Enhancing moral sensitivity in the aftermath of academic misconduct: Results from a quasi-experimental field study

ABSTRACT

Most secondary and postsecondary institutions take a behavioral approach in dealing with student cheating—punishing those caught with grade reductions and/or suspensions. While some form of punishment may be necessary, it is not sufficient. As an instantiation of negative morality, academic misconduct offers an opportunity for moral education. The present investigation builds on the literature related to developmental approaches in responding to academic misconduct. It does so by describing theoretical underpinnings and instructional design of a developmental approach (Intervention), as well as results from a quasi-experimental study of its effects on moral sensitivity. Participants (N = 798) included university students who had been found responsible for academic misconduct and completed the Intervention. As hypothesized, participants not only reported greater attentiveness to moral issues after completing the Intervention, they also demonstrated greater awareness of the moral values related to academic misconduct. The implications and limitations of these findings are discussed.

KEYWORDS:

• Academic misconduct
• moral functioning
• moral sensitivity
• developmental approaches

In a 1996 edition of the Journal of Moral Education, Oser (1996) posited that educational institutions were not capitalizing on what he called ‘negative morality’ or ‘negative moral knowledge’ as an opportunity for students’ moral development. ‘[N]egative moral knowledge’ can be acquired directly through personal engagement in misconduct (e.g., bullying, cheating, lying, stealing) but also vicariously by seeing or hearing about the consequences of negative actions committed by others. This ‘negative moral knowledge’ is, of course, opposite of positive moral knowledge in which we focus on the good things that people do and the rewards that follow. While both can be powerful learning tools, Oser argued that educational institutions seemed to have shied away from the learning power of ‘negative morality’.

Oser’s (1996) claim still holds true today, at least in the arena of academic misconduct. The typical response to academic misconduct (e.g., cheating and plagiarism) is almost always strictly punitive (e.g., grade reduction and/or suspension). Although some form of punishment may be warranted, it is—when delivered alone with no opportunity for structured reflection and learning—insufficient and contraindicated (Bertram Gallant & Stephens, 2020). From a cognitive-development perspective, emphasizing ‘punishment and obedience’ or ‘law and order’ for university students is regressive and unlikely to stimulate moral growth (Kohlberg & Hersh, 1977). In contrast, Oser hypothesized that negative moral experience—responded to with opportunities to learn from it—‘leads to higher sensitivity to moral and social problems’ (p. 72).

In this study, we focus on an educational intervention (Intervention hereafter) that was designed to leverage learning from a moral mistake. Specifically, the Intervention was designed to develop students’ moral functioning (e.g., Rest et al., 1999)—particularly their moral sensitivity—in the aftermath of being found responsible for academic misconduct. Broadly speaking, moral sensitivity is concerned with awareness of how our actions affect other people (Bebeau et al., 1999). While Narvaez and Endicott (2009) have operationalized moral (or ethical) sensitivity as comprising seven skills (e.g., reading and expressing emotions, taking the perspective of others, preventing social bias), the Intervention focuses on enhancing the recognition and ascription of importance to the moral values (e.g., honesty, respect, responsibility, fairness, and trustworthiness) at stake, or in conflict, in a particular situation. As previously argued (e.g., Stephens & Wangaard, 2016), developing moral sensitivity is critical because failure to perceive a situation as presenting a moral problem makes it difficult, if not impossible, to render a principled judgment and be moved to act in accordance with it.

Leveraging moral mistakes for learning

Oser’s call for learning from negative moral experiences (or ‘moral mistakes’) has been echoed by other researchers who study errors at work (e.g., Harteis & Bauer, 2014; Stevenson & Stigler, 1992). The common thread in the literature is that moral mistakes can be leveraged for changing patterns of thinking and behavior if there is a ‘culture of mistakes’ within the organization or institution. Such a culture defines mistakes as normal and helping people learn from them as critical to furthering personal growth and development and to building cultures in which the negative impacts from errors do not fester and spread.

In making these arguments, Oser and others have (often implicitly) applied Kolb’s (1984) experiential learning theory to the theory of moral development. Kolb’s work suggested that the most powerful learning comes not from reading or hearing about others’ experiences, but through personal experience combined with the opportunity for reflection, conceptualization, and experimentation. This idea of learning from experience is constructivist in nature, positing that people best construct meaning from their own experiences, and that negative experiences are particularly powerful for learning because they cause a dissonance that needs to be reconciled (Bertram Gallant, 2020; Mezirow, 1997).

