Enhancing virtue via five modalities: empirical evaluation of utilizing a meta-model for character formation in higher education

A renewed emphasis on character formation in higher education is evident in recent years, with significant funding directed toward character education initiatives, particularly at the collegiate level (Nietzel, 2025). This movement has helped to fuel the creation of numerous programs, frameworks and strategies focused on virtue formation (Brant, Lamb, Burdett, & Brooks, 2020; Cockle, Melton, Glanzer, Hansen, & Schnitker, 2026; Domingo & Fernández Espinosa, 2025; Lamb et al., 2022; Tausen, Erickson, Neinhuis, & Jalalian-Chursky, 2026a; Tausen, Fossum, Hoff, Erickson & Flores, 2026b; Wang et al., 2025). While definitions of virtues and cultivation approaches vary, most conceptualizations emphasize a relatively stable pattern (disposition) of thinking, feeling and acting in the service of a greater good beyond oneself (Lerner, 2019; Ratchford, Pawl, Jeffrey, & Schnitker, 2024). Across definitions, there is often a central belief that virtue can indeed be cultivated or developed, and that growth in character formation is intrinsically meaningful for human beings in general as well as in the context of higher education. Further, an interdisciplinary emphasis on habits as a mechanism for change underscores the need for concrete practices that can be implemented in a classroom or college setting.

Notably, attempts to cultivate character have a growing body of empirical evidence to draw from. The field of positive psychology offers a wealth of data identifying interventions that reliably enhance individual virtues such as patience, love and temperance (Niemiec and Pearce, 2021). Increasingly, such interventions are being integrated into classroom pedagogy and educational programs (Ng, Ng, & Siu, 2025; Qianqian, Bin Abu Bakar, & Caixia, 2025; Seligman, Ernst, Gillham, Reivich, & Linkins, 2009; Shankland & Rosset, 2017; Tausen et al., 2022, 2026a, b; Waters, 2011). These efforts provide valuable evidence-based approaches to character formation that align with broader moral formation and character education goals, goals that extend beyond content mastery or intellectual development (Arthur, 2003). As a bonus, integrating positive psychology interventions into classrooms and educational programming can enhance student well-being (Lo & Leow, 2025; Ng et al., 2025; Sin & Lyubomirsky, 2009), a critical concern in emerging adulthood, in line with Aristotelian conceptualizations of virtue development as a core component of personal flourishing (Kraut, 2021).

While the introduction of empirically grounded techniques into the classroom is valuable, it is possible that their efficacy is limited to particular types of courses where there is strong fit between the interventions and course content (Hammill, Nguyen, & Henderson, 2022). Even if interventions prove to be equally effective regardless of fit or professor delivery, efforts by individual faculty members over the duration of a term (10–16 weeks) are unlikely to produce the sustained, transformative effects necessary to shape college students over the course of their entire education as many institutions promise. Meaningful character formation in higher education instead may require more coordinated, programmatic approaches that intentionally scaffold course-relevant virtue-based interventions across the entire curriculum as well as extracurricular and co-curricular activities (Lickona, 1996; Tausen et al., 2026a, b).

A recurring theme in character education conversations is the idea that students’ character, the constellation of their virtues and vices, is inevitably formed by education and that educators have a responsibility to positively guide that transformation (Kiss & Euben, 2010). Just like most professors are not trained in best pedagogical practices, very few have been equipped with tools to intentionally implement character formation practices in their classrooms (Berkowitz & Bier, 2004). Institutions can help by offering clear definitions of character education, defining contextually valued virtues, scaffolding virtue-relevant content across the curriculum and adhering to frameworks or strategies that can be readily adapted into classroom settings (Lickona, 1996; Tausen et al., 2026a). Two recent positive examples of intentional training efforts include Wake Forest University’s Educating Character Initiative “Character Across the Curriculum” workshop and Notre Dame’s Center for Virtues and Vocation’s “Integrating Virtue Together” workshop. Both institutions have offered training workshops to facilitate a more comprehensive and intentional approach to character education by focusing on virtue cultivation at the individual classroom level.

A core component of these workshops (both attended by the first author) is the presentation of the “Seven Strategies for Virtue Development” proposed by Lamb, Brant and Brooks (2021). Drawing on Aristotelian thought, these strategies conceptualize virtue as habit and outline practical approaches such as friendships of mutual accountability (Hoyos‐Valdés, 2018) and moral reminders that are readily applicable within educational settings. The strategies are philosophically grounded and offer clear practical value to professors seeking to “teach virtue.” In this way, they offer a “bottom-up” approach to character formation by seeding specific techniques that can be employed in a classroom setting to help students cultivate character. Ongoing empirical work probing the efficacy of the seven strategies in distinct settings will continue to strengthen our understanding of the unique contributions of each strategy, as well as how they function in relation to one another (Lamb et al., 2021, 2022). Such work will also have important implications for instructional design at both the classroom and curricular levels.

Another approach to character formation is to provide educators and practitioners with a top-down framework that offers a theory of change. A theory of change seeks to clarify the primary pathways through which change occurs. As Hansen, Melton, Cockle and Davis (2024; Para. 3) explain, “A good theory of change should make program decisions clearer: Given (1) how you think change happens, (2) your goals for change, and (3) your limited time and resources, which activities will you prioritize when designing your programs?” One example is the “head, heart, hands” framework introduced in the Kare model (Melton et al., 2024), which has more recently been named the Process of Self-Alignment for Virtue Model (PSAVM; Cockle et al., 2026). Grounded in virtue science research, this approach identifies three core avenues for change: knowing (head), enacting (hands) and aligning (heart). The framework is rooted in broader theories of moral development and offers a useful meta-framework for thinking about the whole person. It enables educators to consider whether they are prompting students to reflect on their transcendent narrative identity, providing opportunities for active virtue practice and inviting students to “check in” by comparing their actual selves with their ideal or ought selves (Higgins et al., 1986). While further empirical validation is needed, the model represents a meaningful theoretical advance to guide institutional program development.

