Introducing and unpacking the Aristotelian “standard model” of phronesis.

Given the historical role of Aristotle as the forebear of all contemporary philosophical and psychological work on (practical) wisdom, we believe his theory deserves serious scientific exploration; hence the original question in the title of this article. In this section, we explore the theory in more detail, before later subjecting it to a comprehensive empirical investigation.

In Aristotle’s [3] ethical system, phronesis is nothing less than the lynchpin of a flourishing life, actualizing the virtues and representing good character. According to Aristotelian character developmental theory, young people who have acquired the right moral habits through good upbringing need to gradually develop this intellectual virtue to guide their decision-making. Otherwise, their moral lives will be fragmented, uncritical, and lacking in intrinsic value. Rejecting Plato’s idea of a moral master virtue (justice), which trumps the other virtues in times of conflict, Aristotle proposed this intellectual meta-virtue instead. In that sense, then, Aristotelian phronesis is best understood as deliberative excellence in moral decision-making [4].

It is important to note at this point that the foundational concept of Aristotelian and neo-Aristotelian virtue theories (as well as their educational incarnations as “character education”) [7, 44] is not “character” or even “virtue,” but rather “flourishing.” Flourishing is seen as the “ungrounded grounder” of the good life [33], which the other key capacities are either conducive to or constitutive of. Although flourishing (often under the banner of “objective wellbeing”) has been on psychological agendas for decades, its prominence has been on the rise in recent times, as well as efforts to measure it [45, 46]. Although the immediate aim of phronesis, on an Aristotelian understanding, is excellence in moral decision-making, its ultimate aim is not just “prosocial behavior” (an end in which virtue ethicists tend to be only derivatively interested) but rather its constitutive contribution to an overall flourishing life. The overarching research question motivating the present research project, to which the question in the title refers, is whether phronesis predicts the core components of flourishing.

There are already large theoretical literatures on practical wisdom in general [8] and the Aristotelian conception in particular (see [6]). However, most of those literatures are either exegetical or purely philosophical in orientation, and hence outside of our immediate practical and empirical interests. What matters for present purposes is that in philosophy there has gradually evolved what Miller [39] calls a neo-Aristotelian “standard model of phronesis,” which carries independent interest, whatever one may think of some of Aristotle’s own claims. In the standard neo-Aristotelian model, the task of phronesis is complex [47], and a common suggestion from the theoretical literature is that it has at least three functions. First, phronesis helps us spot situations where the relevant virtue is required and how to execute it. For example, courage is the virtue that is appropriate to situations involving risk. Second, phronesis allows us to integrate different virtues that seem to come into conflict in the same situation, such as being courageously generous. This arbitration can also lead to enacting one virtue that is a higher priority and in unresolvable conflict with a second virtue (e.g., mercy versus justice). Only through phronesis do the virtues become a “package deal” ([8], p. 26). Third, the phronimos (person possessing phronesis) re-evaluates and regulates emotional traits acquired early in life, infusing them with reason and justification. Other scholars have added a fourth function, of “deep understanding” [47] of the human condition, to this mix: an understanding of what constitutes human flourishing as an irreducibly moral activity.

Although it has been suggested that Aristotle’s remarks on phronesis are not always particularly illuminating, especially from a contemporary developmental and educational perspective [37], it does seem possible to derive a general account of phronesis from those texts that emphasize its diverse above-mentioned functions—hence refining and concretizing “the standard model.” The best way to convey the nature of those functions in contemporary psychological language is to say that the construct is made up of various (inter-related) components, and we will hence shift to talking about “components” rather than “functions” in what follows. Therefore, we hypothesize, in line with previous theoretical writings [4, 9, 19], that there are four psychometric components of phronesis, reflecting the four functions they are theoretically tasked with carrying out. The four-componential version of the “standard model,” which we call the Aristotelian Phronesis Model (APM), constitutes a hypothesis derived from Aristotelian theory, a hypothesis tested within Study 1. Moreover, the components do not refer to psycho-moral capacities that are completely independent of one another and can be turned “up” or “down” in isolation; rather, they are inter-related as explained below (see further in [4]). The four APM components are outlined below.

Moral perception.

