A virtues approach to children’s kindness schemas

ABSTRACT

Kindness is key to flourishing school communities. A social-cognitive approach to virtue emphasizes the importance of having an elaborate set of accessible mental representations (i.e., schemas) for expressing kindness. We employed a multi-informant, mixed method, longitudinal design across 6 months that focused on 4^th and 5^th graders’ (N = 320) kindness schemas using the open-ended question, ‘What are some ways you can show kindness to others?’ Results indicated that children’s schemas entailed wide-ranging content, expressing virtues of generosity, compassion, inclusion, civility, and harm avoidance. The breadth of children’s schema repertoires was positively associated with peer (but not teacher) ratings of their kindness, and virtues that attend to others’ vulnerability (compassion, inclusion) were the most indicative of children’s kindness from peers’ perspectives. Further, the breadth of kindness repertoires was associated with aspects of classroom ecology (e.g., peer acceptance), suggesting that positive classroom relationships may serve as sites for the cultivation of kindness schemas.

KEYWORDS:

• Kindness
• virtue
• development
• education
• classroom ecology
• middle childhood
• mixed method

Schools are organized communities whose flourishing is built upon norms of virtuous behavior – where teachers, administrators, and children are respected and valued members of the community (Sauve & Schonert-Reichl, 2019). The fields of education (e.g., Warren & Narvaez, 2020), positive psychology (e.g., Peterson & Seligman, 2004), and philosophy (e.g., Kristjánsson, 2015) share the related interest in the cultivation of children’s virtue. Wright et al.’s (2020) interdisciplinary work describes virtues as appropriately motivated, robust traits of character that are consistently enacted across a wide range of situations that call for their expression. Virtues are often sorted into moral virtues that involve treating others well, intellectual virtues directed toward knowledge and wisdom, and enabling virtues that help people enact the other virtues and achieve valued ends (Kristjánsson, 2015; see also, McGrath et al., 2018). This study focuses on a subset of moral virtues that involve benevolence, as they play important roles in promoting positive school cultures and often serve as targets of development through character or virtue education (Berkowitz & Bier, 2007; Kristjánsson, 2015), social and emotional learning – SEL (Layous et al., 2012), positive youth development (Lerner et al., 2015), and mindfulness-based interventions (Flook et al., 2015). Moving forward, we refer to moral virtues focused on benevolence as ‘kindness virtues,’ because kindness is a relatable term within school contexts. The present study explores which virtues are central to children’s conceptions of kindness and examines how mental representations of kindness develop during middle childhood.

A Differentiated Virtues Approach

The current study adopts the premise that kindness reflects far more than ‘being nice’ and instead encompasses several specific virtues, such as compassion, generosity, and gratitude. A virtue approach to morality emphasizes that different situations ‘call for’ the enactment of certain virtues (Fowers et al., 2021; Wright et al., 2020). A child might observe their classmate struggling with math and compassionately help, or they might consider the crossing guard’s concern for their safety and express gratitude while crossing the street. The study of specific virtues elucidates various kindness virtues that differentially matter across various sorts of social interactions in children’s lives. As such, a differentiated virtues approach supplies critical content for moral education (Hamm, 1977), and the terminology of individual virtues offers relatable language and accessible action guidance for schools, teachers, and children.

Kindness Virtues and Salience for Children

Merging psychology and philosophy, Gulliford and Roberts (2018) describe an ‘allocentric quintet’ of virtues rooted in ‘intelligent caring about people’ (p. 223). They describe these kindness virtues as consisting of generosity, compassion, gratitude, forgiveness, and humility. These virtues express kindness in distinct ways and toward particular sorts of targets: ‘Generosity is benevolence expressed in giving …, compassion is benevolence toward sufferers, gratitude is benevolence toward benefactors, forgiveness is benevolence toward offenders’ (p. 217), and humility, while not a form of benevolence per-se, fuels these virtues because each one entails an ‘other’-orientation that is made possible by diminishing the vices of pride (e.g., arrogance, domination).

Although the quintet framework is useful, its relevance for children is unknown. Wright and colleagues (2020) propose a developmental contextual account of virtue development, arguing that virtues are cultivated in relation to the practical situations that children encounter in their everyday lives. To the extent that children routinely encounter situations that call for additional kindness virtues beyond the quintet (e.g., inclusion), or if they rarely encounter situations that call for virtues among the quintet (e.g., humility), then the framework would have limited utility for understanding children’s kindness. Humility, in particular, may be disconnected from children’s kindness because the situations they encounter may be more likely to highlight concrete behaviors (what one should do) rather than the more abstract manner in which they should enact kindness (e.g., without arrogance or domination).

Several streams of scholarship provide insight into the kindness virtues that are relevant for children. Literature on prosocial behavior has examined a wide range of ‘other-benefitting’ constructs, including empathy, compassion, and gratitude but also cooperation and mutual coordination (civility) and the avoidance of harm to others (Keltner et al., 2014; Ladd & Profilet, 1996; Weir & Duveen, 1981). Similarly, kindness interventions focus on benevolent concepts such as respect (civility), inclusiveness, caring, sharing (Flook et al., 2015; Schonert-Reichl & Whitehead Arruda, 2016), or simply, kindness, however children understand the term (Binfet & Passmore, 2019; Layous et al., 2012). Thus, in addition to Gulliford and Roberts’ (2018) quintet, virtues of civility, inclusiveness, and harm avoidance may be highly relevant to children.