Bottom-up approaches, such as positive psychology interventions or the Seven Strategies (Lamb et al., 2021, 2022), and top-down approaches, such as the PSAVM (Cockle et al., 2026), are not mutually exclusive. On the contrary, the PSAVM explicitly identifies overlap with the Seven Strategies, just as our five-modality framework can be readily linked to specific strategies and interventions (see Tausen et al., 2026a). Integrating these two types of approaches is likely to be most effective in educational settings, as together they articulate the core domains necessary for change while also offering practical, easily implemented tools within these broader categories.

Notwithstanding the theoretical underpinnings and strengths of the frameworks and strategies being offered in higher education spaces, little empirical evidence to date exists to directly evaluate the effects of their implementation in higher education. In one meta-analysis of character education, of the 52 studies that met the required thresholds for inclusion in the analysis, only one of them was conducted on college students (Jeynes, 2019). A few recent investigations of college-age adults or older include empirical evaluations of programs that draw heavily upon the Seven Strategies, such as Oxford’s Global Leadership Initiative (OGLI; Brant et al., 2020) and Oxford’s online Leading with Character (LwC) Course (Wang, Birin, & Brooks, 2025). Results from these investigations are promising and show that intentional programmatic approaches to cultivating virtue can and do have detectable effects in self-reports of virtue. For example, students who completed the OGLI showed growth (pre- to post-test) on two of the four virtues targeted in the program (Brant et al., 2020). A report disseminating the results of a large randomized controlled trial (course vs. waitlist groups) examining the LwC course provided quantitative evidence of growth on three of the four course virtues (purpose, love, hope) and qualitative evidence of deep engagement with courage as a virtue (Wang et al., 2025). In a similar vein, Lamb et al. (2022) demonstrated positive changes in course-targeted virtues when comparing students in a university seminar course entitled “Commencing Character” relative to students in other university classes (e.g. alternative first-year seminars, introduction to psychology). Together, this work provides a solid foundation of empirical evidence that suggests character courses and programs can indeed help college-age students and adults grow in self-reported virtue.

Notwithstanding the value of these findings, more work is necessary to determine the efficacy of character education efforts in higher education. This is important from both a methodological and practical standpoint. Practically, this is important if we want to understand drivers of meaningful character formation at the institutional level. Because many universities endeavor to graduate students of character (Arthur, 2024; Nucci, Narvaez, & Krettenauer, 2014), it is important to scrutinize the utility of strategies or frameworks that (1) get woven into typical coursework for university students (rather than classes explicitly about virtue or programs that are taken in addition to core coursework) and (2) are delivered by professors in a wide variety of disciplines (not only the ones whose primary training or scholarship is in the domain of virtue formation).

Methodologically, the challenges associated with assessing targeted character formation efforts often result in a relatively small number of students who self-select into the program or course because they are personally motivated to grow in the ways the program is being advertised or disseminated. Other investigations that do utilize random or quasi-random assignment and large samples may rely on waitlists as a control condition. While this is better than no comparison at all, where active control conditions are possible, they are the most rigorous assessment of a targeted character intervention. Recognizing that it is both difficult and necessary to assess whether programmatic efforts foster sustained growth among university students (Berkowitz & Bier, 2004), we sought to address some of these methodological limitations in our evaluation.

The current manuscript evaluates the utility of a meta-framework to support professors across a variety of disciplines in designing classroom assignments and activities that meaningfully cultivate virtue (Tausen et al., 2026a) and contribute to the moral ecology of an institution (Tausen et al., 2026b). This framework identifies key modalities through which personality and behavior change occurs (see Tausen et al., 2026a for more details). Specifically, it organizes five change modalities across seven institutionally valued virtues–although any virtues could be incorporated into the meta-framework. The five change modalities reflect the fact that professors can utilize change strategies that emphasize environments/situations (e.g. exposing students to virtuous exemplars or stories, or formative relationships or physical spaces), cognition (e.g. knowledge of virtue, training to think virtuously), the affective/somatic domain (e.g. arousing moral emotions), behavior (e.g. rehearsing skills, role-play or enacting virtues) and motivation (e.g. intrinsic and extrinsic incentives).

The seven virtues targeted in this study included the four cardinal (courage, justice, wisdom, temperance) and three theological (faith, hope and love) virtues championed by philosophical and theological thinkers like Plato, Aristotle, Augustine and Aquinas (Erickson et al., 2026; Stalker, 1902; Tausen et al., 2026a, b). These seven were selected given their importance at the institution where our data was collected and the institutional approach to training professors to define and cultivate these virtues as related, but distinct constructs.

As previously noted, the five change modalities could be applied to any virtue list. Institutional context (which promotes theologically informed virtue definitions) as well as our ongoing work demonstrating a seven-factor structure for the virtues in our Christian context (Erickson et al., 2026) also guided decisions in our work to treat these virtues as seven distinct factors throughout the manuscript. Notably, however, there exist alternative models of virtue factor structure, especially recent work supporting a three-factor structure for the VIA (Han & McGrath, 2023; McGrath, Greenberg, & Hall-Simmonds, 2017; McGrath, Brown, Westrich, & Han, 2022), which may be more appropriate for future work in contexts without a specific set of distinct and institutionally valued virtues which guide faculty training and are being taught in classrooms to students.