The present study took place at a university that defines academic misconduct as a mistake that can be leveraged for learning. This focus on a student’s actions as a mistake is meant to emphasize that many students do not intentionally violate standards or ethics but do so because they lack the knowledge or skill needed to achieve with integrity (e.g., Bertram Gallant, 2020; Bertram Gallant & Stephens, 2020; East, 2010; Jamieson & Howard, 2019). When students at this university accept responsibility for such a mistake, or they are held responsible by a hearing board, their learning is facilitated through several different structured opportunities that are assigned according to the type of mistake that occurred.

The Intervention

The Intervention was grounded primarily in neo-Kohlbergian conceptions of moral functioning (Rest et al., 1999), which posits that effective moral functioning involves the integrated use of at least four processes: sensitivity (the ability to perceive and interpret the moral issues or values inherent in a given situation), judgment (the capacity to deliberate or reason through a given dilemma to determine the morally principled course of action), motivation (the desire to prioritize morality above other values and commit to acting in ways consonant with one’s judgment), and character/implementation (a constellation of abilities needed to stay the course and implement the moral course of action). As suggested by the word ‘integrated,’ the four components act as an ‘ensemble of processes’ and the failure to act morally may result from a deficiency in any one of them (Narvaez & Rest, 1995, p. 387).

In keeping with Kohlbergian approaches to moral education, the Intervention used dilemma discussions, role-plays, and other techniques to help students—all of whom had recently been responsible for academic misconduct—learn from their moral mistake. These pedagogical activities took place in-person over several weeks, totaling approximately seven hours of learning. First, students completed a pre-assessment and wrote a reflection about their moral mistake. This was followed by a 20-minute peer-to-peer meeting to discuss the reflection and prepare for the learning experience. Then, the students attended four class meetings (along with 20–50 other students), which included two hours of instruction in ethical decision-making (facilitated by an instructor). Students completed assignments before attending meetings 2–4, which formed the basis of one-and-a-half-hours of peer-facilitated discussions (in groups of 5) about how to recognize and prioritize ethical values in future situations.

Students also received feedback on their assignments during these discussions. For example, one of the assignments asked students to use an ethical decision-making framework to analyze and write about an ethical dilemma they were experiencing, and then share that analysis with their small group. Consistent with Narvaez and Endicott’s (2009) suggestions for enhancing moral or ethical sensitivity, students were instructed to address several questions related to noticing the problem, stating the situation, and identifying the interested parties.¹ After the Intervention, students were asked to complete a post-assessment.

The present investigation

The present quasi-experimental field study sought to test the potential effects of an intervention designed to enhance participants’ moral sensitivity. We chose to focus on moral sensitivity (and not other components of moral functioning) for several reasons. First, while the Intervention was likely to affect the other three components, most of the discussions and assignments centered or depended on perceiving and interpreting academic misconduct as a moral situation.

Second, previous research has found that students who fail to see academic misconduct as a moral situation are more likely to cheat (e.g., McDonald et al., 2014; Schab, 1991; Stephens, 2018). For example, Stephens (2018) found that compared to students who categorized unpermitted collaboration and plagiarism as a ‘personal choice’ or ‘social convention’ (i.e., wrong only because school rules say so), students who perceived those behaviors to be ‘morally wrong’ were significantly less likely to report engagement in them.

Third, and finally, previous research has shown that educational interventions can be effective in increasing moral/ethical sensitivity in numerous disciplines, such as accounting (e.g., Taylor, 2013), dentistry (e.g., Bebeau & Brabeck, 1987), military training (Seiler et al., 2010) and nursing (e.g., Uncu & Güneş, 2021). However, none of these studies (or any we could find) had assessed the effects of an intervention on moral sensitivity delivered in the aftermath of a moral mistake like academic misconduct. Nonetheless, based on this previous research and Oser’s (1996) hypothesis, we believed that completion of the Intervention would be associated with enhanced moral sensitivity. Specifically, we hypothesized that after completing the Intervention participants would:

1. report significantly higher levels of perceptual moral attentiveness (H₁) and reflective moral attentiveness (H₂); and

2. demonstrate significant increases in implicit moral awareness (H₃) and explicit moral awareness (H₄).