Arguably the best way to understand kindness virtues during middle childhood is to directly ask children, which would reveal what counts as kindness based on the practicalities of their everyday lives (Wright et al., 2020). Recent qualitative research with 4^th – 8^th graders has found the most prevalent themes were helping others, encouraging others, and showing respect (Binfet & Passmore, 2019) – with similar themes arising in other age groups (Binfet, 2020; Binfet & Gaertner, 2015) – behaviors that map onto the virtues of generosity and compassion identified in the allocentric quintet, but also highlighting civility (i.e., respect). Notably, none of these studies organized and labeled children’s responses as virtues representing core educable principles. The current study expands this area of research by using an open-ended response format to identify ways in which 4^th and 5^th grade students conceptualize kindness, and coding core virtues that emerge.

Mental Representations of Kindness: Fueling Virtue from Within

Kind behaviors rely upon children’s internal cognitive processing of social situations. We adopt a social-cognitive approach to virtue (Wright et al., 2020), which holds that virtuous responses are produced by a set of virtue-supportive social-cognitive mechanisms including schemas (organized knowledge structures), beliefs, and other mental representations that are activated by and process the current situation. Mental representations filter perceptions of the social environment (Lapsley & Narvaez, 2004) and serve as the ‘cognitive carriers of dispositions’ (Cantor, 1990, p. 737). For example, a child who has a well-developed schema of kindness has a general sense of what kindness means; has seen kindness modeled in various ways by parents, teachers, and peers and therefore it comes to mind often; and can generate different kindness responses that fit various situations. Their elaborate network of mental representations about kindness helps them ‘see’ the kindness-relevance of daily situations, and supports a range of kind behaviors, whether by offering emotional comfort (compassion), playing with the new student (inclusion), or thanking a classmate for their help (gratitude).

As such, mental representations are critical ingredients that serve as the intermediates between environmental inputs and virtuous responses (Cantor, 1990; Wright et al., 2020). Indeed, mental representations are linked to real-world behavior within the moral domain (Farrell, 2013; Hertz & Krettenauer, 2016). For example, a meta-analysis found moral self-schemas are positively associated with moral behavior (Hertz & Krettenauer, 2016). Moreover, focusing on mental representations is important from an educational perspective because they develop through learning processes (e.g., Matsuba et al., 2011). Despite their importance, little is known about children’s mental representations of kindness, and more specifically, the breadth of these schemas.

Development of Kindness Schema Repertoires in the Classroom Ecology

Due to our interest not only in the construct of kindness itself, but in the development of kindness schemas at school, we considered whether the social ecology of the classroom – children’s relationships with peers and their teacher – supports positive changes in children’s representations of kindness. The role of children’s status within the classroom peer context was of particular interest as interactions with peers become increasingly important in middle childhood (DelGiudice, 2018). A contemporary approach to peer relations research has characterized peer relationships as developmental contexts (Bukowski et al., 2011) where the peer group and a child’s status in the group work together to influence development by providing opportunities for children to learn and practice social skills (Wentzel, 2014). Much attention has focused on ‘accepted’ children, or those who are well-liked by their classmates. Because accepted children generally receive many social initiations from peers, they experience plenty of opportunities to practice various forms of kindness. Further, having this reputation likely encourages students to meet this expectation in their interactions, such that the peer group not only presents opportunities to engage in kind behaviors, but also reinforces such behavior. Indeed, positive peer relations in the early grades are associated with the development of later social competence, academic achievement (Johnson et al., 2000; Neil et al., 1997), and even relationship quality in adulthood (Marion et al., 2013). Thus, we consider whether peer acceptance supports the development of benevolent schemas across the school year, as the variety of opportunities that accepted children are presented with which may be met with kindness likely expands their repertoire of kindness schemas.

We also explore whether children’s relationship with their teacher may set the stage for the development of kindness schemas. Research born from attachment theory (Verschueren, 2015), social learning theory (Bandura, 1969) and developmental systems perspectives (Lerner et al., 2015) posit that more supportive and less conflictual relationships with teachers contribute to positive student development, either through establishing a secure base, explicit modeling and reinforcement, or through setting positive expectations for classroom behavior. Supportive relationships with teachers are known to be associated with children’s social and academic skills (Birch & Ladd, 1997; Engels et al., 2016), and teacher-child conflict is associated with reductions in cooperative participation (Birch & Ladd, 1997) and prosocial behavior (Roorda et al., 2014). Additionally, teacher-student relationships become increasingly critical as students enter adolescence (Ruzek et al., 2016). It is possible that such effects on children’s social, behavioral, and academic outcomes manifest through developments in children’s social cognition resulting from strong relationships with teachers. This is the first study that we know of to examine whether children’s relationships with peers and teachers are associated with the development of their kindness schemas across the school year.

The Current Study

The current study expands our understanding of kindness in middle childhood by testing four hypotheses. First, using an open-ended response format, we examined which virtues are most prototypical of (central to) 4^th and 5^th graders’ kindness schemas. The prototypicality hypothesis held that, among Gulliford and Roberts’ (2018) allocentric quintet, the virtues of generosity, compassion, gratitude, and forgiveness will occur most frequently in children’s responses, whereas humility will rarely occur. We also expected to see evidence of additional virtues beyond the quintet, such as civility and inclusion. Second, because virtue development is theorized to occur in relation to the practical situations that children encounter in everyday life (Wright et al., 2020), we posed the visibility hypothesis, which states that children whose kindness schemas incorporate virtues that are most prototypical of kindness would be recognized as kind by peers and teachers. Third, drawing from social-cognitive approaches to virtue (Lapsley & Narvaez, 2004; Wright et al., 2020), we investigated the elaborateness of children’s kindness schema repertoires. The schema repertoire hypothesis posited that the number of distinct kindness schemas in a child’s response would be associated with more positive peer and teacher evaluations of their kindness. Finally, grounded in the developmental contextual view that kindness develops through affordances in children’s social environment, the social development hypothesis posited that peer acceptance and teacher-child closeness would be prospectively associated with expanded kindness schema repertoires, whereas teacher-child conflict would be associated with narrowed kindness repertoires.

