preventative measures for others yielded higher levels of prevention intentions. In addition, Gillman et al. (2022) found that people were more receptive to other-focused messages about COVID-19 than to self-focused messages, suggesting that prosocial frames may be an effective way to communicate about novel public health measures.

One reason why other-focused framings may be more effective than self-focused framings is that other-focused framings are more morally relevant due to their focus on the interpersonal consequences of an action. Luttrell and Petty (2021) found that people were more likely to perceive other-focused messages regarding social distancing as moral arguments than self-focused messages. Thus, other-focused framings seem to tap into people’s moral reasoning, which may make them a particularly effective communication tactic for adults and children alike.

In contrast to this research with adults, it is not known whether similar effects would be found among children. Evidence from other domains, such as gain and loss framing, suggests that children are susceptible to framing effects on decision making, similar to adults (e.g., Schlottmann and Tring, 2005, Wyllie et al., 2015). However, no studies to date have investigated the effectiveness of self- and other-oriented framings on preventative health measures in children. This is important to explore considering that other-oriented frames seem to have moralizing effects in addition to affecting decision making.

Although no study has investigated how framing might influence children’s evaluations of public health measures, developmental research has started to address children’s knowledge of COVID-19 and corresponding health measures. For example, a COVID-19 “quiz” administered to children aged 6 to 16 years in India asked children about their awareness of how COVID-19 is transmitted, its symptoms, and measures to prevent its spread such as social distancing, hand washing, and masking (Shaikh & Likhite, 2020). The results showed that the majority of children’s scores (93%) fell within the “good” and “excellent” knowledge categories, suggesting that children generally had a good understanding of what COVID-19 is and how to prevent it. In addition, qualitative studies with children in several different countries found that school-aged children are concerned about COVID-19 and protecting themselves and others from it (Bray et al., 2021, Sarkadi et al., 2021, Thompson et al., 2021). Finally, children participating in COVID-19-related studies in the United States (Howe et al., 2021) and China (Xue et al., 2021) reported high levels of adherence to preventative health guidelines. Therefore, children pick up on information about COVID-19 and are familiar with preventative measures. Consequently, it would be prudent to explore how children might be susceptible to self- versus other-oriented framing—not only for the current pandemic but also for future situations where novel preventative measures would be necessary.

Will children, like adults, find other-oriented framings to be more persuasive and moralizing? Children are attuned to moral decision making from a young age. Even beginning during infancy, children prefer more moral characters and are capable of evaluating moral situations (Hamlin, 2013). As children progress through childhood, they show increasingly nuanced moral decision making, considering contextual factors as they determine right from wrong across several domains (for an overview, see Killen & Smetana, 2015). Before entering middle childhood, children reason and respond with moral convictions; for example, they intervene in the face of third-party transgressions and police the behaviors of others (Heyman et al., 2016, Vaish et al., 2011).

Further evidence also suggests that framing behaviors as other-oriented would be moralizing and effective for children. It is known that by preschool ages children rate moral transgressions against others as more wrong than violations of personal rules (Tisak, 1993). Because of this, other-oriented frames may provoke a sense of morality that would not be induced in self-oriented frames. Moreover, children by middle childhood view the avoidance of transgressing against others as morally obligatory—even more obligatory than engaging in positive moral acts (Kahn, 1992). In other words, it is more important to children to not harm others than it is to help others. Taken together, although none of this work tested other- versus self-regarding frames, it supports the prediction that children would be more sensitive to other-oriented appeals and that they may elicit moral thinking.

An additional study manipulated whether school-aged children saw somebody lie to benefit themselves or others (Fu et al., 2015). Whereas children judged both types of lying as bad, they judged selfish lying as worse than lying to benefit others and found those who lied to benefit themselves to be untrustworthy. This suggests that the same act can be judged differently depending on who it benefits and may indicate that in our public health context children will be more favorable to behaviors that benefit others.

Although we so far have argued for the hypothesis in favor of other-oriented frames over self-oriented frames, we also must consider the counterhypothesis that self-oriented frames prove to be more effective. This alternative hypothesis is not without merit; young children (i.e., 4- and 5-year-olds) are known to consider their own self-interest in resource sharing and decision making more strongly than older children (i.e., 8- and 9-year-olds) and adults (e.g., McAuliffe et al., 2017). In addition, children who are primed to think of themselves demonstrate less helping and sharing behaviors, suggesting that self-oriented framings may influence children’s social cognition and behavior (Weltzien et al., 2018). Therefore, although other-oriented frames may be effective because they prompt children to moralize preventative behaviors, it might turn out that self-oriented framings are more effective drivers of behavior because they prompt children to think about preserving their own well-being. The current studies aimed to adjudicate between these two hypotheses.

The current studies

Here we tested the effects of self- versus other-oriented framing on children’s moral evaluations of public health measures similar to COVID-19 prevention behaviors. We deliberately chose to use hypothetical scenarios with novel public health measures because of the polarized opinions regarding the pandemic and possible preconceived notions by children about specific rules such as mask wearing and social distancing. This was a particularly strong concern given that our data collection took place during the summer and fall of 2021, after children were already exposed to a multitude of information about COVID-19. Furthermore, as developmental psychologists interested in cognitive development, our aim was to understand the underlying cognitive processes and representations of children rather than children’s specific opinions regarding mask wearing and social distancing. Therefore, instead of asking directly about COVID-19 prevention behaviors, we constructed hypothetical situations that mirror the inherent dilemmas caused by behaviors such as mask wearing, wherein a prevention behavior has a health benefit to the self or others but comes with a cost to the self because of the disutility of needing to wear a mask or not being able to move freely without needing to keep social distance. This methodological approach has a long tradition in moral psychology: To get at the underlying concepts, it is beneficial to use hypothetical scenarios that are structurally similar to real-life moral problems rather than directly asking individuals about current controversies (e.g., abortion, euthanasia, fair wages) for which people might just repeat rehearsed answers or have preconceived notions that mask their actual reasoning (see Greene, 2013, and Mook, 1983, for extensive discussions on the advantages of such methods to understand psychological mechanisms).

The general method across our series of experiments was to use illustrated hypothetical scenarios involving novel aliens—an approach that has been shown to be useful in other moral decision-making studies with children (i.e., Rottman & Kelemen, 2012). Specifically, children viewed vignettes of alien characters who faced dilemmas such as covering up spikes on their body with heavy caps to prevent poking themselves (self-oriented frame) or poking other aliens (other-oriented frame). We also manipulated the severity of these situations, with spikes either poking (high severity) or brushing against (low severity) the recipient. The reason for this manipulation was to determine whether children would be attentive to the potential harm caused by the action and adjust their responses accordingly. We used several structurally similar scenarios with different kinds of dilemmas. Children were then asked to evaluate the behaviors and express their preferences for agents who did or did not perform the protective behaviors.