When looking at how professors naturally distribute their assignments across these five modalities and seven virtues (prior to the workshop), our own pilot data indicated that professors tended to rely heavily on cognitively oriented strategies and target cognitive virtues such as wisdom, while underemphasizing other virtues and modalities (Tausen et al., 2026a). This is noteworthy because it suggests room for growth in the ways that professors design their assignments to ensure they are incorporating multiple modalities (in line with the theory of change assumptions implicit in our meta-framework) and multiple virtues.

Professors were introduced to the meta-framework (a core set of seven virtues and five modalities of formation) via a day-long “Classrooms that Cultivate Character” (CCC) workshop (Jalalian-Chursky, Tausen, & Erickson, 2026; Tausen et al., 2026a). The workshop aimed to help professors develop a common language around the core virtues, strengthen connections between course assignments and multiple virtues, and most explicitly, to “stretch” assignments beyond cognitive modalities toward a more multi-modal model of virtue formation. The CCC workshop was intentionally designed as an interactive, hands-on professional development experience that emphasized active learning, collaboration and practical application rather than passive instruction (see Jalalian-Chursky et al., 2026 for full details). Throughout the day, faculty explored the historical, theological and psychological foundations of the university’s seven core virtues before examining the five modalities of virtue formation and their application to teaching and learning. The workshop included sessions on developing a shared conceptual understanding of the seven virtues, introducing the five modalities of virtue formation, analyzing exemplar virtue-based assignments through the lens of the five-modality framework, collaboratively redesigning existing course assignments, presenting redesigned assignments for peer feedback and planning for implementation within participants’ courses. This progression allowed participants to move from understanding the theoretical foundations of character formation to creating discipline-specific assignments that could be immediately implemented in their courses. Practically, these newly designed assignments needed to engage multiple modalities and multiple virtues, to help professors move beyond their over-reliance on the cognitive modality and virtues like wisdom in their courses (see cultivatecharacter.com for example assignments and materials; Tausen et al., 2026a). Finally, professors were given “boiler plate” information to incorporate into their syllabi, which listed the seven intuitional virtues and their definitions as well as a statement about the importance of character formation at the institution.

A distinguishing feature of the CCC workshop was its intentional interdisciplinary design. Rather than focusing on a single department or disciplinary perspective, the workshop brought together small cohorts of faculty from across the university, creating a collaborative learning community in which instructors examined character formation through the lens of their own disciplinary expertise while engaging with colleagues from diverse academic backgrounds. Faculty who attended initial workshops and participated in this study ^[1] represented psychology, statistics, philosophy, history and theology. This breadth allowed the framework to be tested across a variety of disciplines, introductory and upper-division courses, and general education and major-specific curricula.

Rather than prescribing a standardized curriculum, faculty were encouraged to adapt the meta-framework to their existing course objectives while maintaining disciplinary integrity. The interdisciplinary cohort model fostered cross-disciplinary collaboration, enabling participants to exchange pedagogical perspectives, collaboratively refine assignments and identify authentic opportunities to embed virtue formation within their own courses. This design increased the ecological validity of the intervention by demonstrating that the framework could be adapted across a wide range of pedagogical contexts without requiring substantial curricular overhaul.

To control for extraneous variability in virtue content and instructor variance, we recruited professors who taught (1) a required general education or psychology course and (2) at least two sections of the same course within 12 months. This recruitment strategy facilitated the collection of student data at the beginning (pretest) and end (post-test) of courses both before and after faculty attended the workshop and utilized the meta-framework to construct an assignment for their course. This approach, while not fully randomized, yielded a relatively rigorous quasi-random control group as there was no indication to the students prior to enrollment that the course was being augmented with additional virtue-based assignments. Notably, control courses were taught prior to faculty workshop attendance (with the exception of two courses where both sections were taught in the same quarter). Even when control and intervention classes were taught concurrently, only the intervention class contained an explicit virtue assignment informed by the five-modality × seven virtue framework and incorporated explicit definitions of the seven core virtues into course syllabi (see Tausen et al., 2026a). Moreover, all participating courses fulfilled either institutional general education requirements or major requirements, reducing the likelihood of self-selection into virtue-focused courses. These matched-control classes resulted in more conservative comparisons than previous empirical work evaluating character education interventions because they accounted for the natural variation in virtue-related content across diverse academic disciplines (e.g. statistics, theology, history, psychology and philosophy) rather than comparing fundamentally different course types.

We articulated several hypotheses about the effects of applying the CCC workshop’s multimodal framework on student character outcomes in courses revised using the meta-framework, relative to pre-training “as usual” versions ^[2]. First, we hypothesized that students in the revised “Intervention” courses that integrated the meta-framework would perceive greater focus on character relative to those in “as usual” courses (control group) and greater retrospective ratings of virtue growth. Second, we hypothesized pre-to-post increases in character outcomes in the active intervention group, including total self-reported character virtues (e.g. wisdom, courage, justice) and character strivings (striving to be a better person, personal importance of character), providing a direct test of our program. Lastly, we hypothesized a condition-by-time interaction in which participants in the active character integration courses would show greater increase in character virtues and strivings relative to those in “as usual” control courses.