Materials and Methods

Study materials and a supplement containing a coding manual, expanded analytic plan, and supplementary tables and figures are available in an online repository (https://osf.io/ubfzk/?view_only=71ef5a1243664330958d05c9e3c9f420). Note that data are not publicly available, as parents did not consent to having their children’s de-identified data shared.

Study Design and Recruitment

Data were drawn from a randomized controlled trial of an SEL program – Kindness in the Classroom. The study was described to principals of 17 schools from two public school districts in British Columbia, Canada. Principals shared information about the study with the 4^th and 5^th grade teachers in their schools. Informed consent was collected from teachers (for their own participation) and from parents/guardians of students; students provided informed assent. Study design and recruitment are further described elsewhere (Braun et al., 2020). This study was approved by the Behavioural Research Ethics Board (BREB) at the University of British Columbia – Vancouver Campus (approval #H14-02370).

Participants

Because the current study focused on a descriptive assessment of children’s understanding of kindness and the natural development of kindness, only teachers and students in the control condition were included. Participants included 15 4th and 5th grade teachers (M = 16.42 years of teaching experience, SD = 6.58; 86% female; 67% white; 80% with a post baccalaureate diploma or graduate degree; 20% 4th grade, 13% 5th grade, and the remaining taught a combined class of 4^th and 5^th grade students). Teachers from the two districts did not differ in gender or ethniccomposition, years of teaching, nor reports of teacher-child conflict, closeness, kindness, or the number of kindness responses students produced. However, districts differed on peer acceptance. Except for two teachers who taught at the same school, all teachers taught at different schools.

Class sizes ranged from 23 to 29 students, M = 27, SD = 2. Participation rates within each classroom were: M_T1 = 78%, Range_T1 = 48%-96%; M_T2 = 76%, Range_T2 = 36%-93%; M_T3 = 75%, Range_T3 = 40%-93%. Seventeen students were omitted from analyses because they reported that reading English was ‘hard’ or ‘very hard’ for them, calling into question the validity of their responses. In total, 320 students (M = 9.93 years-old at the start of the study, SD = 0.57; 48% female; 83% English as their first language; 46% 4th grade) were included in analyses.

Procedure

Data were collected from teachers and students at three time points across a single school year: December (Time 1), March (Time 2), and May/June (Time 3). The school year in Canada begins the first week of September and ends at the end of June. At Time 1, teacher-rated kindness and peer-nominated kindness were collected; at Times 1–3, teachers and students completed surveys measuring breadth of kindness schema repertoires, teacher-child closeness and conflict, and peer acceptance. Trained research assistants masked to study condition administered the survey to students by reading each item aloud to control for differences in reading ability.

Measures

Kindness

Kindness was assessed in several ways, incorporating children’s free responses, peer nominations, and teacher-ratings.

Ways to be Kind. Children provided open-ended written responses to the question, ‘What are some ways you can show kindness to others? List as many as you can think of in the space below.’ This item was written for the current study. The breadth of each child’s kindness schema repertoire was calculated in accordance with the Kindness Virtues Coding Scheme detailed in the Analytic Plan and reflected the number of distinct ways the child identified to be kind. In addition, the presence of each virtue (among eight virtue categories) in children’s lists was coded (0 = virtue absent, 1 = virtue present).

Peer-Nominated Kindness. Peer-nominated kindness was measured using a 3-item, unlimited, cross-gender peer nomination procedure (e.g., Schonert-Reichl et al., 2012; Wentzel, 1993). Children were given a class roster and were instructed to circle the names of students who (1) ‘share and cooperate’, (2) ‘help other kids when they have a problem,’ and (3) ‘are kind.’ Children could circle as many names as they wanted. These items were selected from a larger set of social competency items (Schonert-Reichl et al., 2012), based on their conceptual alignment with kindness. For each item, students’ scores reflected the proportion of classmates who nominated them. The three items were averaged and exhibited good internal consistency (⍺ = .93).

Teacher-Rated Kindness. Teachers’ ratings of students’ kindness were gathered using a measure created for the current study that aligned with the peer nomination items. Teachers rated each student on three prompts using a 5-point scale (1 = definitely does not apply, 5 = definitely applies): ‘This student is a student who … ’ (1) ‘shares and cooperates,’ (2) ‘helps other kids when they have a problem,’ and (3) ‘is kind’. The three items were averaged and exhibited good internal consistency (⍺ = .84).

Classroom Ecology

Three aspects of students’ classroom ecologies were measured: peer acceptance, teacher-child closeness, and teacher-child conflict.

Peer Acceptance. Peer acceptance was measured using the same peer nomination procedure described above for peer-nominated kindness (e.g., Schonert-Reichl et al., 2012; Wentzel, 1993). Children were instructed to ‘circle the names of students who you would like to be in school activities with.’ For this single item, scores were computed in the same manner as peer-nominated kindness.