Because this topic has not yet been addressed in children, we decided to recruit a broad age range to track any developmental changes that may occur throughout middle childhood. Previous research suggests that moral reasoning abilities are fairly well developed by 5 years of age (see Killen & Smetana, 2022). In addition, piloting suggested that 5-year-olds were a viable age group for online testing, whereas younger children had difficulties. For these two reasons, we used 5 years as the lower bound of our age range. The upper bound of our age range was set to 10 years because this generally constitutes the end of middle childhood and we were most interested in developmental change across this life period.

We conducted two studies. The main goals of Study 1 were to validate our paradigm and assess whether framing the protective behaviors as either self- or other-oriented influenced children’s evaluations. In a within-participants 2 × 2 design, we manipulated the framing (other-oriented vs. self-oriented) and severity (low vs. high) of the aliens’ dilemmas. Our dependent measures were children’s endorsement of the behavior and their evaluations of individuals who did or did not perform the protective behaviors. We hypothesized that children would be more likely to endorse the protective behaviors and show more positive sociomoral evaluations of followers over violators in the other-oriented condition over the self-oriented condition. We also hypothesized that children would be more likely to endorse the behaviors and show more sociomoral reasoning in the high-severity condition compared with the low-severity condition and that there would be an interaction such that the high-severity/other-oriented condition would yield the highest rate of endorsement and sociomoral judgment. We had no predictions about developmental changes but included age as an exploratory variable. The same applied to gender.

Study 2 was conducted with a separate sample and had two main parts. The goal of Study 2a was to conduct a direct replication of the main results from Study 1 in a larger sample. Study 2b was conducted directly after Study 2a with the same children in the same test session and aimed to extend our previous findings by further exploring children’s moral judgments within the other-oriented framing condition. All methods and hypotheses were preregistered at AsPredicted (Study 1: https://aspredicted.org/DMD_YYK; Study 2a: https://aspredicted.org/4P8_SRY; Study2b: https://aspredicted.org/NVJ_7XN).

Study 1

Method

Participants

The final sample was 48 children aged 5 to 10 years (M = 8.02 years, SD = 1.73, with equal numbers of girls and boys per age). One additional participant was excluded from analyses because of failure to pass the comprehension checks. Parents identified their children as follows: 10% as Asian American, 4% as Black/African American, 69% as White/Caucasian, and 17% as multiple races/ethnicities. Participants were recruited primarily through a research database hosted at the authors’ home institution, the lab website, and social media ads.

Design

We employed a 2 × 2 design with the variables severity (high or low) and framing (other-oriented or self-oriented) tested within participants. Therefore, each child saw four different aliens representing the four different combinations of manipulations (high-severity/other-oriented, high severity/self-oriented, low severity/other-oriented, and low-severity/self-oriented). We counterbalanced the order and combinations of conditions, as well as the order and conditions of the aliens, to create 16 unique sequences that were randomly assigned to children within each of three age groups created for counterbalancing purposes (5–6, 7–8, and 9–10 years).

Procedure

This study was conducted online by an experimenter over Zoom. Parents completed consent forms prior to the study session. During the test session, the researcher verified the parent’s consent and obtained verbal assent from the child. The study was presented via Qualtrics, with the experimenter sharing their screen with the child. All stimuli were read aloud, and the child’s responses were obtained verbally and recorded in Qualtrics by the researcher as the study proceeded.

We included two manipulation check questions and one comprehension check for each vignette (described below). If the child answered a manipulation or comprehension check question incorrectly, that portion of the vignette was repeated and the child was re-asked the question. If the child failed the question a second time, the incorrect answer was recorded and that trial was subsequently excluded from analysis. Of the 192 trials (4 per child, 48 children), 8 trials were excluded due to failed manipulation checks and 7 trials were excluded due to failed comprehension checks for a total of 15 excluded trials (<8%).

The procedure was the same for all four vignettes presented in counterbalanced order (see above). At the end of the session, we thanked the child for participating and e-mailed a $5 Amazon online gift code and a form explaining the purpose of our research to the child’s parents.

Stimuli

The stimuli for this study consisted of cartoon-like drawings of aliens. There were four different types of aliens, each with some bodily appendage that presented a potential problem for themselves or others, depending on the condition. As shown in Fig. 1 , the Furpees had large spikes on their bodies, the Ollers had spiky balls on the end of their tails, the Blickets had spiky antennas, and the Zibs had tails with spiky protrusions. The type of harm or inconvenience (i.e., poking or brushing against) was held constant. Children saw illustrations of the aliens alongside text vignettes that were read aloud by the experimenter. We provide the full text and image stimuli, as well as dependent variables, as presented to the children in the online supplementary material.

Fig. 1.

Examples of stimuli used in the studies. The top row is the introduction image for each alien (Furpees, Ollers, Blickets, or Zibs), and the bottom row is the prevention behavior for each alien.

Each vignette opened with an introduction to the alien, the planet the alien lives on, and a short description of the alien’s appendage. An illustration depicting three of these aliens in different colors accompanied the description (see the top row in Fig. 1). This introduction was held constant in all conditions for each of the aliens.

In the next step, the alien either poked or brushed against the self or another alien. This step introduced the two manipulations: framing and severity. The framing manipulation refers to whether the alien’s appendage affects the self (self-oriented) or another alien (other-oriented); the severity manipulation refers to whether the alien’s appendage pokes (high severity) or brushes against (low severity) the self or the anther alien (e.g., “Sometimes the Furpees’ spikes poke other Furpees” [high-severity/other-oriented]).

Following the depiction of the manipulation, a speech bubble reemphasized the severity (“Ouch!” in the high-severity condition and “Oh!” in the low-severity condition) along with an additional description stating whether this hurts (high severity) or does not hurt (low severity). Two manipulation checks assessed whether the children understood (a) which alien is affected (e.g., “Who does the Furpee poke?”) and (b) whether or not harm is caused (e.g., “Does it hurt when the Furpee pokes other Furpees?”).

Next, the children saw the protective behavior that prevented the alien from poking or brushing against the self or others. The protective behaviors, intended to reflect the inherent conflict of health measures such as mask wearing in that they provide protection but are uncomfortable, were as follows: The Furpees could cover up their spikes with caps, the Blickets could put a hat over their antennas, the Zibs could put a wrap over their wings, and the Ollers could put a cover over their tail. The comprehension check question here assessed whether the children understood the discomfort of the alien (“How do you think they feel? Comfortable or uncomfortable?”). At the end of each vignette, children saw the result of doing the protective behavior (i.e., not poking or brushing against the self or others).

Dependent variables

Immediately after each story vignette, children saw a different member of the alien type (denoted by a different color) debating whether to perform the protective behavior. We asked children whether or not the alien should do the behavior (endorsement), followed by a free response question of “Why?” to keep their attention. Because their justifications were purely exploratory and not part of our hypothesis, we did not include them in our main analysis; however, we do provide a description of general trends in the Results section.