Students enrolled at a small, private institution in the Pacific Northwest participated in this study as part of their enrollment in a college course (N = 542 after omitting duplicate responses). While survey completion was required for class, all respondents could consent to or decline their data being used for research purposes. Students who did not consent to data usage, were not 18 years of age or were outliers in survey completion duration were dropped. Data analyses were performed on the 451 remaining participants. Participants in our final sample ranged in age from 18 to 45 (M = 20.40, SD = 2.62) and identified as 72% women, 26% men, 1% non-binary or agender and <1% other. Regarding race/ethnicity, participants identified as 34% non-Hispanic White, 21% Hispanic/Latino/a, 18% Asian or Asian American, 11% Black or African American, 10% Multiracial, 1% Middle-Eastern/North African, 1% American Indian or Alaska Native and 1% non-responders.

The study employed a 2 (intervention condition: experimental vs. control) × 2 (time: pretest vs. posttest) mixed design. Control conditions entailed a “class as usual” paradigm where professors taught their course how they typically teach it (before integrating the meta-framework). The experimental or intervention condition entailed a “course modified to enhance virtue” paradigm that relied on the integration of at least one meta-framework compliant classroom assignment (see intervention described above).

Students in all classes were asked to self-report on their virtues at pretest and post-test and reflected on their perceptions of course virtue integration and impact of the course on their own virtue growth only at post-test. Data presented here were derived from a larger study administered via Qualtrics to assess various factors associated with virtue perceptions and formation for college students. The survey was open for one week at the beginning (Pretest) and at the end of term (Posttest). At each data collection, participants were asked to indicate whether or not they consented to their survey responses being utilized for research purposes. Upon survey completion, participants received a unique code and submitted it via the institution’s online learning system to receive research participation credit. In this way, instructors had no knowledge of who consented to their data being utilized, protecting the anonymity and autonomy of the participants.

This study probed students’ perceptions of virtue integration in courses, self-reported virtues and character strivings.

Seven items were created by the authors to capture the extent to which students perceived virtue as explicitly integrated into a particular course and shaping their experiences. Examples include, “Character formation (helping students develop virtue strengths) is a primary aim of this course” and “My experiences in this class have helped me deepen my commitment to the greater good.” Responses were collected on a 5-point scale from 1 (Completely Disagree) to 5 (Completely Agree). An additional set of seven items used the stem, “To what extent did the content in this course (lectures, activities, or assignments) emphasize the following virtues or strengths or character…” and named faith, hope, love, courage, justice, wisdom and temperance separately. Items were rated from 1 (Not at all) to 5 (Very Much). Internal consistency estimates were strong for both course virtue (α = 0.90) and course experiences (α = 0.89).

The VIA-IS-V6 is a 48-item scale that assesses six hypothesized virtue domains (McGrath, 2017). Responses are collected on a five-point scale from 1 (very much unlike me) to 5 (very much like me). The VIA scale showed strong internal consistency in this study (α = 0.91), justifying the creation of a single composite score. Subscales also showed internal consistency (courage: α = 0.74, humanity: α = 0.72, temperance: α = 0.65, transcendence: α = 0.72, justice: α = 0.72, wisdom: α = 0.74); we created composite scores for each of the six primary VIA virtues to afford a more nuanced exploration of potential virtue change over time.

The short-form Compassionate Love of Humanity scale (CLS-H-SF; Chiesi, Lau, & Saklofske, 2020) is a nine-item assessment tool that probes a sense of care and compassion for the suffering of strangers rather than just members of one particular group. Responses are collected on a 7-point scale from 1 (not at all true of me) to 7 (very true of me). The CLS-H-SF scale showed strong internal consistency in this study (α = 0.94). Chiesi et al. (2020) also provided evidence for a one-factor structure and convergent validity of the CLS-H-SF, although test–retest reliability was not assessed.

The Eschatological Hope scale (vanOyen-Witvliet et al., 2022) is a six-item assessment tool that probes dimensions of hope central to a Christian theological tradition (e.g. “I am confident that God will overcome evil”). Responses are collected on a 5-point scale from 1 (not at all like me) to 5 (exactly like me). The Eschatological Hope scale showed strong internal validity in this study (α = 0.97). vanOyen-Witvliet et al. (2022) reported a single-factor structure, good test–retest reliability and convergent validity for the Eschatological Hope Scale, including expected associations with related measures of hope, religiosity and spirituality.

The Santa Clara Strength of Religious Faith scale (Plante & Boccaccini, 1997) is a ten-item assessment tool that probes religious engagement and identity. Responses are collected on a 4-point scale from 1 (strongly disagree) to 4 (strongly agree). The SCSRF items showed strong internal validity in this sample (α = 0.97). Plante and Boccaccini (1997) reported excellent internal consistency and convergent validity for the SCSRFQ.

Two items constructed by the authors were utilized to probe students’ perceived importance of (i.e. “being a person of strong moral character is important to me”) and striving for (i.e. “I am always striving to be a better person”) character growth. Responses were collected on a visual analog scale from 0 (completely disagree) to 100 (completely agree).

Planned analyses included several steps. First, we screened data to remove survey responses taking an extremely low or high amount of time and then checked for normality and presence of outliers. Next, to determine whether students consciously perceived active intervention courses as involving relatively greater virtue integration, we compared intervention versus control groups at posttest on student perceptions of whether their course focused on character virtues as well as retrospective perceptions of whether the course enhanced their own character. Because these perceptions were assessed only at one time point, independent-samples t-tests were the appropriate method.