Teacher-Child Closeness. Teachers reported teacher-child closeness using the 8-item closeness subscale of the Student-Teacher Relationship Scale (Pianta, 2001). Teachers rated their relationships with each of their students using a 5-point scale (1 = definitely does not apply, 5 = definitely applies) to respond to the items (e.g., ‘I share an affectionate, warm relationship with this child.’). The eight items were averaged and exhibited adequate-to-good internal consistency (⍺_Time1 = .88; ⍺_Time2 = .74; ⍺_Time3 = .90).

Teacher-Child Conflict. Teachers reported teacher-child conflict using the 7-item conflict subscale of the Student-Teacher Relationship Scale (Pianta, 2001). Teachers rated their relationships with each of their students using a 5-point scale (1 = definitely does not apply, 5 = definitely applies) to respond to the items (e.g., ‘This child and I always seem to be struggling with each other.’). The seven items were averaged and exhibited good internal consistency (⍺_Time1 = .92; ⍺_Time2 = .92; ⍺_Time3 = .91).

Analytic Plan

Data from Time 1 were used to test the first three hypotheses (prototypicality, visibility, schema repertoire), since these required only cross-sectional data.¹ The longitudinal, social development hypothesis was tested with data from all three timepoints.

Kindness Virtues Coding Scheme

A coding scheme was developed for this study to code open-ended responses to the kindness prompt. We employed inductive thematic analysis (Clarke & Braun, 2017), a bottom-up process by which we examined responses from a random subset of 75 children, took inventory of the sorts of responses that emerged, and identified which virtue themes were expressed.

Eight high-level virtue themes emerged in our data. The first five, (1) Generosity, (2) Compassion, (3) Gratitude, (4) Inclusion, and (5) Civility represent different expressions of caring for others. Related to civility is the even more basic avoidance of unkindness, which was coded as (6) Harm Avoidance. Vague benevolent actions (e.g., be kind, be nice), infrequently mentioned benevolent qualities (e.g., forgiveness, loyalty), and generic benevolent actions (e.g., smile, hug) were assigned to the (7) Miscellaneous Forms of Kindness category. Responses that were not relevant to kindness were assigned to the (8) Not Kindness-Relevant category. High-level virtue themes were further divided into subcategories described in detail in the Kindness Virtues Coding Manual (see Supplementary Materials) and were used to ascertain non-redundant responses.

Each line below the prompt provided to children was considered a coding unit. Because most responses were brief (e.g., ‘play with others’), typically a single code was assigned to each coding unit. Coding units involving long or complex responses were assigned the two most relevant codes (e.g., ‘open the door for someone or play with someone new’ was coded both as Generosity and Inclusion). Notably, 10% of children provided no responses to the open-ended item. If these children completed at least one of the two previous survey items, they were included in the analysis and treated as listing zero ways to be kind (and no virtues present); those who did not complete the previous two survey items were treated as missing data.

Coding. After coders were trained and achieved adequate inter-rater reliability (see Supplementary Materials), a primary coder officially coded all three waves of responses. A secondary coder independently coded a random subset of 20% of the sample. Inter-rater reliability was computed based on the eight high-level virtue themes. Inter-rater reliability was high (Cohen’s κ = .92; exact agreement = 94%), indicating that the coding protocol was consistently applied across coders. The primary coder’s data were used in analyses.

Prototypicality Hypothesis

To test which kindness virtues emerged in children’s kindness schemas, non-redundant kindness-relevant responses reported at Time 1 were tallied. Prototypicality was operationalized in three ways: (1) percentage of children who mentioned each virtue category – computed by dividing the number of children who ever mentioned the virtue category, by the total number of children in the sample, (2) percentage of total responses for each virtue category – computed by dividing the total number of responses for each virtue category (for all children), by the total number of responses across all categories, and (3) priority scores, which indexed the average order in which the virtue category first emerged in children’s responses. See Supplementary Materials for further information.

Visibility Hypothesis

To test whether children whose kindness schema repertoires incorporate the most prototypical virtues were recognized as kind by peers and teachers, we examined associations between the eight binary variables that indexed the presence of each virtue, and each of peer-nominated kindness and teacher-rated kindness at Time 1. Given the hierarchical structure of the data (i.e., students nested within classrooms), multilevel modeling (MLM) was used, with students (Level 1) nested within classrooms (Level 2). Unconditional models found an intraclass correlation (ICC) of .29 for peer-nominated kindness and .16 for teacher-rated kindness, indicating substantial between-classroom variance (i.e., ICC > .05; Peugh, 2010) and the need for MLM to examine conditional models.

In the conditional models, eight binary predictors indexed the presence of each virtue, and gender, grade level, and district were included as covariates given their putative associations with kindness and verbal ability, and to account for idiosyncratic design features of the study. Covariates were binary (0 = boy, 1 = girl; 0 = 4^th grade, 1 = 5^th grade; 0 = District 1, 1 = District 2). See Supplementary Materials for models.

Schema Repertoire Hypothesis

To test whether the breadth of kindness schema repertoires were associated with peer-nominated kindness and teacher-rated kindness, associations between these constructs at Time 1 were examined with MLM. Responses coded as Not Kindness-Relevant (8) did not contribute to the tally of the breadth of children’s repertoires. MLM models paralleled models for the Visibility Hypothesis, but included the breadth of kindness schema repertoires as the main predictor of interest. Breadth of kindness schema repertoires was group-mean centered around the classroom mean. See Supplementary Materials for models.