Phronesis in the APM involves the cognitive discriminatory ability to perceive the ethically salient aspects of a situation and to appreciate these as calling for specific kinds of responses [9]. In the phronimoi, this becomes a moral cognitive excellence in that, after having noted a salient moral feature of a concrete situation calling for a response, they will be able to weigh different considerations and see that, say, courage is required when the risk to one’s life is not overwhelming but the object at stake is extremely valuable; or that honesty is required when one has wronged a friend. Herein, we adopted Moral Perception as a more generally understood term. We could also refer to this component as moral sensitivity, a term often found within standard moral psychology/education literatures [31].

Moral emotion.

Individuals foster their emotional wellbeing through phronesis by coordinating their emotional responses with their understandings of the ethically salient aspects of their situation, their judgment, and their recognition of what is at stake [9]. This is partly because they will have developed habituated virtues, meaning, inter alia, that their emotions have been shaped to align with the motivations and behaviors characteristic of a virtuous person. Additionally, these emotional habits are reinforced and solidified through understanding and reasoning, providing a robust intellectual foundation for their responses. For example, a phronimos might recognize that her appraisal of the situation is problematic, giving rise to an emotional response that is inappropriate to the situation. The emotion-regulative component can then help her adjust her emotion by, for instance, giving herself an inner “talking to” or asking herself questions about what is prompting the ill-fitting emotional response. For this reason, we can also refer to this component, in a more standard Aristotelian way, as infusing emotion with reason [48]. The term we will use hereinafter is Moral Emotion, encapsulating the emotional regulative function emphasized by philosophers, and the positive and negative emotions associated with moral actions.

Moral identity.

The synthesizing work of phronesis operates in conjunction with the agent’s overall understanding of the kinds of things that matter for a flourishing life: a person’s own ethical aims and aspirations, her understanding of what it takes to live and act well, and her need to live up to the standards that shape and are shaped by her understanding and experience of what matters. This amounts to what we call a blueprint of flourishing [9]. A “blueprint” has more similarity to what psychologists call “moral identity” than a full-blown theoretical outline of the good life [49, 50], and we use Moral Identity for simplicity in subsequent sections. Phronetic persons possess a general justifiable conception of the good life (eudaimonia) and adjust their overall reactions to that blueprint, thus furnishing it with motivational force. This does not mean that each ordinary person needs to have the same sophisticated comprehension of the “grand end” of human life as a philosopher might have in order to count as possessing phronesis. Rather than being an “elite sport,” the sort of grasp of a blueprint of the aims of human life informing phronesis is within the grasp of the ordinary individual. It draws upon the person’s standpoint on life as a whole and determines the place that different goods occupy in the larger context.

Moral adjudication.

Assume that we have identified a moral problem correctly as one potentially requiring input from two or more apparently conflicting moral virtues. Let us further assume that we have infused our relevant emotions with reason and that they are not obstructing the decision process. Finally, assume that we have a clear, non-self-deceptive identity of who we want to be—a blueprint of the good life—and an overall motivation to bring our reactions into line with that identity. That leaves just the final component of the four-componential construct: the integrative component—what we could also call its adjudicative component [9] or, in line with standard moral psychology, simply denote as a form of “moral reasoning.” Through this component, an individual integrates different virtue-relevant considerations, via a process of checks and balances, especially in circumstances where different ethically salient considerations, or different kinds of virtues or values, appear to be in conflict and agents need to negotiate dilemmatic space. For the sake of simplicity, we will use the term Moral Adjudication, hereinafter.

The APM is built upon the foundational work of the Jubilee Centre Phronesis Report [51] and the influential framework of Darnell et al. [9, 19]. These earlier studies provided a pivotal theoretical base and important preliminary empirical insights into the structure of phronesis. A particular strength of Darnell et al.’s [19] work was its use of a confirmatory factor analysis (CFA) within a structural equation modeling framework to assess the viability of the four-component APM. By drawing on existing validated measures aligned with the theorized components of phronesis, this study offered an innovative first attempt to operationalize the APM and demonstrated that a second-order model, with four interrelated components feeding into an overarching phronesis construct, could be empirically supported.