We tested our core hypotheses about student changes over time in self-reported character virtues (VIA virtues, SCR Faith, EHS Hope and CLS Love) and visual analog scales related to character strivings (importance of character, striving to be a better person) via multilevel modeling (MLM). MLM was appropriate for our study because it permits missing and unbalanced data (i.e. differing time points) and models within-person dependencies in repeated measures, nesting time points (Level 1: pretest and posttest) within participants (Level 2). We considered 3-level models with repeated measures nested within individuals that were nested within professors, but the number of professors did not warrant a third level, and brief inspection of such models suggested little variance attributable to specific professors. Models calculated maximum likelihood estimates. Group (0 = control, 1 = intervention), time (0 = pretest, 1 = posttest) and group × time interactions were modeled as predictors, with character variables as outcomes. We report Cohen’s d as effect sizes with each model, using conventional heuristics for small (0.2), medium (0.5) and large (0.8) effects.

We planned a set of related models to test core hypotheses. We first tested unconditional models (i.e. with no predictors) to check whether participants differed significantly from each other in pretest levels of character variables (i.e. random intercepts). Next, we tested MLMs in the intervention group with only Time as a predictor, testing whether students increased on character variables across the academic term, on average. This permitted probing for character growth in the active condition prior to tests of whether those in the intervention group reported significantly greater growth relative to the control groups. Lastly, we modeled group, time and group × time effects on outcomes. Because both group and time each only had two levels, significant interaction terms were interpretable without requiring simple effect tests, as described below in the Results. Monte Carlo simulations for multilevel models (see Arend & Schäfer, 2019) suggested that for 400 participants with 2 repeated measures, the study would achieve 0.80 power or greater to detect small standardized effects. Because this was the first test of this program of research and given the conservative design of testing the intervention against a control group that did not completely exclude character formation, we initially report significance test results without correction. However, we also report which findings remained significant after correction for familywise Type I error via the Benjamini–Hochberg procedure (Benjamini & Hochberg, 1995), which compares the p-value for each statistical test to one of a set of increasingly conservative cutoffs for significance. In this way, there is no single p-value cutoff for each test, but a unique and increasingly smaller p-value threshold for each test.

Inspection of skew and kurtosis suggested normally distributed variables, and only six significant (p < 0.001) outliers were detected, which was reasonable for a sample of this size. We removed cases with extremely rapid (<200 s) or slow (>10,000 s) responses, resulting in a final sample of N = 451 (202 intervention and 249 control group participants). Overall, over 98% of participants provided data for MLM analyses; because MLM handles unbalanced data, we analyzed data in raw form. Unconditional multilevel models (with no predictors) suggested significant variability in intercepts (i.e. participants differed in pretest levels of character outcome variables), which justified use of MLM and retaining random intercepts in all models. Although model-fit indices differ in important ways – Akaike information criterion (AIC) is favored for predictive accuracy, whereas Bayesian information criterion (BIC) is appropriate for theory testing but incorporates sample size penalties (Vrieze, 2012) – we consulted both indices (see Supplemental Materials). Models without random effects for Time were lower in every case for AIC and in most cases for BIC, suggesting better fit. Thus, we included only fixed effects for Time (i.e. the average pre-to-post change represented well the change across participants).

Two independent samples t-tests probed whether participants perceived intervention versus control courses as differentially focusing on character formation. Contrary to hypotheses, groups did not differ on perceptions of their course as focused on character virtues, t(336.83) = 1.11, p = 0.268, d = 0.12. Means in intervention (M = 3.79, SD = 0.79) and control (M = 3.88, SD = 0.80) groups were similar and moderately high, suggesting perceived virtue integration in both conditions. Surprisingly, students’ retrospective perceptions of class experiences enhancing their character were not significantly different between groups, t(337.75) = 1.90, p = 0.058, d = 0.21, instead trending opposite of the hypothesized direction with slightly higher perceptions in control courses (M = 4.08, SD = 0.68) than intervention courses (M = 3.94, SD = 0.70).

Separate MLMs tested for intervention group pre–post change in each outcome (see Table 1). Significant increase over time (i.e. time effect) was present for VIA total, justice, courage, temperance, transcendence and wisdom but not humanity. Similarly, SCR Faith and CLS Love demonstrated significant within-person increases, on average, but EHS Hope did not. Lastly, single ratings of character importance, but not striving to be a better person, increased significantly. After controlling for type I error across this family of tests, each of these effects except for character importance remained significant. These results suggest the presence of character growth for specific virtues in general but left open questions of differential change of intervention course participants relative to those in control courses, as addressed in the subsequent analysis.

Next, each character outcome was investigated in a separate MLM with group, time and group × time terms (see Table 2). Group was coded as 0 (control) and 1 (intervention), and time coded as 0 (pretest) and 1 (posttest). As a result, intercept scores estimated the level of each outcome variable at pretest for the control group only, and thus, group effects test how much the intervention group’s mean differed from the control group mean at pretest. The intervention group was significantly lower than the control group in VIA Wisdom (and marginally lower on VIA Courage) at pretest; otherwise, groups did not differ significantly in pretest scores, suggesting relatively successful equivalence of starting levels across groups.

Given binary coding of predictors and two time points, Time effects in these models represent the pre-to-post difference in each outcome variable for the control group. Table 2 effects for Time show that control group participants increased significantly on VIA Justice (and marginally on CLS Love and EHS Hope); they decreased over time in striving to become better people.