Social Development Hypothesis

To test whether peer acceptance, teacher-child closeness, and teacher-child conflict were prospectively associated with breadth of kindness schema repertoires, random intercepts cross-lagged panel models (RI-CLPM; Hamaker et al., 2015) tested prospective reciprocal associations between breadth of repertoires and each of these social ecology constructs in three separate models (see Figure 1 and Supplementary Figures 1 and 2). See Supplementary Materials for information about model specifications and tests of competing longitudinal models (Supplementary Table 1).

Figure 1. Random intercepts cross-lagged panel model (N = 320) with breadth of kindness schema repertoire and peer acceptance. Unstandardized estimates are displayed. Cross-lagged and autoregressive paths were constrained equal over time, whereas within-wave covariances were freely estimated. Breadth = breadth of kindness schema repertoire; Acceptance = peer acceptance. *** p < .001.

Results

Descriptive statistics and bivariate correlations are displayed in Supplementary Table 2.

Prototypicality Hypothesis

Partial support was found for the hypothesis that Generosity, Compassion, Gratitude, and Forgiveness (and not Humility) would occur most frequently, and that additional virtues would also be evident. The most frequently mentioned responses to the prompt ‘What are some ways you can show kindness to others?’ exemplified Generosity (57% of children). Compassion (43%) and Miscellaneous forms of Kindness (40%) were mentioned by a large fraction of children, followed by Inclusion (31%) and Civility (24%). Far less frequently mentioned was Harm Avoidance (14%), responses that were Not Kindness-Relevant (9%), and a very small fraction of children offered examples of Gratitude (3%). Forgiveness was mentioned so infrequently that it did not have its own code. The same general pattern of findings was observed when prototypicality was indexed by the percentage of all responses (see, Table 1).

Table 1. Prototypicality of Virtues Exemplifying Kindness for 4^th and 5^th Graders.

	% of children	% of all responses	Priority
Generosity	57.09	27.87	1.62
Compassion	42.91	22.41	1.65
Misc Kind	39.86	16.48	2.25
Inclusion	31.42	12.05	2.20
Civility	23.99	9.32	2.54
Harm Avoidance	14.19	6.78	2.64
Not Kindness Rel	9.46	4.24	2.86
Gratitude	3.38	0.85	3.10

Note. Misc Kind = miscellaneous forms of kindness. Not Kindness Rel = not kindness relevant. Priority scores refer to the position in children’s lists, such that virtues with lower priority scores were listed earlier.

Prototypicality was also operationalized using priority scores, which reflected the order in which each virtue emerged in children’s lists. Results were similar to the frequency findings, with the noteworthy exception that Inclusion occurred slightly earlier than Miscellaneous forms of Kindness. The three prototypicality indices were highly correlated (|r| = .96 – .98), indicating strong convergence across operationalizations of prototypicality.²

Visibility Hypothesis

Next, we found partial support for the hypothesis that children with schemas incorporating the most prototypical virtues would be recognized by peers and teachers as kind. The presence of Compassion and Inclusion responses predicted higher peer-nominated kindness (B = .03, p = .028; B = .05, p = .002, respectively) but not teacher-rated kindness (see, Table 2). The remaining virtues were not uniquely linked to peer-nominated kindness or teacher-rated kindness.³

Table 2. Multilevel Regression Models Predicting Peer- and Teacher-Rated Kindness as a Function of Individual Virtues in Students’ Responses.

	Peer-Nominated Kindness			Teacher-Rated Kindness
	B	SE	p	B	SE	p
Intercept	.32	.03	< .001	4.11	.12	< .001
Demographics
Gender	.16	.01	< .001	.25	.08	.002
District	.07	.05	.208	−.32	.17	.083
Grade	.03	.02	.102	−.11	.10	.248
Kindness Virtues
Generosity	.02	.02	.217	.04	.08	.604
Compassion	.03	.01	.028	.10	.08	.216
Gratitude	−.02	.04	.567	.07	.23	.756
Inclusion	.05	.02	.002	.16	.09	.071
Civility	.02	.02	.195	.03	.10	.793
Harm Avoidance	−.01	.02	.617	.01	.12	.916
Misc Kind	−.01	.02	.491	−.08	.09	.401
Not Kindness Rel	−.02	.03	.429	−.03	.14	.829
Model Fit
AIC	−359.58			612.38
BIC	−308.06			663.71

Note. Bold values are statistically significant (p < .05). Unstandardized estimates and standard errors are displayed. Gender (0 = boy, 1 = girl), District (0 = District 1, 1 = District 2), Grade (0 = 4^th grade, 1 = 5^th grade). All kindness virtues are dichotomous (0 = virtue absent, 1 = virtue present). Misc Kind = miscellaneous forms of kindness; Not Kindness Rel = not kindness relevant. AIC = Akaike’s Information Criteria; BIC = Bayesian Information Criteria.

Schema Repertoire Hypothesis

On average, children listed about three ways of being kind (M = 2.82, SD = 1.98). The hypothesis that breadth of schema repertoire would be associated with higher other-reports of the child’s kindness was supported when other-reports came from peers (B = .01, p < .001) but not teachers (B = .03, p = .150; see, Table 3).⁴

Table 3. Multilevel Regression Models Predicting Peer-Nominated and Teacher-Rated Kindness as a Function of Breadth of Kindness Schema Repertoire.