Methods

To assess the potential effects of the Intervention on participants’ moral sensitivity, a quasi-experimental (single group, pre-posttest) research design was employed.

Participants

The final sample of participants (N = 798) included 491 (61.5%) males, 306 (38.3%) females, and 1 (0.1%) gender diverse student; 125 (15.7%) Freshmen, 169 (21.2%) Sophomores, 198 (24.8%) Juniors, 179 (22.4%) Seniors, 68 (8.5%) fifth year students, and 59 (7.4%) graduate students; and 356 (44.6%) students who indicated that they had attended secondary school outside the United States.

Procedures

The final sample of participants described above was derived from a larger population (N = 1,134) of students who were responsible for engaging in some form of academic misconduct (e.g., plagiarism, cheating, collusion) and subsequently enrolled in the Intervention. The data comes from the pre- and post-assessment that the university assigned to all students enrolled in the Intervention. Students were required to complete the pre-assessment before starting the Intervention (Time 1; T1 hereafter), and the post-assessment shortly after completion (Time 2; T2 hereafter). Both assessments were the same and administered online (via SurveyMonkey), but the post-assessment was not required. Of the 336 students (29.6%) who were removed from the final sample, 305 (26.9%) were eliminated because they did not complete the T2 assessment and 31 (2.7%) because one or both of their assessments had significant amounts of missing or invalid data.²

Measures

As detailed below, a mix of existing measures were employed to assess changes in participants’ moral attentiveness and awareness. While moral or ethical sensitivity has typically been assessed with self-report questionnaires using Likert-type response scales (e.g., Maxwell et al., 2021), we also wanted to incorporate the use of constructed or open response questions that require participants to demonstrate moral sensitivity. The latter can be more challenging and time-consuming for participants to complete, but they are also less susceptible to social desirability bias. We were especially interested in assessing implicit moral awareness as previous research suggests that many unethical acts take place without the conscious awareness of the actor (Banaji et al., 2003). We decided to use a word-fragment completion task, which has been shown to assess implicit cognitive processes (e.g., Bassili & Smith, 1986). Gino and Bazerman (2009) developed such a task for their study on unethical behavior and we’ve adopted it here to test if participants’ implicit thought processes (as assessed by their choice of words) may have been affected by the Intervention.

Moral attentiveness

An existing measure (adapted from Reynolds, 2008) was used to assess the extent to which participants noticed and thought about the morality of their circumstances or actions. Specifically, participants used a seven-point scale (1 = Strongly disagree to 7 = Strongly agree) to respond to eight items that assessed two types of moral attentiveness (MA): Perceptual MA (4 items; T1 α = .86, T2 α = .86), which included items such as, ‘In a typical day, I face several ethical dilemmas’ and ‘I frequently encounter ethical situations’; and Reflective MA (4 items; T1 α = .81, T2 α = .80), which included items such as, ‘I think about the morality of my actions almost every day’ and ‘I often reflect on the moral aspects of decisions.’

Implicit moral awareness

A measure developed by Gino and Bazerman (2009) was adapted to assess participants’ implicit moral awareness (i.e., the extent to which moral values or principles were more readily activated or accessible than equally valid non-moral words or concepts). Specifically, participants were presented with word fragments and asked to replace missing spaces with letters to form complete words. This word-completion task consisted of seven items, four of which could form words related to morality:

1. C A _ _(CARE as opposed to CAPE, CARP, CART, etc.)

2. _ _ R P

3. G _ E _ T

4. _ _ RAL(MORAL as opposed to MURAL, CORAL, RURAL, etc.)

5. _I_ _ UE(VIRTUE as opposed to TISSUE, BISQUE, RISQUE, etc.)

6. S_ _S_ _

7. E_ _ _C_ _(ETHICAL as opposed to EFFECTS, EFFACED, etc.)

Participants received a score ‘1’ for each of the four moral-related words they spelled, which were subsequently summed to create a single variable (range = 0 to 4). As reported in Gino and Bazerman (2009), the inter-rater reliability coefficient for the original measure was 0.99. In the present study, the reliability was 1.00; that is, the two raters had 100% agreement that the four words being scored ‘1’ (i.e., care, moral, virtue and ethical) were related to the moral domain, while other words being scored ‘0’ (e.g., cape, mural, tissue, and effects) were not related to morality or ethics.