Group by Time interactions in MLMs tested our core hypothesis of students’ greater increase in character outcomes in the intervention group relative to the control group. Supporting hypotheses, positive interaction term coefficients in several models suggested greater improvement in the intervention group (see Tables 2 and 3). Namely, the interaction was significant for SCR Faith, VIA Wisdom, VIA Temperance and VIA Courage and marginally significant for VIA Total (see Figure 1 for plots). In addition, the interaction was significant for striving to become a better person and marginally significant for self-reported importance of character. Contrary to expectations, the interaction term did not approach significance for EHS Hope, CLS Love or VIA Humanity, Justice and Transcendence. After controlling for type I error in interaction effects, only the interaction for striving to become a better person remained significant, although all trended in the hypothesized direction (see Figure 1). We note that this was a highly conservative test given that our intervention condition was tested against an active control condition that did not exclude character formation. In an effort to limit both Type I and Type II error, we present the uncorrected p values as well as note whether or not they reached significance with corrected thresholds. Recognizing that this is the first study to test a model of character formation against an active control group at a university that already emphasizes character formation, we interpret the data through the lens of overall patterns – which demonstrated greater growth in the intervention group.

In summary, although students in intervention and control classes did not consciously perceive marked differences between them on courses’ virtue integration and experiences, measurement of actual pre–post change suggested increases in student self-reported virtues over time in the intervention group – where faculty were trained to integrate assignments based on the 5 × 7 meta-framework – demonstrated relatively greater shifts in character-related outcomes.

The present study provides an initial empirical examination of the effectiveness of training professors to use a meta-framework to promote student virtue development over a 10-week course. Results indicated that, although participants in the intervention and control classes did not perceive substantial differences in the emphasis on virtue within courses or their explicit beliefs about how much the course helped them grow, students in the active intervention condition showed comparatively greater gains at or approaching conventional significance levels on several self-reported virtue-related outcomes (VIA Total, Courage, Temperance and Wisdom; SCR Faith; striving to become a better person). Even after applying a conservative correction for multiple comparisons, growth in virtues in the intervention courses over time remained. Consistent trends of greater growth in the intervention (vs. control) can be seen in Figure 1 as well as the uncorrected p values, although they dropped below significance levels under the more stringent correction. A cautious interpretation of both the hypothesized main effects and the consistent pattern of interactions across measures suggests amplified growth for students in the intervention condition. This accelerated virtue growth in courses where faculty integrated character formation into course assignments using the five-modality framework occurred against a pattern of minor increases in self-reported virtue in the control group – and even a decline in one measure (striving to become a better person).

Amid burgeoning interest in positive psychology and character education, a growing, but small body of empirical research has evaluated the integration of interventions to enhance virtues at the class or program level (Brant et al., 2020; Lamb et al., 2022; Wang et al., 2025). To our knowledge, the present study is the first to directly assess whether training educators on a meta-framework can have detectable character formation effects in their students. What is more, our study appears to be the first to investigate a coordinated approach to character formation where participants did not knowingly enroll in a class or program about character and where there was an active control group. Prior evaluations relied solely on pre–post designs within the intervention condition (i.e. open trial design) or on unmatched control groups risking attributing observed growth to the program when alternative factors, such as time investment or participants’ intrinsic motivation to cultivate character, may be responsible for the effects. Comparing results of intervention courses where the meta-framework was applied to courses conducted “as usual” where it was not provided a conservative test of whether explicit character formation efforts generally, and the application of a meta-framework more specifically, impacted student character outcomes.

Unexpectedly, retrospective ratings of course focus on character (e.g. character formation is a primary aim of this course) and consciously attributed virtue growth (e.g. my experiences in this course have helped me deepen my commitment to the greater good) were not higher in the intervention (vs. the control) group. Retrospectively perceiving one’s class as fostering growth even trended lower for intervention than control students. These findings were surprising because one might expect items directly tapping global course perceptions to be most sensitive to courses with explicit virtue focus, particularly compared to the challenge of detecting changes over only ten weeks in student character virtues. Examining these measures in isolation might lead one to expect that no meaningful pre–post changes in character occurred, as students did not perceive courses using virtue-focused pedagogy differently, on average, from how control group students perceived their courses, nor did they directly report that the class enhanced core aspects of character (commitment to a greater good, listening to different perspectives). Contrary to this interpretation, however, students in the intervention condition showed increases in self-reported virtues from pre-test to post-test, with those in the intervention condition demonstrating relatively larger gains in core virtues of interest, such as faith and temperance.

One possible explanation for the lack of student-perceived differences between conditions is that students in control classes may have interpreted virtue as broadly present throughout their coursework (e.g. virtue is everywhere) in the absence of explicit references to virtue definitions or character-based assignments in the syllabus. Alternatively, students in both control and intervention classes completed the virtue surveys at the beginning and end of the quarter, with the final survey administered alongside questions assessing course content. Although surveys were intentionally titled “personal reflection surveys” and students evaluated their course before reporting on their own virtue in an effort to minimize potential priming effects, it remains possible that students might have recalled the consent form or pretest measures of virtue, shaping global ratings, particularly at a university whose mission centers “competence and character.”

Regardless of the mechanism driving the lack of differences between conditions in student course perceptions, the finding that students in intervention courses reported greater pre-to post-test increases in virtue suggests that growth measured over time did not necessarily align with students’ retrospective perceptions of growth or the degree to which they perceived virtue to be emphasized in their course. This raises important methodological questions, particularly for programs that measure growth retrospectively at a single time point. It also raises interesting theoretical questions about conscious, backward-looking awareness of growth and whether it is necessary in order for growth to actually occur. Identifying which measures best align with other indicators of virtue growth (e.g. behavior, informant reports) remains an important direction for future research.