	Peer-Nominated Kindness				Teacher-Rated Kindness
	B	SE	p		B	SE	p
Intercept	.35	.03	< .001		4.19	.11	< .001
Demographics
Gender	.16	.01	< .001		.26	.08	.001
District	.07	.05	.218		−.32	.17	.083
Grade	.04	.02	.025		−.08	.10	.381
Kindness Virtues
Breadth of Kindness Schema	.01	.00	< .001		.03	.02	.150
Model Fit
AIC	−365.87				602.97
BIC	−340.11				628.64

Note. Bold values are statistically significant (p < .05). Unstandardized estimates and standard errors are displayed. Gender (0 = boy, 1 = girl), District (0 = District 1, 1 = District 2), Grade (0 = 4^th grade, 1 = 5^th grade). Breadth of Kindness Schema = breadth of kindness schema repertoire (group-mean centered). AIC = Akaike’s Information Criteria; BIC = Bayesian Information Criteria.

Social Development Hypothesis

Prospective bidirectional associations between breadth of kindness schema repertoire and (a) teacher-child closeness, (b) teacher-child conflict, and (c) peer acceptance were examined. No support was found for the hypothesized prospective within-person associations, but between-person associations were observed.

Peer Acceptance and Breadth of Kindness Schema Repertoire

There were no cross-lagged or autoregressive within-person effects between peer acceptance and breadth of schema repertoire across time, indicating that deviations from one’s average level of each construct did not predict deviations from average on either construct at the subsequent timepoint (see, Figure 1). At the between-person level, peer acceptance was positively related to breadth of schema repertoire (r = .35, p < .001); children who had higher average peer acceptance scores across time tended to produce a higher average number of ways to be kind across the three timepoints.

Teacher-Child Closeness and Breadth of Kindness Schema Repertoire

For the teacher-child closeness model, there were no cross-lagged or autoregressive within-person effects (see Supplemental Figure 1). At the between-person level, teacher-child closeness was positively related to breadth of schema repertoire (r = .26, p < .001) such that children with higher average scores on teacher-child closeness across time produced a higher average number of ways to be kind across timepoints.

Teacher-Child Conflict and Breadth of Kindness Schema Repertoire

Regarding teacher-child conflict (see Supplemental Figure 2), there were no cross-lagged within-person effects, but there was an autoregressive within person effect for teacher-child conflict (β_T1-T2 = .58, p < .001; β_T2-T3 = .69, p< .001).⁵ Children who scored higher relative to their average level of conflict at a given timepoint tended to subsequently have higher teacher-child conflict (relative to their own averages) two months later. At the between-person level, children who averaged higher teacher-child conflict across time tended to list fewer ways to be kind across the three timepoints (r = −.42, p = .006).

Discussion

This study maps out the conceptual terrain of kindness schemas in the eyes of children in 4^th and 5^th grades, while testing aspects of a social-cognitive, developmental contextual account of virtue (Wright et al., 2020). In contrast to a recent philosophically derived conceptual analysis proposing an ‘allocentric quintet’ of kindness virtues (Gulliford & Roberts, 2018), we found that children described kind behaviors that both overlapped with the quintet model and deviated in developmentally appropriate ways. We also found that those children who were viewed as kind by their peers produced more ways to be kind in their open-ended responses, and specifically listed compassionate and inclusive responses, suggesting that there is value both in the quantity and quality of children’s kindness repertoires. Further, although there were no effects of children’s social ecologies on changes in their kindness repertoires, evidence indicated that children with more robust kindness repertoires were – across the entire study period – more accepted by their peers and had closer and less conflictual relationships with their teachers. These findings advance understanding of 4^th and 5^th graders’ kindness schemas and highlight the value of conceptualizing kindness as a broad virtue comprised of several narrow virtues (e.g., generosity, compassion, inclusion).

Prototypicality: Rethinking the Allocentric Quintet for Children

We posited that Gulliford and Roberts’ (2018) allocentric quintet (generosity, compassion, gratitude, forgiveness, and humility) might be a developmentally poor fit for children. As expected in the prototypicality hypothesis, generosity and compassion from the quintet were among the most frequently expressed by 4^th and 5^th graders, whereas references to humility were not evident in children’s conceptualizations of how to be kind. Humility involves the manner in which behaviors are enacted (not what concrete behaviors one enacts), implicating a level of abstract thinking that may be inaccessible to 4^th and 5^th grade children. Moreover, concerns that normatively spike during adolescence, such as social status and accurately reading and responding to social cues from peers (Burnett et al., 2011; Yeager et al., 2018) – conditions in which humility could matter strongly for the formation and maintenance of smooth relations (e.g., Allgaier et al., 2015) – may not yet be present in 4^th and 5^th grade children.

Counter to our hypothesis, neither gratitude nor forgiveness featured strongly in children’s responses, suggesting that these virtues are not central to kindness schemas among 4^th and 5^th graders. Perhaps children’s untutored conceptions of gratitude and forgiveness are more tightly tied to the domain of social conventions (e.g., to say ‘thank you’; to accept others’ apologies) rather than to the domain of morality where considerations of others’ welfare reside (e.g., Smetana et al., 2014), resulting in a dearth of these two virtues in children’s conceptualizations of kindness at this stage in development. Alternatively, forgiveness in particular may be implicit in notions of other virtues such as compassion and inclusion, without being explicitly identified as an expression of kindness.

As expected, we found strong evidence of additional virtues beyond the quintet – namely inclusion, civility, and harm avoidance – aligning with themes covered in age-appropriate kindness education programs (e.g., Flook et al., 2015; Schonert-Reichl & Whitehead Arruda, 2016) and with scholarship on prosocial behavior (e.g., Keltner et al., 2014; Ladd & Profilet, 1996). Children of this age clearly conceptualize kindness in multifaceted meaningful ways that extend far beyond simplistic notions of ‘being nice.’