Explicit moral awareness

To assess participants’ moral awareness in a more explicit manner, we used a hypothetical vignette adapted from Stephens and Wangaard (2016). The vignette was designed to present participants with a moral dilemma—a situation wherein an agent (Laura, in this case) has moral reasons for taking two distinct actions, but cannot do both actions (McConnell, 2022). Specifically, participants were asked to read the vignette below and then to ‘describe the moral/ethical values or principles … present in the situation.’

Laura and Her Friend

Laura is really smart and her classmates know it. Her reputation, however, sometimes causes unwelcome attention and difficult choices. Here’s an example: After taking a midterm test, one of Laura’s friends starts grilling her about it. The friend is taking the midterm later that day and she wants Laura to tell her all of the test questions and answers that she can remember. Laura feels torn and doesn’t know what to do.

The authors used both inductive and deductive thematic analysis—the latter rooted in the ‘fundamental values of academic integrity’ (International Center for Academic Integrity [ICAI], 2021) and moral foundations theory (Haidt, 2013)—to create an original coding scheme consisting of eight moral or ethical values (e.g., fairness, honesty, loyalty). Each value was coded (0 = absent or 1 = present) based on its inclusion or expression in the participants’ responses. Inter-rater agreement was assessed by calculating Cohen’s (1960) κ on a random subset of the data (40 responses) that the authors classified independently with results indicating ‘strong’ or ‘almost perfect’ agreement (range of κ = .86 to 1.00; McHugh, 2012). Please see Appendix for details of the coding scheme, including examples and kappa statistics for each code.

Data analyses

Confirmatory factor analysis (CFA) was used to confirm the structure and fit of the two-factor measurement model for moral attentiveness. Based on recommendations by Hu and Bentler (1999) and Ullman and Bentler (2003), normed chi-square values and several other indices were used to determine model fit, where χ²/df < 5.0, CFI > .90, TLI > .90, RMSEA < .08, and SRMR < .08 constituted an ‘acceptable’ fit. To ensure that the latent constructs were measured equivalently across the two time points, multigroup confirmatory factor analysis (MGCFA) was employed to test for measurement invariance. Based on recommendations by Chen (2007), change in CFI, RMSEA, and SRMR values were used to determine the level of invariance achieved: ΔCFI of <−.010 and ΔRMSEA of < .015 for each successive level (i.e., metric, scalar, and residual), and ΔSRMR of <.030 for metric invariance and <.015 for scalar or residual invariance.

Finally, to test study hypotheses, both frequentist (e.g., repeated measures ANOVA and paired-samples z-test of proportions) and Bayesian statistical techniques were employed. With respect to the latter, BF₁₀ scores (i.e., the Bayes factor in favor of the alternative over the null hypothesis) were calculated, where BF₁₀ values >150 = very strong evidence against the null hypothesis, 20–150 = strong evidence, 3–20 = positive evidence, and 1–3 = anecdotal evidence—‘not worth more than a bare mention’ (Kass & Raftery, 1995, p. 777). All frequentist statistical analyses were conducted using SPSS and AMOS (Version 26), and all Bayesian statistical analyses were conducted in JASP (Version 0.16.4).

Results

Results from the CFA and MGCFA are reported first, followed by results pertaining to the hypotheses tested.

Confirmatory factor analyses

CFA was used to test the validity of the two-factor measurement model for moral attentiveness at Time 1 (T1) and Time 2 (T2). The data at both time points offered an acceptable fit to the model: χ²/df = 4.42, CFI = .983, TLI = .974, RMSEA = .049 (90%CI = .040–.059), and SRMR = .050 at T1; and χ²/df = 2.99, CFI = .981, TLI = .971, RMSEA = .051 (90%CI = .037–.064), and SRMR = .051 at T2. Table 1 provides a summary of the factor items, loadings, and alphas at both time points. Except for item 4 on reflective moral attentiveness, all factor loadings were good to excellent (range = .71 to .88). Similarly, the Cronbach’s alpha values for all four factors were also good (range = .80 to .88).