Pre-to-post increases were more apparent in the students in the intervention group for eight virtue outcomes, but interpretations of Figure 1 show that some also trended in the positive direction for several virtues in the control group. This might mean that professors fostered virtue formation regardless of condition, and it might be too difficult to stop formation from happening in “as usual” classrooms. Character formation may happen in the university setting with or without training on frameworks for virtue cultivation. While we champion this growth as educators, we acknowledge as researchers that the design choice (intervention vs. treatment as usual) resulted in an extremely conservative test of the efficacy of our framework. Unlike passive control groups (e.g. no-intervention wait-lists), our far more rigorous active control condition presumably already provided opportunities for character development even when not explicitly named. While making it more challenging to find significant differences, this setup also boosts confidence that the training workshops (Jalalian-Chursky et al., 2026) and employed frameworks are actually useful above and beyond what professors are naturally doing in their courses. Importantly, however, the professors who participated in our workshop were early adopters, so they might be the most “bought in” to character education and were likely already emphasizing intentional character formation in their classrooms, even if via a different lexicon or set of tools. As a result, we cannot assume that, left to their own devices, many classes or professors would show a similar increase in their students’ self-reported virtue growth, further emphasizing the need for fully randomized investigations where professors are equally likely to receive training on teaching character through the lens of the five-modality framework.

An alternative explanation for virtue increases seen here might be “demand characteristics.” However, had students in the intervention group known that the “right answer” was to say they had grown a lot in the course, one would also expect them to report greater emphasis on virtue content in their courses and greater growth as a result of the course. Additionally, the fact that pre–post changes were only reported for some variables and that there was greater growth reported by students in the intervention rather than the control classes argues against a generalized demand characteristic to report positive growth globally. It is also worth noting that visual inspection of Figure 1 suggests students in the control condition appeared to decrease on several virtues (e.g. humanity and temperance) as well as reports of striving for moral formation, suggesting that positive growth was not inevitable.

Together, these findings show promise for training professors to use our five-modality framework (Tausen et al., 2026a). Doing so affords an opportunity to evaluate and stretch course assignments to incorporate more virtues and modalities to enhance student formation. Utilizing the range of modalities might simply be chalked up to good pedagogy, and yet, we know that there are many constraints (e.g. time, money, lack of training for best practices, blind spots in one’s own pedagogical toolbox) that can impede a professor’s ability to utilize a full range of modalities in their teaching. Even as instructors of the framework, we note its utility for identifying gaps in our own assignments, which often neglect the affective/somatic modality prior to deeper scrutiny.

Theoretically, our framework is complementary to the bottom-up tools afforded in the Seven Strategies (Lamb et al., 2022) and top-down approaches like the KARE and Process of Self Alignment for Virtue Models (Cockle et al., 2026; Melton et al., 2024). Informed by a psychological and intervention science approach to changing personality and behavior (see Tausen et al., 2026a), the five-modality framework is more akin to a top-down approach, suggesting several paths or avenues by which change can occur. At face value, the five-modality framework appears to capture additional domains beyond cognitive (head/heart) and behavioral (hands) by adding the environmental, affective/somatic and motivational domains, for example trying to cultivate feelings of compassion (affective), learning how to internally regulate our nervous systems when anger sets in (somatic) or being exposed to moral exemplars (environmental) to elicit “moral elevation” (affective) that inspires emulation of virtue (Erickson, Valkren, & Mills, 2025), providing important alternate strategies.

Further, as psychologists, we note that motivation for change can come from within or beyond an individual. Intrinsic motivation to cultivate virtue is likely associated with a transcendent narrative identity or a sense of participation in a story beyond the self (Schnitker, King, & Houltberg, 2019). This motivation is reflected in the “head” component of the Process of Self Alignment for Virtue Model (Cockle et al., 2026) and in the cognitive recognition of a discrepancy between the ideal and actual selves during processes of realignment (Higgins et al., 1986). Of course, professors also have a great deal of power over the setting of extrinsic motivators and rewards in classes (e.g. grades, points) that can help nudge students toward engaging in activities to cultivate virtue. Students might be required to attend a free community dinner (Tausen et al., 2022), engage in service-learning (Agnew Cochran & Fozard Weaver, 2017; Dubinsky, 2002) or attend an exhibit providing an experiential sense of the trials and challenges facing refugees and immigrants (Fossum & Tausen, 2026). Such activities, even if not initially internally motivated, might still have a powerful impact on students’ character formation. Motivations can also be activated or reinforced through interventions like moral reminders (Lamb et al., 2022) which may appeal to intrinsic (personal satisfaction of living in alignment with our ideals) or extrinsic (e.g. social approval) rewards. Undoubtedly, there is much overlap across the frameworks and strategies currently circulating in institutions of higher education, and future work that disentangles the overlap and unique contributions remains an important area for future research.

Practically speaking, frameworks can be useful pedagogical tools to help professors ensure a well-rounded approach to character formation efforts in their classrooms. Pilot data from our own work has shown that prior to targeted training, professors understandably tend to gravitate toward cognitive modalities and intellectual virtues (Tausen et al., 2026a). Much like the research on learning styles has helped to combat myths that people have a singular learning modality (Husmann & O’Loughlin, 2019; Kirschner, 2017), we suspect that character formation efforts do well to limit overreliance on cognitive activities that fail to engage the whole person in ways necessary for formation to occur. The five-modality framework can help professors stretch their assignments to more holistically engage students and create more robust character education.