Children then saw both a violator alien (an alien who failed to do the prevention measure) and a follower alien (an alien who decided to do the prevention measure). For each of the four vignettes, the children first separately evaluated the follower and the violator (order counterbalanced) on Likert scales assessing their action rating (on a 5-point scale from very bad to very good), friendship quality (on a 4-point scale from very bad friend to very good friend), and smartness (on a 4-point scale from very not smart to very smart). Then, children responded to two forced-choice questions where the participants chose which alien (follower or violator) they would rather be friends with (friend choice) and which is smarter (smarter alien choice). The questions asking how good or bad the follower and violator aliens’ actions were and the questions assessing how good or bad friends the aliens were constitute sociomoral questions and allowed us to examine whether our manipulations affected children’s moral thinking about others. The inclusion of the smartness questions allowed us to assess whether the effect of framing is specific to sociomoral evaluation or is evident in other more general evaluations of competency.

For the Likert scale ratings, difference scores were calculated by subtracting the children’s ratings of the violator alien from their ratings of the follower alien. These difference scores were used as the dependent variables; higher positive scores reflect more positive ratings of the follower and more negative ratings of the violator, and lower scores reflect more similar ratings of the two aliens. Higher positive difference scores on the action rating variable would suggest that children see a refusal to perform the protective barrier as morally bad, and higher positive difference scores according to framing on the friendship quality variable along with a higher proportion of follower on the friend choice variable would suggest that this extends to a moral judgment about the alien as a social partner. Effects of the independent variables on smartness and smarter alien choice would suggest that children’s evaluations of the aliens are more global and not specific to sociomoral considerations. See Table 1 for a summary of the six dependent variables.

Table 1.

Summary of dependent variables

Variable name	Description	Text of question	Measurement
Endorsement	Whether or not the alien should do the behavior	What do you think? Should the [alien] [do the protective measure]?a	Binary: Yes/No
Action rating	Rating of the follower’s and violator’s action	Do you think what the [color] [alien] did was good, bad, or just OK? A little good/bad or very good/bad?	Continuous: Difference score between two 5-point Likert scales
Friendship quality	Rating of the follower’s and violator’s friendship quality	Do you think the [color] [alien] is a good or bad friend? A little good/bad or very good/bad?	Continuous: Difference score between two 4-point Likert scales
Smartness	Rating of the follower’s and violator’s smartness	Do you think the [color] [alien] is smart or not smart? A little smart/not smart or very smart/not smart?	Continuous: Difference score between two 4-point Likert scales
Friend choice	Forced choice of which alien is a better friend	Which of the two [aliens] would you rather be friends with?	Binary: Follower/Violator
Smarter alien choice	Forced choice of which alien is smarter	Which of the two [aliens] is smarter?	Binary: Follower/Violator

^aExact wording depends on the alien type (see online supplementary material).

Analysis approach

We used generalized linear mixed models for the three binary dependent variables (endorsement, friend choice, and smarter alien choice) and used linear mixed effects models for the three continuous dependent variables (action rating, friendship quality, and smartness). All analyses were performed in R Version 3.6.3 using the package lme4 (Bates et al., 2015).

For each model, the fixed effects were trial, gender, age, framing, severity, and the interaction between framing and severity. Participant ID was included as a random effect to account for within-participants testing. For each analysis, we first compared a full model that included all predictors of interest with a null model containing only trial and participant to test whether the predictors combined had a significant effect on our dependent variable, preempting concerns of multiple analyses inflating the rate of false positive results (Forstmeier & Schielzeth, 2011). If this comparison showed a significant effect, we proceeded by comparing the full model against a series of hypothesis-driven models to compare model fit using likelihood ratio tests and assessing the effects of each predictor on the dependent variables. Age and gender were preregistered as exploratory analyses because we did not have any prior hypotheses regarding their potential effects. We include a regression table of the final models (age and gender were included in final models only if they showed an effect in model comparisons) for each of the following analyses in the supplementary material. In these regression tables, the beta values can be interpreted as effect sizes because the variables are standardized. A summary of the significant effects of Studies 1 and 2a can be found in Table 2 .

Table 2.

Summary of results from Studies 1 and 2a broken down by dependent variable and listing all significant effects

Dependent variable	Study 1	Study 2a
Endorsement	• Severity	• Severity
	• Age × Severity
Action rating	• Framing	• Framing
	• Severity	• Severity
	• Gender	• Age × Severity
Friendship quality rating	• Framing	• Framing
	• Severity	• Severity
	• Gender
Friend choice	• Severity	• Framing
	• Gender	• Severity
		• Age
Smartness rating	• Severity
	• Gender
Smarter alien choice	• Severity

Results

Endorsement of behavior

Our first question was whether our experimental manipulation influenced children’s endorsement of the protective behaviors (Fig. 2 A). Specifically, we built a generalized linear mixed model with the aforementioned predictors and children’s endorsement of the protective behaviors as a binary outcome. There was a significant difference between the full model and the null model, χ ²(5) = 62.21, p < .001. Model comparisons revealed a main effect of severity, reflecting stronger endorsement of the behavior in the high-severity condition compared with the low-severity condition, χ ²(2) = 59.08, p < .001. There was no effect of framing, gender, age, or the interaction between framing and severity. However, there was a significant interaction between age and severity; younger children showed high levels of endorsement even in the low-severity condition, and with age children endorsed less in the low-severity condition while consistently showing near ceiling levels of endorsement in the high-severity condition, χ ²(1) = 15.02, p < .001. Overall, severity appears to matter the most for children’s endorsement of the protective behaviors.

Fig. 2.

Bar charts of model-predicted values for Study 1 by framing (self-oriented vs. other-oriented) and severity (low vs. high). The dependent variables are endorsement (A), action rating (B), friendship quality (C), friend choice (D), smartness (E), and smarter alien choice (F). Error bars represent 95% confidence intervals. Note that the y axis for Panels B, C, and E are difference scores calculated by subtracting the violator rating from the follower rating.

Sociomoral evaluations

The next series of analyses assessed children’s sociomoral evaluations of aliens who did (follower) or did not (violator) perform the protective behavior by assessing children’s rating of the action rating (Fig. 2B), friendship quality (Fig. 2C), and friend choice (Fig. 2D).

Action rating

We built a linear mixed model to examine how children differentially evaluate the follower’s and violator’s decisions to do or not do the protective behavior, using the difference score as a continuous outcome variable. There was a significant difference between the full model and the null model, χ ²(5) = 68.35, p < .001. Model comparisons revealed a significant main effect of severity; in the high-severity condition, children had higher positive difference scores, reflecting higher ratings of the follower over the violator, whereas in the low-severity condition, children rated the followers and violators similarly, χ ²(2) = 54.25, p < .001. In addition, there was a main effect of framing; when the behavior was framed as other-oriented, children had more positive difference scores, whereas when the behavior was framed as self-oriented, children’s difference scores hovered around 0, χ ²(2) = 7.79, p = .020. There was no effect of age or the interaction between severity and framing. However, there was an effect of gender, with girls rating violators and followers more evenly than boys, who tended to show a slight favoring of violators overall, χ ²(1) = 8.20, p = .004. Taken together, these analyses show that children evaluate aliens who do protective behaviors more positively and evaluate aliens who do not do protective behaviors more negatively when the behaviors prevent harm or protect others. By contrast, children show similar ratings of followers and violators when the behaviors only prevent mildly inconvenient outcomes or protect the self.