Table 1. Factors, items, and loadings for moral attentiveness by time.

Factor/Item	Loading Time 1	Loading Time 2
Perceptual Moral Attentiveness
1. In a typical day, I face several ethical dilemmas.	.74	.79
2. I regularly face decisions that have significant ethical implications.	.78	.83
3. My life has been filled with one moral predicament after another.	.88	.86
4. I frequently encounter ethical situations.	.71	.75
(Cronbach’s alpha)	(.86)	(.88)
Reflective Moral Attentiveness
1. I regularly think about the ethical implications of my decisions.	.77	.79
2. I think about the morality of my actions almost every day.	.83	.80
3. I often reflect on the moral aspects of my decisions.	.73	.65
4. I like to think about ethics.	.52	.58
(Cronbach’s alpha)	(.81)	(.80)

N = 798. Loadings are standardised beta weights.

MGCFA was conducted to test (progressively) the metric, scalar, and residual invariance of the two-factor measurement model across time. As detailed in Table 2, based on change (Δ) in CFI, RMSEA, and SRMR values, the model demonstrated ‘strict’ invariance (i.e., equivalence of residuals). As only ‘strong’ invariance (i.e., metric and scalar equivalence) is required to allow comparison of latent factor means across time, we proceeded as planned with hypothesis testing.

Table 2. Results from tests of measurement invariance for moral attentiveness based on time.

Model	CFI	RMSEA	SRMR	ΔCFI	Δ RMSEA	Δ SRMR	Decision
Unconstrained	.953	.067	.046	–	–	–	–
Metric Invariance	.952	.063	.048	−.001	−.004	.002	Accept
Scalar Invariance	.949	.063	.053	−.003	−.000	.005	Accept
Residual Invariance	.943	.062	.057	−.006	−.001	.004	Accept

N = 798.

Tests of hypotheses

As hypothesized, results indicated statistically significant increases over time on all four of the measures related to moral sensitivity. As detailed in Table 3, participants reported greater perceptual and reflective moral attentiveness and demonstrated greater implicit and explicit moral awareness. As indicated, all the differences were statistically significant (all p-values < .001); however, given the large samples, even small changes in mean values can achieve statistical significance. Accordingly, partial eta-squared values and Bayes Factors were also examined to determine the magnitude of observed differences and probability that T1 and T2 means were different. With respect to the former, the increases in perceptual moral attentiveness and explicit moral awareness were large in magnitude (η² = .290 and .281, respectively), while the increases in reflective moral attentiveness and implicit moral awareness were small and medium (η² = .025 and .084, respectively).

Table 3. Descriptive statistics for moral awareness and attentiveness by time with results from frequentist and Bayesian repeated measures ANOVA.

		Time 1			Time 2
	n	Mean	SD	95% CI	Mean	SD	95% CI	F-value	Partial ƞ^2	BF10
Moral Attentiveness
Perceptual	798	3.60	1.22	3.52–3.69	4.30	1.22	4.21–4.38	325.47***	.290	2.355 × 10^+34
Reflective	798	4.76	1.08	4.60–4.84	4.87	1.04	4.80–4.94	20.21***	.025	2.80
Moral Awareness
Implicit	798	1.57	1.17	1.49–1.65	1.91	1.23	1.82–1.99	73.20***	.084	6.413 × 10^+13
Explicit	721	1.34	0.85	1.27–1.40	2.04	0.91	1.98–2.11	281.94***	.281	1.037 × 10^+52

CI = Credible Interval. Partial ƞ² values between .01 and .06 constitutes a small effect, .07 to .12 a medium effect, and .13 or greater a large effect. BF10 values >150 = very strong evidence for the hypothesis and values between 1 and 3 = anecdotal evidence for the hypothesis.

***p < .001.

Similarly, the BF₁₀ values for perceptual moral attentiveness and explicit and implicit moral awareness were much larger than 150, indicating ‘very strong evidence’ for the hypothesis that their T1 and T2 means were different. In contrast, BF₁₀ value for reflective moral attentiveness was only 2.80, indicating ‘anecdotal evidence’ for the hypothesis. Finally, examination of the 95% Credible Intervals of the T1 and T2 distributions provides further evidence of the foregoing results—except reflective moral attentiveness, none overlapped. In sum, results of both frequentist and Bayesian statistical analyses provide strong and consistent evidence for three of the four hypotheses; only H₂ related to reflective moral attentiveness produced weak and conflicting evidence.