Considering the lens of practical or clinical significance, we note that our interaction effects were small. Applying an increasingly stringent correction for the number of tests made these interaction effects no longer statistically significant. Importantly, this correction was highly conservative given our research design (comparison to an active control) and also elevates the risk of Type II errors, where one fails to detect effects that exist. Moreover, most of the within-group increases in the intervention group remained significant after correction, and the trend across outcomes suggests a consistent pattern of greater gains (see Figure 1) but also the need for larger samples in future studies to increase power. Given the “low touch” nature of the pedagogy intervention, the rigorous comparison condition and tools predominately designed to measure stable dispositions rather than short term changes, it is remarkable that the intervention resulted in any detectable change beyond “treatment as usual” at a college that already touts character formation. It is also notable that participants already scored moderate to high on character variables at pretest–some of which might be attributable to self-enhancement effects–making significant character growth more noteworthy. Nonetheless, replication in another university setting would be useful in confirming the size of the effects.

We also acknowledge the challenges of attempting to measure character, particularly given the limitations of self-report and the current set of available tools, which are not specifically designed to detect short-term or even medium-term changes. While the VIA remains one of the most widely utilized virtue assessments, internal consistency for individual VIA scales was lower relative to total VIA and other three virtue measures. We augmented the VIA with additional virtue measures that capture theologically-informed operationalizations of hope, faith and love (compared to VIA Transcendence and Humanity) more appropriate to our context, but future work should examine the benefits and limitations of such measures in samples such as ours, which include both religious and non-religious participants. This might be particularly important given that the potential for growth in theological virtues (e.g. eschatological hope) might be limited to a subset of participants who have a strong religious identity and are taking classes that emphasize the theological nature of these constructs. Targeting religious students exclusively for the potential development of theological virtues as well as the development of tools specifically designed to detect short-term changes, perhaps akin to patterns of states or daily diary methods that can detect fluctuations in virtue expression at the level of the individual, could be particularly fruitful and would likely show even more robust growth in character growth relative to the current tools that are designed to tap into relatively stable dispositions.

Future work should also evaluate programs like ours through the lens of alternative VIA factor structures, like the three-factor structure demonstrated by McGrath et al. (2022) or other multi-virtue measures with stronger factor structure (e.g. Erickson et al., 2026), as well as perceptions of and the interaction with institutional “character climate” (Tausen et al., 2026b). Indeed, debates in the literature about how to measure virtue and a replicable factor structure for the multitude of virtues remain ongoing. Support for the six-factor structure of the VIA has been inconsistent, with a more recent and robust line of research proposing and replicating thread broad virtue factors related to caring, inquisitiveness and self-control (McGrath et al., 2022). In addition, work that triangulates virtue formation with behavioral measurements, informant reports or includes longer-term follow-up will further elucidate the staying power of character formation interventions delivered via a classroom setting.

Additional limitations pertain to study design. Strictly speaking, we did not fully randomize participants to conditions, although we have no reason to suspect that students were more likely to register for courses that were part of our intervention versus control conditions, and we included only required courses, making self-selection unlikely to drive observed findings. Students were also blind to whether their course occurred before or after the character workshop, further supported by the lack of differences in retrospective course perceptions. Additionally, for practical reasons, we did not counterbalance condition order (professors taught a control class before an intervention class), although two of the professors did teach intervention and control classes simultaneously. However, groups did not differ significantly on character variables at pretest, ruling out the threat that lack of full randomization could lead to different starting levels.

We also note pros and cons of each faculty serving as their own control group. The strength of this approach is that it may hold instructor variance constant, minimizing extraneous differences between conditions. On the other hand, it remains impossible to fully control for character content and processes. Moreover, faculty likely varied in integrating the framework and how many virtue-based assignments they required. Professors in the intervention condition might have gravitated toward particular virtues, accounting for why some changed more (e.g. faith, courage), potentially washing out effects of lesser-championed virtues. If few intervention classes targeted justice, this might explain weaker growth effects. Alternatively, many courses might already emphasize justice, leading professors in the intervention condition to incorporate virtues that were previously less central to their courses, such as courage or faith. Leaving professors the choice of which virtues (of the seven) to emphasize in their revised course supported buy-in and adoption of the workshop material, but limited statistical power. Future studies should examine course content in detail to determine “dosage.” Greater consistency of virtue selection across intervention courses may yield stronger effects. Practically, however, we champion a larger virtue set from which professors select virtues most relevant to their courses. Character education does not happen in just one class and repetition across classes over multiple years and contexts, where there is strong content fit (Hammill et al., 2022) may best serve character formation.

A final limitation is that our work, while having an active control condition (e.g. not students on a waitlist to take a class), did not directly compare our model to other models or frameworks for change. Thus, we are not able to speak directly to the advantage of using our model over other models but rather using an explicit model versus no explicit model. Future work will be necessary to evaluate if certain modalities within our own model are most effective, or if there is some threshold or tipping point for the number or combination of modalities leveraged that generate the most substantial virtue growth.

In a context of heightened attention to character formation in higher education and a great need for civic responsibility and moral formation, our work addressed a critical empirical gap, demonstrating the utility of training professors on a meta-framework that emphasizes a five-modality theory of change for virtue development. Utilizing an active control group, our results provided initial evidence that requiring assignment revision through the five-modality meta-framework following a short workshop intervention training can accelerate self-reported growth of some virtues over a ten-week period. These findings reinforce the benefits of practices that help professors stretch their assignments in ways that engage the whole person. The most robust character formation may require not only cognitive intervention but also environments, emotions, bodies, action and incentives toward creating habits of being that transcend a classroom, inviting individuals to think, act and be in ways that reliably serve a greater good beyond oneself.

The opinions expressed in this publication are those of the author(s) and do not necessarily reflect the views of Wake Forest University or Lilly Endowment Inc.

The supplementary material for this article can be found online.