Friendship quality

We constructed a linear mixed model to examine whether severity or framing predicted differences in children’s rating of how good or bad of friends the violator and follower are, using the difference score for this measure as a continuous outcome variable. There was a significant difference between the full model and the null model, χ ²(5) = 57.50, p < .001. Model comparisons revealed that there was a significant main effect of severity, such that children in the high-severity condition had higher positive difference scores, rating the follower as a better friend and the violator as a worse friend, than children in the low-severity condition, whose difference scores were near zero, rating the aliens more similarly, χ ²(2) = 25.65, p < .001. There was also a main effect of framing, such that children in the other-oriented condition had higher positive difference scores, rating the follower as a better friend and the violator as a worse friend, than children in the self-oriented condition, whose difference scores were near zero, rating the two aliens more similarly, χ ²(2) = 30.30, p < .001. There was no effect of the interaction between framing and severity and no effect of age. However, like the ratings of the aliens’ actions, there was a slight main effect of gender where boys tended to have more negative difference scores, slightly favoring the violators, whereas girls tended to have more positive difference scores, slightly favoring the followers, χ ²(1) = 4.63, p = .031. Overall, children think that aliens who do protective behaviors are better friends than those who do not when the behaviors prevent harm or protect others, and they think that follower and violator aliens have about the same friendship quality when the behaviors prevent inconvenience or protect the self.

Friend choice

The binary measure of whether the child would rather be friends with (the violator or the follower) mirrored the effects found using the difference score measure. The full model significantly differed from the null model, χ ²(5) = 59.55, p < .001. Model comparisons revealed a main effect of severity, such that children’s preference for the follower as a friend was more pronounced in the high-severity condition compared with the low-severity condition, χ ²(2) = 50.39, p < .001. In addition, there was a marginal main effect of framing, such that children chose the follower alien as the better friend more when the behavior was framed as other-oriented compared with when the behavior was framed as self-oriented, χ ²(2) = 6.17, p = .046. Although this effect is slightly weaker than the scale response questions, Fig. 2D demonstrates that this is because a majority of children chose the follower in the high-severity condition regardless of framing, suggesting a near ceiling effect. Overall, our results show that children think that aliens who perform protective behaviors are better friends when those behaviors prevent harm or when those behaviors protect others. There was no effect of age or the interaction between severity and framing. However, there was an effect of gender, with girls choosing the follower more and boys choosing the violator more, χ ²(1) = 5.70, p = .020.

Across three sociomoral evaluation measures, both severity and framing affected children’s ratings and choices regarding the follower and violator aliens. Children thought that followers were better friends and performed more morally good actions than violators if the protective behaviors aimed at protecting others from harm.

Evaluation of smartness

In the final series of analyses, we examined children’s evaluations of the aliens’ smartness with the goal of assessing whether framing effects are specific to sociomoral evaluations or might engender generally more positive evaluations of norm followers. Our dependent variable was the difference score of children’s smartness ratings for the violator and follower (Fig. 2E) and the binary measure of which alien is smarter (Fig. 2F).

We ran a linear mixed model to examine whether severity or framing predicted differences in children’s rating of how smart or not smart the violator and follower are. The dependent variable in this model was a difference score calculated by subtracting the children’s 4-point Likert scale rating of the violator’s smartness from their 4-point Likert scale rating of the follower’s smartness. There was a significant difference between the full model and the null model, χ ²(5) = 61.42, p < .001. Model comparisons revealed a main effect of severity; children in the high-severity condition had higher positive difference scores, rating followers as smarter and violators as less smart, compared with children in the low-severity condition, whose difference scores were near zero, rating the follower and violator aliens about the same, χ ²(2) = 52.55, p < .001. There was also an effect of gender, driven by boys’ slightly negative difference scores, revealing that they had higher ratings for the violator overall, χ ²(1) = 5.78, p = .016. A marginal trend of framing also emerged here, with children in the other-oriented condition having higher positive difference scores, rating followers more positively than violators in the other-oriented condition, and children in the self-oriented condition having difference scores closer to zero, rating followers and violators about the same, χ ²(2) = 5.75, p = .057. It is worth noting that although this trend looks similar to the results of the sociomoral evaluation measures, it is a smaller effect and does not reach statistical significance, suggesting that framing has a lesser impact on smartness evaluations. There was no effect of age or the interaction between severity and framing. Thus, children think that followers are smarter than violators when the behaviors prevent harm to the self or to others. In contrast to sociomoral evaluations, framing these behaviors as other-oriented does not have a significant effect on evaluations of aliens being smart.

The binary measure of which alien is smarter provided more evidence that the self- and other-oriented framings were not as influential when it comes to the issue of competency. We constructed a generalized linear mixed model with participants’ choice of the smarter alien (follower or violator) as the dependent measure. The full model significantly differed from the null model, χ ²(5) = 42.70, p < .001, and subsequent model comparison revealed a significant main effect of severity; children said that the follower alien was the smarter alien more in the high-severity condition than in the low-severity condition, χ ²(2) = 39.18, p < .001. Nearly all children in the high-severity condition chose the follower as the smarter alien, whereas slightly more than half of children chose the follower in the low-severity condition (Fig. 2F). However, there was no effect of framing, age, gender, or the interaction between framing and age. Therefore, we conclude that children think that aliens who do protective behaviors are smarter than those who do not when those behaviors prevent harm, but whether the behaviors protect the self or others does not have any bearing on the aliens’ perceived smartness.

Justifications

In addition to the quantitative measures aimed at testing our hypotheses, we also added a “Why?” question to keep children engaged. We coded these responses in the interest of transparency and for exploratory purposes even though we had no specific predictions on how children would respond. Research assistants transcribed children’s responses from video-recordings, and those transcriptions were used to generate a coding scheme. The first author then coded each response. In most cases, children’s responses were repetitions of aspects of the stories; most commonly, justifications for “yes” focused on the outcome of the preventative measure either for others (46% of responses; e.g., “So other Furpees don’t get hurt) or for the self (32% of responses; e.g., “So they don’t poke themselves), and justifications for “no” focused on the discomfort for the protagonist (33% of responses; e.g., “Because it’s very uncomfortable”). A few responses reflected a cost–benefit analysis, with children weighing the pros and cons of the behavior (17% of responses; e.g., “Even though it’s uncomfortable, it’s better than getting hurt”). In addition, 12% of responses mentioned an alternative solution to the dilemma (e.g., “Because they could tie up their antennas and make a braid or something so it doesn’t brush against anyone”) and 3% directly cited moral principles (e.g., “Even if it feels bad, you should always help others”). Finally, 2% of respondents were unable to give any justification (e.g., “I don’t know”). Overall, these justifications did not add much to our understanding of the phenomena but did provide additional evidence beyond our manipulation checks that children were attending to the dilemmas and manipulations.

Discussion

These analyses indicate that the effect of framing behaviors as other-oriented or self-oriented affects sociomoral evaluations but does not affect a more general evaluation of the aliens’ competency. The current findings seem to suggest that children find protective measures to be wise and practical regardless of framing; they endorse the behaviors and rate them as smart in both framings provided that they prevent harm (i.e., high-severity context). However, the self- and other-oriented framings do have an effect when it comes to how children evaluate the social and moral characteristics of aliens who do or do not do the behaviors. Thus, we can tentatively conclude that framing novel public health behaviors as other-oriented taps into children’s sociomoral thinking.