Increased complexity in explicit moral awareness

As reported above, participants demonstrated greater explicit moral awareness after the intervention as measured by their capacity to identify and name moral values at stake in a hypothetical dilemma involving academic misconduct. However, additional analyses indicated that the observed change was not distributed evenly among the eight values that were coded and scored. As depicted in Figure 1, there were significant increases from T1 to T2 for six of the eight moral values or principles and a significant decrease for one of them (i.e., integrity: ∆ = −12.0%; z = −5.07, p < .001). The only value that did not change significantly over time was care (∆ = −0.4%; z = −0.49, p = .627). Among the values that increased significantly, honesty (∆ = 29.5%; z = 12.95, p < .001) and loyalty (∆ = 30.4%; z = 12.26, p < .001) had the largest gains in terms of percentage points as well as effect size (Cohen’s d’s = 0.55 and 0.51, respectively).

Figure 1. Moral values participants reported as present in Laura’s situation: mean percentages by time.

Error bars represent standard errors. CI = Confidence Interval (Lower Limit, Upper Limit). For Cohen’s d, 0.20 = small effect, 0.50 = medium effect, and 0.80 = large effect.

**p < .01 ***p < .001.

Although not hypothesized, as we coded participants’ responses to ‘Laura and Her Friend’ we noticed not only an increase in the number of moral values being expressed but also an increase in the complexity of their expression. Specifically, two distinct but related types of conceptual complexity appear evident in participants’ responses: a decrease in the generic use of the word ‘integrity’ (e.g., ‘It violates the rule of academic integrity’ 803_T1) and a concomitant increase in the juxtaposition of moral values in the form of a moral dilemma, such as honesty vs. loyalty (e.g., ‘In this situation Laura is faced with a choice between loyalty to her friend and honesty to herself, classmates, and professor’ 745_T2). As illustrated in Figure 2, while instances of ‘integrity vs. loyalty’ did not increase significantly from T1 to T2 (∆ = 1.8%; z = 0.09, p = .369), the increases were significant for both ‘fairness vs. loyalty’ (∆ = 14.4%; z = 7.71, p < .001) and ‘honesty vs. loyalty’ (∆ = 24.8%; z = 12.69, p < .001).

Figure 2. Participants’ perception of Laura’s situation as a moral dilemma by time.

Error bars represent standard errors. CI = Confidence Interval (Lower Limit, Upper Limit). For Cohen’s d, 0.20 = small effect, 0.50 = medium effect, and 0.80 = large effect.

***p < .001.

While the foregoing quantitative data provide clear evidence that participants’ responses, on average, demonstrated an increased complexity, the change is also evident at the individual level. For example, participant 850 wrote only one word (‘Honesty’) at T1 but the following sentence at T2: ‘It’s an ethical issue since Laura is facing a decision between loyalty and honesty’ (850). Similarly, participant 823 wrote one sentence to describe Laura’s situation at T1: ‘It’s a choice between helping a friend and honoring the academic integrity.’ At T2, this participant wrote a lengthy response that not only acknowledged the presence of a dilemma, but demonstrated a clear understanding of several moral values at stake in the decision:

The conflict I see here is friendship vs. honesty/responsibility/fairness. If Laura doesn’t help her friend, their friendship might be undermined. However, giving away the test questions/answers will violate academic integrity … . Laura would be violating her responsibility to not share test material with other students and to report students who cheat in exams. Last but not least, their action is not fair for other students who pay a lot of effort to get a good grade in the class. (823_T2)

In this T2 response, the participant identified that loyalty (friendship) was at odds with (other) moral values such as honesty and fairness. This demonstrates a more sophisticated understanding of academic integrity and a greater awareness that the situation goes beyond a simple right versus wrong to a moral dilemma—Laura has moral reasons for taking two distinct actions but can only choose one.