With regard to kindness education, the most prototypical kindness virtues identified in this study (generosity, compassion, inclusion) may be solid entry points for education, as these virtues are shown here as particularly relatable and widely agreed-upon expressions of kindness for 4^th and 5^th graders. In contrast, the least prototypical kindness virtues (harm avoidance, gratitude, and perhaps forgiveness⁶) may reflect areas for growth and require additional scaffolding and support to expand children’s notions of kindness. Given that children listed just under three ways to be kind to others, educational efforts could go a long way towards expanding 4^th and 5^th graders’ notions of kindness.

Visibility of Compassion and Inclusion

A parallel educational consideration is understanding the virtues that hold the most value for children. We investigated this issue by examining which kindness virtues are most visible to children’s peers and teachers. We found partial support for the visibility hypothesis, as children who generated some of the most prototypical virtues (which we now know to be generosity, compassion, and inclusion) were perceived as more kind by their peers and teachers. Specifically, compassion and inclusion were linked to higher peer nominations of kindness, but generosity was not. In contrast, no virtues were significantly associated with teacher ratings of children’s kindness, although inclusion was marginally significant in the positive direction. The most prototypical virtues were only intermittently linked to peers’ evaluations, and even weaker associations were found for teachers’ evaluations.

An alternative explanation for these findings is that visibility is driven by the virtues that hold the greatest benefit to peers and teachers. Generosity may be less visible because its benefits can be bestowed in response to both satisfactory and unsatisfactory conditions. In contrast, compassion and inclusion are responses to suffering and isolation (Gest et al., 2014; Strauss et al., 2016), respectively, that would not likely go unnoticed by the recipients of such kindness. This explanation also accounts for the finding that peers’ evaluations were more strongly linked to compassion and inclusion than were teachers’ evaluations: Peers are more likely to be the beneficiaries of children’s compassion and inclusion, whereas teachers customarily inhabit an observer or leadership role in the classroom. One reason why inclusion was marginally associated with teachers’ evaluations may be because teachers are responsible for managing classroom social dynamics in ways that support isolated students (Gest et al., 2014). Thus, teachers may be aware of inclusive children as they can help teachers with this goal, for example, by being paired with an isolated student (Braun et al., 2019).

Broadly, finding that only certain virtues predicted peer nominations of kindness underscores the importance of differentiating among the virtues and considering their unique value (e.g., for different social partners; for different sorts of virtue-relevant situations; Wright et al., 2020), rather than lumping the virtues together under the moniker of prosocial behavior (Fowers et al., 2021).

The Breadth of Kindness Schema Repertoires has Real-World Consequences

Differentiating virtues in children’s mental representations also provides a meaningful method to operationalize the repertoire of distinct ways in which a child thinks about kindness, albeit as a step toward constructing a single index of kindness schema repertoires. We found partial support for the schema repertoire hypothesis: children with more varied kindness schemas were evaluated as more kind by their peers but not their teachers. The breadth of schema repertoires is significant from a social-cognitive perspective (Lapsley & Narvaez, 2004; Wright et al., 2020), as children with more varied kindness schemas may interpret social information in kindness-relevant ways and have cognitive access to a wider array of kind behaviors. Further, results provide evidence that schemas have such real-world consequences, as they appear to manifest in behaviors that are acknowledged by peers. These findings are consistent with prior research showing that mental representations in the moral domain predict real-world behavior (e.g., Hertz & Krettenauer, 2016).

It should also be noted that in the interest of parsimony, we restricted our presentation of findings regarding the schema repertoire hypothesis to data from Time 1. Sensitivity analyses found that the breadth of children’s kindness schema repertoires was positively associated with both peer and teacher ratings of kindness at Times 2 and 3. Future research is required to determine whether, and under what circumstances, schema repertoires are associated with teacher ratings of students’ kindness, yet these findings provide support to a social-cognitive approach to virtue.

Development of Breadth of Kindness Schema Repertoires in the Classroom

We examined whether the breadth of children’s kindness schema repertoires develop through informal social processes. Results did not support the social development hypothesis, as peer acceptance, teacher-child closeness, and teacher-child conflict earlier in the school year did not predict within-person longitudinal changes over time in the breadth of children’s kindness schema repertoires. We discuss these findings in relation to the limitations of the study design in more detail below.

However, between-person associations were found; across all timepoints, children who were more accepted by their peers and had closer and less conflictual relationships with their teachers had larger kindness schema repertoires throughout the study. These between-person findings are consistent with early peer relations research on accepted children. Children who are well-liked by their classmates receive more social initiations and engage in more social conversations with peers (Gifford-Smith & Brownell, 2003), affording them a wider variety of interactions that can be met with various forms of kindness. In addition, research indicates that kindness is highly valued among children’s peers (Wagner, 2019) and increases peer acceptance (Layous et al., 2012), pointing to a benevolent cycle in which peer acceptance serves as both a consequence of kindness and an impetus for it (see Rudolph, 2021).

Similar remarks apply to the role of children’s relationships with teachers. Such relationships not only suggest that teachers may be secure attachment figures (Verschueren, 2015) who model (Bandura, 1969) and explicitly set positive classroom norms (Lerner et al., 2015) that could foster the development of children’s kindness schemas, but conversely, children who enact many forms of kindness are likely the easiest students with whom teachers can establish close and low-conflict relationships (Weyns et al., 2018). Relationships and children’s kindness appear to be mutually influential, and intervening on either side this dynamic may be appropriate, although our null longitudinal findings offer no support for causal relations.