There were a few limitations to Study 1 that we hoped to address in Study 2. First, we recruited a modest sample size for Study 1; in the absence of any prior work on this topic, our sample of 48 was chosen based on convention rather than power analyses using preexisting effect sizes. Therefore, our aim in Study 2a was to replicate the findings of Study 1 in a larger sample determined with power analyses using the effect sizes of the initial study. In addition, we aimed to further explore children’s moral thinking by focusing on the other-oriented condition in particular. One important aspect of children’s moral reasoning is the attribution of intentionality; children consider the role of intention in moral violations, showing more negative evaluations of intentional violators over accidental violators (e.g., Cushman et al., 2013, Grueneich, 1982, Piaget, 1932/1965). In Study 2b, we manipulated the intention of aliens who did not perform the protective measures to observe whether children would show similar patterns of responding as they do in other moral contexts.

Study 2

Our first study indicated that other-oriented framing elicits children’s sociomoral reasoning. As a next step, we aimed to study how children reason about the intentions of the protagonists who fail to engage in protective health measures. This is because the ability to take into account the intentionality behind a transgression is regarded as a fundamental aspect of moral reasoning. If children show sensitivity to intention in this domain, it would suggest that other-oriented frames induce a moral norm and lead children to think about violators of public health behaviors in moral terms.

A long tradition of developmental research has highlighted how reasoning about intentions influences children’s moral judgment differentially over development; younger children tend to focus more strongly on the outcomes of an action, whereas older children are more likely to evaluate norm violators based on whether they are well- or ill-intentioned (Piaget, 1932, Piaget, 1965). For example, younger children are more likely to make more negative evaluations of people who have caused negative outcomes even with the knowledge of these actions as unintentional (Grueneich, 1982). By around 4 to 8 years of age, children make more intent-based judgments of the actors, and this developmental shift to a more intent-based moral judgment (e.g., Costanzo et al., 1973, Hebble, 1971, Piaget, 1965, Yuill and Perner, 1988) highlights an important developmental shift toward evaluating actions based on intentions rather than on mere outcomes.

Although various research has addressed the role of intentionality in children’s judgment, the concept of an explicit shift from outcome to intention-focused judgments and the trend of this shift developmentally is still debated. Work with infants’ behavioral responses showcased an understanding of intentional and goal-related actions at around 9 to 12 months of age (Behne et al., 2005, Kuhlmeier et al., 2003). Intentionality is an important factor in interpreting others’ behavior. The trend of considering intentionality in moral evaluations may occur earlier than previously predicted considering the importance in understanding intentions for a child’s social development. Children as young as 4 or 5 years consider intentionality as a basis of evaluating behavior when the amount of harm or damage caused is held constant (Costanzo et al., 1973). However, although 4-year-old children can consider relevant social information in their moral evaluations, with age they value intention more and by 8 years of age they tend to think of accidental harmers as less naughty and punishable and of intentional harmers as more naughty and punishable (Cushman et al., 2013). These studies articulate how this outcome intention shift is less precise than we previously have assumed. There are important developments involving intentionality from infancy to puberty. Therefore, this prior work suggests that across our sample of 5- to 10-year-olds, children might already be able to take intention into account when making moral decisions. Moreover, the wide age range allowed us to observe possible developments in incorporating more social information into children’s moral evaluations. As such, we included age as an exploratory variable with no specific predictions about it to investigate further into this wide range of development.

Thus, the aims of Study 2 were to (a) conduct a replication of the framing effects from Study 1 and (b) further investigate whether other-oriented framing induces moral reasoning. We predicted that Study 2a would replicate the findings from Study 1—framing would have an effect on our sociomoral variables but not on endorsement, and severity would have an impact on all variables. In Study 2b, we predicted that children would be sensitive to intention in their evaluations of violators, which would provide further evidence of moral reasoning in this context.

Study 2a

Method

Participants

The final sample for Study 2a was 61 children aged 5 to 10 years (M = 7.79 years, SD = 1.71; 30 girls). Parents identified their children as follows: 3% as Asian American, 3% as Black/African American, 5% as Hispanic/Latino, 72% as White/Caucasian, and 16% as multiple races/ethnicities. As with Study 1, participants were recruited primarily through a research database hosted at the authors’ home institution, the lab website, and social media ads. The sample size was determined by a power analysis using the SimR function in R, based on the results from Study 1. We computed the minimum sample size necessary to detect an effect of framing with at least 80% power on one of our primary variables, action rating. We chose this variable because it had a smaller effect than friendship quality, thereby ensuring it had sufficient power to adequately assess all variables. We chose to use the effect of framing in this power analysis because it is the variable we were most interested in conceptually. We did not incorporate age into our power analysis because the age effects we found in Study 1 were not consistent across dependent variables, and therefore we did not have predictions about age for Study 2a. The power analysis revealed that a sample size of 54 provided adequate power to detect an effect of framing [84.2%, confidence interval = (81.79, 86.41)]. Thus, we chose a target sample of 60 children to meet this threshold and allow for adequate counterbalancing across age groups.

Procedure, stimuli, and dependent measures

The procedure and stimuli used for this portion of Study 2 were identical to those used in Study 1. As in Study 1, we asked both manipulation and comprehension checks throughout the procedure, and we excluded trials in which children failed either manipulation or comprehension checks. There were 6 trials where children failed manipulation checks, and 4 trials where children failed comprehension checks. One of the failed manipulation checks was the same trial as one of the failed comprehension checks. In addition, 1 participant completed only 2 of 4 trials. As such, 11 of 244 trials were excluded from analysis (<5%).

All the measures for this study were identical to those used in Study 1, with the exception that we did not include the measures assessing smartness because this was a variable that was not of primary concern. We also excluded the “Why?” justification question to reduce the overall session length given that it did not yield meaningful insights in Study 1. In Study 2a, we focused on the four measures that were central to testing our question about sociomoral reasoning: endorsement, action rating, friendship quality, and friend choice.

Analysis

Our analysis approach was identical to that in Study 1, including our preregistration with age and gender as exploratory variables (https://aspredicted.org/4P8_SRY). See the supplementary material for the regression tables for each of the analyses below.

Results and discussion

Endorsement of behavior

We created a generalized linear mixed model to determine whether severity and/or framing affected children’s endorsement of protective behaviors (Fig. 3 A). The predictors in this model are specified above, and endorsement (yes or no) was the binary dependent variable. Results showed a significant difference between the full model and the null model, χ ²(5) = 70.45, p < .001, so we proceed with model comparisons. As in Study 1, there was a main effect of severity, such that children were more likely to endorse the protective behavior in the high-severity condition compared with the low-severity condition, χ ²(2) = 68.35, p < .001. Nearly all children endorsed the behavior in the high-severity condition, whereas only a little more than a quarter of children endorsed it in the low-severity condition. There was no effect of framing, gender, age, or an interaction between framing and severity on endorsement. However, there was a marginal interaction between severity and age, similar to Study 1; with age, children endorsed less in the low-severity condition, χ ²(1) = 3.77, p = .052.