Discussion

The results from this quasi-experimental field study indicate that it may be possible to develop students’ moral sensitivity in the aftermath of academic misconduct. Consistent with our hypotheses, participants—after completing the four-week Intervention—reported increases in moral attentiveness and demonstrated increases in moral awareness. In particular, the T1 to T2 changes in perceptual moral attentiveness and explicit moral awareness were both large (in effect size as derived from frequentist statistics) and very strong (as evidence against the null hypothesis as determined by Bayesian statistics). Finally, results from additional analysis indicated that participants’ explicit moral awareness not only increased quantitatively (in terms of the number of moral values they perceived) but also qualitatively (in terms of complexity with an enhanced perception of Laura’s situation as a moral dilemma).

Significance

The present study makes three important contributions to the literature related to moral education and academic integrity. First, while previous studies have shown educational interventions to be associated with increases in moral sensitivity in various professions (cf. Bebeau & Brabeck, 1987; Seiler et al., 2010; Taylor, 2013; Uncu & Güneş, 2021), this study is the first do so in the domain of academic integrity. To be clear, other researchers have studied the effects of academic integrity instruction on other outcomes, including self-reported academic misconduct (e.g., Dee & Jacob, 2010; Stephens et al., 2021), but none have focused on or assessed changes in moral sensitivity.

Second, this is the first study to focus on or assess changes in moral sensitivity in the aftermath of academic misconduct, a moral mistake. In fact, to the best of our knowledge, this is the first study to test and provide support for Oser’s (1996) hypothesis that negative moral experience—responded to with opportunities to learn from it—‘leads to higher sensitivity to moral and social problems’ (p. 72). That is, we do not know of any other empirical research that used academic misconduct or any other moral mistake as a teachable moment for enhancing moral sensitivity (or any other component of moral functioning). In providing strong evidence of this possibility, the present study offers an important example—and inspiration, we hope—for other researchers and practitioners.

The third contribution of the present study concerns the enhanced complexity of participants’ responses to ‘Laura and Her Friend’ at T2. Although not hypothesized, we believe the observed shift (among some participants) from a simplistic view of Laura’s situation as ‘right vs. wrong’ to its perception as a moral dilemma (e.g., loyalty vs. honesty) necessitating a difficult choice is an important one. This shift in perception represents not only enhanced cognitive complexity but also entails greater moral responsibility—there is no longer a single moral mandate to follow, but rather a genuine choice to be made. And with that choice, as Isaiah Berlin declared and several of our participants recognized, the potential for ‘irreparable loss’ (e.g., a friendship, in Laura’s case).

Limitations and future directions

As with all studies, the present study is not without its limitations, which of course provide intellectual fodder for possible future research directions. First, and most importantly, the use of a quasi-experimental research design precludes the possibility of making any firm claims that the intervention caused the observed changes. It is possible that other factors (e.g., history and testing effects; see Cook & Campbell, 1979) are responsible (at least in part) for observed increases associated with moral sensitivity. Where possible, future research should employ true experimental designs, such as randomized waitlist control trials, to test more rigorously the associations observed here.

Second, the measures used in the present assessed only a fraction of perceptual and interpretative capacities associated with moral sensitivity. Future research should employ other measures associated with moral sensitivity and perhaps other components of moral functioning. It is possible, for example, that this intervention—or interventions like it—affect other processes associated with moral functioning, such as judgment and motivation. Relatedly, future studies could consider comparing student gains in a more intensive and prolonged training as the ethical development literature suggests that at least eight weeks of instruction is needed to develop moral reasoning skills (Christensen Hughes & Bertram Gallant, 2015).

A third limitation is that our study assessed gains immediately after the Intervention was completed. It is unknown if the effects would persist over time or if students would be able to apply their enhanced moral sensitivity to new situations (hypothetical or real). Similarly, because only some of the Intervention students completed the T2 assessment, our results may be affected by selection bias - students who followed through were more engaged in the learning and therefore demonstrated gains. Future research could address both limitations by including longitudinal studies that assess all the Intervention students over an extended period, ranging from a single six-month follow-up to repeated measures over the remainder of their studies. Such research designs would also allow for more robust testing of the effects of enhancing moral sensitivity on engagement in academic misconduct.