Limitations and Future Directions

This study embraced a mixed-method, multi-informant, longitudinal design which centered children’s own voices in the examination of kindness and its development. Yet, findings should be considered in light of several limitations. The study began in December (nearly halfway through the school year), when classroom social dynamics, and their potential consequences for kindness schemas, were likely already established, potentially leaving little room to detect changes. Thus, without further theory-driven testing, we are hesitant to draw firm conclusions that relationships with peers and teachers are not instrumental for the development of kindness schemas. Future studies should not only commence data collection at the start of the school year but also measure the constructs with sufficient frequency to track changes in rapidly-forming relationships and their associations with kindness schema repertoires.

A second set of limitations concerns the measurement of kindness schemas, which, in asking students ‘What are some ways you can show kindness to others?’ may have primed concrete behaviors and sidelined virtues such as humility or gentleness that pertain to the tone, or manner, in which one is kind. Relatedly, the question was other-focused, precluding mention of self-oriented expressions of kindness (e.g., self-compassion; Neff, 2003). Combined with the location of this prompt at the end of a long survey, this suggests that the breadth of children’s kindness schema repertoires may have been underrepresented in this study. Future research examining prototypicality should employ a more inclusive open-ended stem early-on in the survey and should use the full set of prototype analysis methods (e.g., centrality ratings, recognition tasks; Gulliford et al., 2021; Lambert et al., 2009).

Finally, schemas are merely one part of virtue and should not be viewed as a stand-in for virtuous, situationally appropriate behavior. Virtue scholars emphasize that constitutive of virtue are (a) proper motivation to act for the right reasons (Snow et al., 2020), and (b) practical reason⁷ to perceive which virtue(s) are called for by the situation (Snow et al., 2021). In addition to examining motivation and practical reason, future research should directly investigate behavior and whether schemas for different kindness virtues are differentially linked to schema-expressive behaviors.

Conclusion

This study contributes to positive psychology theory on virtue and has implications for cultivating children’s kindness through SEL, character and virtue education, positive youth development, and mindfulness-based interventions. Findings revealed that 4^th and 5^th graders’ kindness schemas include a suite of moral virtues that vary in cognitive accessibility. Generosity, compassion, and inclusion were most accessible and thus central to children’s kindness schemas, and may be ideal launching points for interventions, whereas scaffolding may be required when introducing civility, harm avoidance, and gratitude as forms of kindness. Virtues that serve as responses to vulnerability (compassion, inclusion) may be especially valuable for children’s peers. Further, positive relationships with peers and teachers may serve as sites for the practice and cultivation of kindness virtues.

Open Scholarship

This article has earned the Center for Open Science badge for Open Materials. The materials are openly accessible at https://osf.io/yc4eg/?view_only=866f73554bc64504be85ce9fee9f671b.

Acknowledgments

Thank you to the district administration and all participating teachers, students, and families, and all the research assistants who contributed to this study.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Data Availability Statement

Due to the nature of this research, parents did not agree to have their children’s data be shared publicly, so supporting data is not available.

Supplementary Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/17439760.2023.2170822

Additional information

Funding

This research was made possible through generous support from the Random Acts of Kindness Foundation. Opinions expressed are those of the authors and do not necessarily represent the granting agency.

Notes

1. Data were originally analyzed at each timepoint for the prototypicality, visibility, and schema repertoire hypotheses. In the interest of parsimony and because results from Time 1 reflected the most conservative and trustworthy set of findings (i.e., fewer associations were statistically significant and when they were they were replicated in at least one subsequent timepoint), we later decided to restrict results presented in the text to the Time 1 data only for these three hypotheses. For transparency, we present findings from Time 2 and Time 3 data as ‘sensitivity analyses’ in subsequent footnotes.

2. When combined with Time 2 and Time 3 data, the ordering of the virtues in terms of prototypicality was virtually identical to the Time 1 data alone, with the exception that a slightly smaller percentage of children produced compassionate than miscellaneous responses when averaged across all three timepoints: Generosity (57.87%), Miscellaneous Kindness (43.32%), Compassion (41.20%), Inclusion (34.94%), Civility (30.49%), Harm Avoidance (16.33%), Not Kindness-Relevant (8.60%), and Gratitude (2.52%). The ordering of the virtues was identical (across all three timepoints vs. Time 1 alone) for the other two prototypicality indices.

3. Sensitivity analyses replicated the positive association between Compassion and peer-rated kindness at Times 2 and 3, and replicated the positive association between Inclusion and peer-rated kindness at Time 2.

Additional intermittent associations were also observed at Times 2 and 3: Generosity was linked to higher peer-nominated kindness at Time 3 only, Generosity and Civility were linked to higher teacher-rated kindness at Time 2 only, and Compassion was linked to higher teacher-rated kindness at Time 2 only. We suggest that these intermittent associations are not reliable findings.

4. Sensitivity analyses found that breadth of kindness schema repertoire was positively associated with both peer-nominations and teacher-ratings of kindness at Times 2 and 3.

5. Note that these are standardized estimates and therefore they differ somewhat over time. Unstandardized estimates in all models were constrained equal over time.

6. Note that forgiveness responses were very infrequent and did not have their own unique code. Forgiveness responses were assigned the miscellaneous code.

7. We use the term ‘practical reason’ rather than ‘practical wisdom’ in recognition that the latter is conceptualized as a highly developed achievement not likely possessed by children (Wright et al., 2020).