Fig. 3.

Bar charts of model-predicted values for Study 2a by framing (self-oriented vs. other-oriented) and severity (low vs. high). The dependent variables are endorsement (A), action rating (B), friendship quality (C), and better friend choice (D). Error bars represent 95% confidence intervals. Note that the y axis for Panels B and C are difference scores calculated by subtracting the violator rating from the follower rating.

Thus, Study 2a replicated the effects found in Study 1 and demonstrated that in this higher-powered sample children show similar levels of endorsement across framing conditions but endorse much higher when the protective behavior prevents harm than when it only prevents inconvenience. In fact, our model predicted nearly ceiling levels of endorsement in the high-severity condition, suggesting that children find these behaviors to be prudent when they prevent harm regardless of whether that harm is directed toward the self or others.

Sociomoral evaluations

Action rating

We built a linear mixed model to examine how children differentially evaluated the follower’s and violator’s decisions, using the difference score for this measure as the continuous outcome measure (Fig. 3B). There was a significant difference between the full model and the null model, χ ²(5) = 57.61, p < .001. As in Study 1, there was a main effect of both severity, χ ²(2) = 49.84, p < .001, and framing, χ ²(2) = 10.22, p = .006. In both the high-severity and other-oriented framing conditions, children showed higher positive difference scores, rating the follower more positively and the violator more negatively, whereas in the low-severity and self-oriented conditions, children had difference scores closer to zero, reflecting more even evaluations of the follower and violator. There was no effect of gender or age, nor was there an interaction between framing and severity. However, there was an interaction between age and severity; with age, children showed more differential ratings of the follower and violator in the high-severity condition and more equal ratings in the low-severity condition, χ ²(1) = 6.92, p = .009. Overall, the primary effects of Study 1 were replicated, with children showing higher ratings of the follower’s action and lower ratings of the violator’s action when the behavior prevented harm or protected others and children showing more similar ratings of the follower’s and violator’s actions when the behavior prevented inconvenience or protected the self.

Friendship quality

We constructed a linear mixed model to examine whether severity or framing predicted differences in children’s rating of how good or bad of friends they viewed the violator and the follower, using the difference score for this measure as the continuous outcome variable (Fig. 3C). There was a significant difference between the full model and the null model, χ ²(5) = 33.44, p < .001. As in Study 1, we found main effects of severity, χ ²(2) = 24.74, p < .001, and framing, χ ²(2) = 9.63, p = .008. In both the high-severity and other-oriented framing conditions, children showed higher positive difference scores, rating the follower as a better friend and the violator as a worse friend, whereas in the low-severity and self-oriented conditions, children had difference scores closer to zero, reflecting more even evaluations of the follower’s and violator’s friendship quality. Unlike in Study 1, there was no effect of gender. There were also no effects of age or an interaction between framing and severity. Thus, the primary effects of Study 1 were replicated, with children rating the followers as better friends and the violators as worse friends when the protective behaviors prevented greater harm over minor harm and when it protected others over the self. The same pattern can be seen in both the difference score measures, suggesting that children judged the violator and follower aliens similarly on multiple sociomoral dimensions.

Friend choice

The binary measure of which alien is a better friend mirrored the effects found using the difference score measure (Fig. 2D). After constructing a generalized linear mixed model, the full model significantly differed from the null model, χ ²(5) = 53.52, p < .001. As in Study 1, there was a main effect of severity, χ ²(2) = 41.12, p < .001, and framing, χ ²(2) = 7.84, p = .020, with stronger evidence for framing than in Study 1 and an overall similar pattern of results. An overwhelming majority of children chose the follower as the better friend in both the high-severity and other-oriented conditions, whereas just over half of the children chose the follower in the low-severity and self-oriented conditions. There was also a marginal main effect of gender, χ ²(1) = 3.78, p = .052): as in Study 1, girls preferred the follower overall. One effect that emerged in Study 2a that was not seen in Study 1 was an effect of age: in this sample, younger children showed a preference for the follower across conditions, whereas older children showed a preference for the violator, χ ²(1) = 4.39, p = .036. This may be explained by older children being more willing to choose the violator as a better friend in the low-severity and other-directed conditions than younger children, who may be less willing to choose the violator as a better friend regardless of context. Overall, we see the main effects from Study 1 as replicated and see stronger evidence that framing protective behaviors as other-oriented yields higher rates of children choosing the follower as the better friend.

In summary, all the primary effects found in Study 1 were replicated in our larger sample, suggesting that those effects are indeed robust and not Type I errors. We demonstrated even stronger evidence that children readily endorse protective behaviors as long as they prevent harm but do not differentially endorse protective behaviors when they are framed as self- versus other-oriented. We find that framing protective behaviors as other-oriented produces more moral thinking, with children showing more positive sociomoral evaluations of followers over those of violators. For a visual summary of the results found across Study 1 and Study 2a, see Table 2.

Study 2b

The aims of Study 2b were to further investigate how other-oriented framing induces sociomoral reasoning. We hypothesized that children would rate violators less positively and as worse friends when violations were committed intentionally in the high-severity condition; violators who intentionally failed to protect others would be judged as more wrong than those who intended to protect others but forgot, especially when there was a potential for harm. Studying 5- to 10-year-olds enabled us to investigate possible changes in children’s moral judgments, so we included age as an exploratory variable to further investigate. Because intention-based reasoning is a critical aspect of children’s developing moral judgment (Killen & Smetana, 2015), our central focus was whether intentionality and the severity of harm would influence children’s moral evaluations of those who do not follow other-oriented framed protective measures.

Method

Participants

Study 2b was run directly after Study 2a with the same children. One child dropped out of Study 2a early and so did not complete Study 2b; thus, the final sample for Study 2b was 60. There were equal numbers of children across three age groups (5–6, 7–8, and 9–10 years; i.e., 20 participants per age group).

Stimuli

Study 2b used the same type of stimuli as Study 2a and builds off of the stories presented to children. One of the four types of aliens (i.e., Furpees, Blickets, Zibs, and Ollers) corresponded to each severity condition (i.e., one high-severity alien and one low-severity alien). An example would be that the Furpees with their spikes would poke and hurt others (high severity). Within each alien type, participants saw two different colored aliens representing the intention conditions. Thus, the participants could see a green Furpee who accidentally forgot to do the target behavior (accidental) and a yellow Furpee who decided not to do the target behavior on purpose (intentional).

Design

We employed a 2 × 2 design with the variables severity (high vs. low) and intention (intentional vs. accidental) tested within participants. Therefore, each child saw four different aliens representing the four different combinations of manipulations (high-severity/intentional, high-severity/accidental, low-severity/intentional, and low-severity/accidental). We counterbalanced the order and combinations of the conditions and the aliens to create 16 unique sequences that were randomly assigned to children within each of three age groups (5–6, 7–8, and 9–10 years) created for counterbalancing purposes.