Finally, this study was limited to a single university in the United States. This presents two limitations. First, the students’ actions determined to have been moral mistakes may not have been determined as such in a different institution. And second, the effects of the Intervention may be conditional on the context and the facilitators of the Intervention. Thus, the generalizability of the results remains uncertain. However, the sample was diverse in terms of educational backgrounds, ethnicity, and academic majors, so it seems reasonable to extrapolate that a similarly structured intervention at different institutions could be similarly successful. Nonetheless, we hope more studies of this nature will take place in other contexts with different populations and with different interventions to establish generalizability of the broader concept tested here—that moral mistakes can be leveraged as learning opportunities to enhance moral sensitivity.

Conclusion

In closing, our findings lend empirical support to Oser’s (1996) theoretical concept that moral mistakes can be leveraged for learning, and in particular, the development of moral sensitivity. The Intervention investigated in this study offers only one example of how this might be accomplished. We hope the results associated with its completion will inspire others (institutions and individuals) to take up Oser’s call to create a ‘culture of mistakes’ and, in doing so, turn moral mistakes into teachable moments.

Acknowledgments

We would like to thank the students who participated in the program and Amanda Brovold and Alan Gutierrez for their assistance with data collection and entry.

Disclosure statement

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

Additional information

Notes on contributors

Jason M. Stephens

Jason M. Stephens is an Associate Professor in the School of Learning, Development and Professional Practice at the University of Auckland. His primary research interests include human motivation, moral functioning, cheating behavior, and the promotion of academic integrity during adolescence.

Tricia Bertram Gallant

Tricia Bertram Gallant is the Director of the Academic Integrity Office and Triton Testing Center at the University of California, San Diego. Her primary research interest centers on academic integrity as a teaching and learning issue, which means she explores how pedagogy and assessment design can enhance learning and reduce cheating, as well as how students can learn from the experience of cheating.

Notes

1. Readers who are interested in replicating this Intervention, with or without a corresponding empirical study, can contact the authors for more details.

2. Given the data are sensitive and owned by the University, those wishing access should contact the authors directly. All requests will be considered on a case-by-case basis.

Appendix

Coding Scheme for the “Laura and Her Friend”.

Code	Meaning	Examples	κ
Fairness	The quality or state of being fair; lack of favouritism or bias; equal or reasonable treatment of others.	“Giving the answers to her friend would give said friend an unfair advantage over other students” (802_T1); “If Laura tells her friend the questions in the exam, it will not be fair to other students taking the same exam” (807_T1).	1.00
Honesty	The quality of being honest, free from fraud or deception, legitimate, truthful.	“It is not honest to share the test contents” (1119_T1); “If Laura goes through with the request her friend made, then she would be violating the moral issue of being honest” (893_T1).	.98
Trust	The assured reliance on the character or truth of someone or something.	“Breaking the trust of the Professor” (922_T1); “If Laura tells all of the test questions that she can remember to her friend, then Laura and her friend have the value ‘trust’ between them” (937_T2).	1.00
Responsibility	The quality or state of being responsible or reliable; moral accountability.	“Laura would be violating her responsibility to not share test material with other students” (823_T2); “Laura…feels she has some sort of responsibility to be kind and help her friend” (1037_T2).	1.00
Respect	High or special regard; the quality of being esteemed.	“She’s showing a lack of respect for herself… a lack of respect for all the students” (713_T2); “It is ethical and moral issue because it would be disrespectful to the class, the professor” (1048_T2).	1.00
Care	To feel interest or concern; to be concerned about.	“Laura cares for her friend and wants her to succeed” (775_T1); “Laura will be hurting the people around her” (1083_T2).	.88
Loyalty	The quality or state or an instance of being loyal; faithful to a person, group or cause).	“Laura wants to assist her friend, and not be a ‘bad friend’ or ‘uptight’ by refusing her help” (1069_T1); “Laura, I presume, would want to give her friend the test questions because they are friends and she wants to be loyal to her” (621_T1).	.86
Integrity	Firm adherence to a code of especially moral or artistic values	“The moral/ethical issue is that she knows by academic integrity standards she shouldn’t tell her friends” (723_T1); “Laura needs to select between friendship and integrity” (728_T1).	.94

Note. Meanings adapted from merriam-webster.com. T1 = Time 1, T2 = Time 2. κ = Cohen’s (1960) kappa coefficient of interrater reliability, where 0-.20 = none, .21–.39 = minimal, .40–.59 = weak, .60–.79 = moderate, .80–.90 = strong, and above .90 = almost perfect (McHugh, 2012).