Procedure

Participants from Study 2a then immediately participated in Study 2b. We re-introduced children to a specific alien type (e.g., Furpees with spikes) that was assigned to either the high-severity or low-severity condition in Study 2a. Two comprehension check questions were included to ensure that the children remembered the severity manipulation and the discomfort of the behavior from Study 2a. Children were then told a story about either the intentional or accidental alien. Following the introduction of the alien, we asked a manipulation check question to assess whether children understood the intention. We employed the same method of evaluating failed comprehension or manipulation checks as in Study 1. Of the 300 trials (5 per child, 60 children), 12 trials (4%) were excluded due to a failed comprehension or manipulation check. Following each vignette, we asked several questions about each vignette (see supplementary material for a walk-through of the vignettes and images shown to children). Once children finished answering questions for all four aliens, the study concluded and participants’ parents were e-mailed a debriefing letter and the Amazon gift code.

Dependent measures

To measure children’s moral judgments of the violator, following the story of the accidental or intentional alien, children evaluated the morality of the intentional and accidental aliens’ actions (morality rating) on a 5-point Likert scale (really right, a little right, just OK, a little wrong, or really wrong). Then, to measure how children perceived the violator as a friend, children indicated for both the intentional and accidental aliens how good or bad of friends the aliens are (friendship quality) on a 4-point Likert scale (very good, a little good, a little bad, or very bad). Following the friendship rating, children were asked to choose, between the accidental and intentional aliens, which they rather preferred as a friend (friend choice).

Analysis

Our analysis approach was similar to that in Study 1. We used generalized linear mixed models for the binary dependent variable (friend choice) and linear mixed effects models for the two continuous dependent variables (morality rating and friendship quality). For each model, the fixed effects were trial, gender, age, intention, severity, and the interaction between intention and severity. Participant ID was included as a random effect to account for within-participants testing. Age and gender were preregistered as exploratory analyses (Study2b: https://aspredicted.org/NVJ_7XN). The regression table for the final models can be found in the supplementary material.

Results

Morality rating

We built a linear mixed model to examine how children morally evaluate the violator’s action. There was a significant difference between the full model and the null model, χ ²(5) = 82.61, p < .001. There was a main effect of both severity, χ ²(2) = 16.89, p < .001, and intention, χ ²(2) = 70.45, p < .001. Children rated intentional violators and high-severity violators as more morally wrong. There was a significant effect of the intention and severity interaction, χ ²(1) = 9.42, p = .002. In high-harm situations, children rated intentional norm violators as more wrong (Fig. 4 A). There was a significant interaction between age and severity, χ ²(1) = 4.42, p = .035. In the high-severity condition, older children tended to rate all transgressors more neutrally compared with younger children. In the low-severity condition, across all ages children tended to rate all transgressors as a little bit wrong. There was no effect of gender. However, there was a marginal main effect of age, χ ²(1) = 3.59, p = .058, with a trend of younger children rating all transgressors as slightly worse compared with older children (Fig. 4B). These differences were most apparent in the high-severity and intentional conditions, but among the rest of the conditions children across all ages tended to rate all violators’ actions as more neutral. Overall, both intention and severity appear to be important for children’s rating of the alien’s actions as good or bad.

Fig. 4.

Charts of model-predicted values for each of the Study 2b dependent variables. Panel A depicts the effects of intention and severity on morality rating, B depicts the interaction between intention and age on morality rating, C depicts the effects of intention and severity on friend rating, and D depicts children's tendency to choose the accidental alien based on seveity. Error bars represent 95% confidence intervals.

Friendship quality

We constructed a linear mixed model to examine whether severity or intention predicted differences in children’s rating of how good or bad of friends they viewed the violators. There was a significant difference between the full model and the null model, χ ²(5) = 60.73, p < .001. We found main effects of severity, χ ²(2) = 11.36, p = .003, and intention, χ ²(2) = 54.98, p < .001. In both the high severity and intentional conditions, children rated the violators as worse friends (Fig. 4C). There was a marginal interaction effect of severity and intention, χ ²(1) = 3.72, p = .054. Children tended to rate intentional violators as worse friends than accidental violators, especially when the behavior prevented severe harm. There was no significant effect of the children’s age or gender on their friendship ratings.

Friend choice

Overall, children were more likely to choose the accidental violator (M = .86, SD = .35) as a friend, t(107) = 10.80, p < .001. We created a generalized linear mixed model to determine whether severity affected children’s choice of the accidental or intentional friend. Results showed a significant difference between the full model and the null model, χ ²(3) = 10.86, p = .013. There was a main effect of severity, such that children were less likely to choose the intentional violator as a friend, χ ²(1) = 7.78, p = .005, when the behavior prevented severe harm to others (Fig. 4D). There was no significant effect of the children’s age or gender on their choice of friend.

A visual summary of the effects found in Study 2 can be found in Table 3 .

Table 3.

Summary of results from Study 2b broken down by dependent variable and listing all significant effects

Dependent variable	Study 2b
Morality rating	• Intention
	• Severity
	• Intention × Severity
	• Age × Severity
Friendship quality rating	• Intention
	• Severity
Friend choice	• Intention
	• Severity

Discussion

The aim of Study 2b was to examine how intention and severity of harm affect children’s sociomoral evaluations of those who do not follow novel protective measures. Overall, children tended to rate the violators neutrally and saw them as at least good friends regardless of their behaviors. However, children rated the intentional violator’s actions more negatively than accidental transgressions when the behavior could have prevented severe harm. In terms of children’s evaluations of the violators as friends, children rated violators in low-severity and accidental conditions as better friends. Altogether, children tended to evaluate most violators as good friends. Because there was an apparent ceiling effect for the friend rating question, this rating scale contained limitations in finding potential differences between variables because of the lack of variance. In the forced-choice friend question, children in high-severity situations preferred the accidental aliens over intentional aliens as friends. Similar to prior work on moral transgressions, children appear to be aware of the circumstances surrounding a transgression (i.e., the intention of the actions, the severity of harm caused) (Grueneich, 1982).

When analyzing children’s evaluations, children rated intentional violators and high-severity violators as more morally wrong and worse friends. Thus, children as young as 5 years are attending to intention in their moral judgments. Recent research suggests that 4-year-old children considered intention in their evaluations but less so than older children’s judgments that closely resembled adults (Nobes et al., 2017). Although, our paradigm required verbal communication from the children to communicate their evaluations, there has been an increasingly number of studies researching intention using nonverbal responses, suggesting that children younger than 4 years could be sensitive to intention in their evaluations of others (Behne et al., 2005, Hamlin, 2013, Kuhlmeier et al., 2003). When manipulating the simplicity of the task, Margoni and Surian (2020) found that 3-year-olds were successful in expressing intent-based judgment. Children’s sensitivity to intention on tasks can be more poignant on tasks with reduced processing demands. Considering this evidence, the trend of children’s use of intention in their judgments may occur earlier than previously studied but could be limited by the cognitive ability needed to complete morality tasks.

Children’s judgments may be mediated by how harmful the moral transgression of spreading illnesses is believed to be. Accordingly, our evidence illustrates that when there was no harm caused, children were more likely to judge both the accidental and intentional violators’ actions as okay. For this reason, how harmful children view an illness could affect how severely they evaluate those who do not follow preventative health measures. Our finding that intention and severity affect children’s sociomoral evaluations has implications for children’s judgments and their adherence to public health rules. When wearing protective measures is framed as preventing harm to others, children prefer being friends with someone who at least intends to follow protective measures and tries to protect others from harm. Therefore, framing public health measures in terms of being a good friend and protecting others could be an effective way to induce children’s adherence to public health measures.

In summary, children consider both the amount of harm caused and the intent of the action when they are asked to evaluate transgressors morally and who they would choose as friends. Children see those who refuse high-severity protective measures on purpose as more morally wrong and as worse friends.

General discussion

Our findings suggest that framing novel protective behaviors as self- versus other-oriented affects children’s moral reasoning but does not affect children’s explicit endorsement of these behaviors. Namely, when novel protective behaviors are framed as other-oriented, children show evidence of moral reasoning by (a) rating violators more negatively and followers more positively on sociomoral measures and (b) showing sensitivity to the intention of violators. Therefore, these results show that focusing on the prosocial aspects of an act helps to activate children’s moral reasoning. Other-oriented frames tend to emphasize the harm done to others, and as such children may adopt a moral lens when situations are framed in such a way.

In addition to the effect of framing, the severity of harm that the protective behavior would prevent was important to children, having a large effect on all our variables. When protective behaviors prevent more severe harm, children readily endorse them and rate violators more negatively and followers more positively. This is not the case when the behaviors prevent only a minor inconvenience.

Our findings on severity and framing were robust, remaining consistent across the two studies. In addition, Study 2b expanded these findings by demonstrating that in the other-oriented condition children reason about intentionality in morality-based ways. This provides further evidence that other-oriented frames induce moral thinking and that this moral thinking looks similar to other instances of moral reasoning that have been studied in the past.

To our knowledge, this is the first set of studies to explore the effect of framing on children’s reasoning about novel public health measures. Our results extend the findings from the adult literature. Similar to Luttrell and Petty (2021) with adults, we found other-oriented frames to be more impactful than self-oriented frames when it comes to children’s moralization of novel public health measures. This may suggest that the mechanism leading to this moralization in adults may already be present during middle childhood.

Although some of the adult literature suggests that framing protective measures as other-oriented is influential in decisions about these measures (e.g., Ceylan and Hayran, 2021, Jordan et al., 2021), we did not find an effect of framing on endorsement. One possible explanation is that the harm was less ambiguous in our study than in the adult studies directly addressing COVID-19 prevention measures. In our study the level of severity was manipulated and clearly communicated, whereas in the COVID-19 studies participants may have interpreted the severity of the pandemic differently. In our high-severity condition, children showed near ceiling levels of endorsement, so it may be that when severe harm is apparent this overrides the effect of framing. This hypothesis is further supported by the fact that Jordan et al. (2020) did not find this framing effect in their studies taking place later in the pandemic, when participants likely had a better understanding of the severity of the pandemic.

Our studies have implications for the literature on children’s morality as well as for public policy. First, our findings suggest that manipulating framing to focus on the effects of an action for others can induce moral reasoning in children. This indicates context specificity in children’s moral thinking, such that a simple manipulation of focus can affect whether moral thinking is activated. Moreover, children’s sensitivity to intentionality—well established in other domains of moral research—extends to our novel public health context.

Although our study did not directly ask participants about COVID-19 protective measures, our results suggest some potential implications for effectively communicating novel protective measures to children. First, emphasizing the severity of the prevented harm is crucial. Children both endorse and moralize novel public health measures when the measures prevent harm rather than inconvenience. Emphasizing the consequences for others, rather than the self, also has moralizing effects. Thinking of the measures in moral terms may lead children to value compliance with said measures more; therefore, emphasizing the severity of harm as well as the effects for others is a promising way to communicate novel protective measures to children.

Our study had a few limitations that could be addressed with future research. First, we chose to use hypothetical rather than real-world situations because we wanted to get an unbiased look at children’s evaluations that did not depend on their preexisting ideas about COVID-19 measures. This decision allowed us to examine children’s reasoning under ideal and controlled circumstances, but it limits our ability to predict how children might react in real-world contexts such as the COVID-19 pandemic. In addition, our sample demographics were not representative of the larger U.S. context. Children’s demographics, such as their parents’ political ideation and the voting behavior of the county where they live, affect their views on COVID-19 protective measures (Gollwitzer et al., 2022), so it is possible that children from different backgrounds might respond differently in our paradigm. Our data suggest that there may also be some gender-related differences, although more research is needed on that front given that most of our gender differences did not replicate from Study 1 to Study 2a. Finally, we did not power our studies specifically to look for age effects, so it is possible that we were underpowered to detect more complex interactions with age.

The findings of our study offer opportunities for future research. First, future studies could assess how children reason about other types of harm, such as psychological harm and disease transmission, that are more causally opaque than physical harm. Prior research shows that harming others is a particularly salient moral violation for children (e.g., Smetana & Ball, 2019). We chose to use physical harm in our paradigm so that we could examine whether an effect would be present in obvious instances of harm. However, paradigms in which the harm is vaguer would map more closely onto real-world contexts such as the COVID-19 pandemic.

In addition, future research can recruit more diverse samples and examine whether there are any effects of political or demographic background on these hypothetical situations. By doing so, researchers could address whether any observed effects based on sociodemographic factors are due to explicit communication about COVID-19 protective measures or due to differences in reasoning based on ideology or context.

Finally, whereas our study focused on one type of framing as self- versus other-oriented, future research could examine how other types of framing could affect these behaviors. A well-known area of framing research is gain–loss framing such as the Tversky and Kahneman (1981) study in which participants saw language describing the number of people who were saved versus people who died. A recent study employed gain–loss framing, along with additional contextual manipulations, on adults’ willingness to take the COVID-19 vaccine (Huang & Liu, 2022). Future research could examine the effects of gain–loss framing on children’s COVID-19 protective behaviors as well such as by emphasizing the risk of not performing the behaviors versus the protection offered by the behaviors.

Our studies are the first to examine children’s evaluations of novel protective behaviors similar to those introduced by the COVID-19 pandemic. Our findings expand on the adult literature looking at other- and self-oriented frames for novel protective behaviors by demonstrating that children consider these frames in their sociomoral evaluations but not in their endorsements of the behaviors. Specifically, children reason in moral terms when novel protective behaviors are framed as other-oriented rather than self-oriented. We also found that children show higher levels of endorsement in addition to moral thinking when the severity of harm prevented by novel public health measures is high. Overall, the findings from the current set of studies help to shed some light on how children reason about novel public health measures, which can help to inform communications with children should novel public health measures be introduced again in the future.

Appendix A. Supplementary material

The following are the Supplementary data to this article:

Supplementary Data 1

Complete stimuli and scripts as presented to the child as well as regression tables for all studies

Data availability

Data will be made available on request.