However, while this approach provided strong theoretical and empirical groundwork, it also presented certain limitations that future research needed to address. First, by relying on existing measures originally developed for other purposes, the study risked a degree of conceptual misalignment between the measures and the specific nuances of the APM components. Second, the use of CFA, while appropriate for testing a pre-theorized model, meant that the data had less opportunity to inform the factor structure inductively. Third, the resulting measure took approximately 45 minutes to complete (hereinafter referred to as the Long Phronesis Measure, or LPM), raising concerns about its practical feasibility for large-scale studies or applied settings. Fourth, the participants were not representative of the general population, so the data could not speak to phronesis in general. Finally, the item set and scoring instructions were not made fully accessible, which limited the replicability of the findings and constrained the development of a broader empirical program of research on phronesis.

These strengths and limitations highlight both the significance of the Darnell et al. [19] study and the need for further work to build upon its contributions. Future research required measures that were specifically tailored to reflect the APM framework, an approach that would balance theoretical precision with practical utility. Additionally, an inductive exploration of the dimensions of phronesis—through methods such as exploratory factor analysis (EFA)—would allow the data itself to reveal the structure of the construct, providing a stronger empirical foundation for subsequent confirmatory analyses. Addressing these issues is essential to advance both the theoretical understanding of phronesis and its measurement in a way that supports scalable, replicable, and substantive empirical research.

Participants.

Participants for Studies 1a-1c were recruited online using Prolific Academic. Studies have shown that Prolific provides high quality data with high levels of participant comprehension and attention, and low levels of dishonesty [53]. Participants in Studies 1a and 1b were compensated for their time at a rate of £9/hr for one hour of their time, while Study 1c participants were compensated at the same rate for 20 minutes of their time. All data for this project (including subsequent studies) were collected between August 2023 and February 2024.

For Study 1a, after excluding those who elected to withdraw, did not have a valid Prolific ID, or had less than 95% survey completion, the final sample comprised 1998 participants. The sample was representative based on age, sex, and ethnicity, reflecting UK census data (see Table 1). For Study 1b, we recruited two new representative samples, one from the UK (see Table 2 for demographics) and a second from the US (see Table 3). The UK sample included 1,000 participants from Prolific Academic, with 997 retained after removing participants who wished to withdraw. Likewise, we recruited 1000 participants from the US, with 988 retained after removing participants who wished to withdraw. For Study 1c, we recruited 300 participants. Participants were only eligible to take part if they were already part of Study 1b’s UK-representative sample. We gained Study 1c participants’ permission to link their Study 1b data with their Study 1c responses, after which point their responses would be fully anonymized. After removing participants with incomplete data, the final number of participants for Study 1c was 295. The mean age of participants recruited was 45.73 (SD = 13.55), with 137 females, 156 males, and two identifying as “other” or “prefer not to say”. For ethnicity, 254 participants identified as “White”, 14 as “Black”, 11 as “Central Asian”, eight as “East Asian”, seven as “Mixed/Multiple ethnic groups”, and one as “Prefer not to say”.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Demographics of the exploratory UK sample compared to demographic data from the Office for National Statistics (ONS) to evidence representativeness.

https://doi.org/10.1371/journal.pone.0317842.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Demographics of the confirmatory UK sample compared to demographic data from the Office for National Statistics (ONS) to evidence representativeness.

https://doi.org/10.1371/journal.pone.0317842.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Demographics of the confirmatory US sample compared to demographic data from the US Census Bureau’s American Community Survey 2022 to evidence representativeness.

https://doi.org/10.1371/journal.pone.0317842.t003

Measures and item development.

For the SPM, we designed a questionnaire comprising 189 items that aimed to capture various theorized components of practical wisdom. These components were identified based on an extensive review of the literature, some of which has been summarized above. Following this, the first author (a psychologist experienced in psychometric measure development) manually generated an array of possible item formats and example items. These examples, along with the definitions of the constructs they purported to measure, were entered into a Large Language Model (LLM), Open AI’s Chat GPT 4.0, where the first author generated a wide array of items of the same types. The first author suggested refinements to item exemplars where he deemed necessary (e.g., if the item array did not appear to be diverse enough to capture how someone generally behaves, across contexts). Once this process was complete, the item list was shared with the rest of the research team (the second author is an Emeritus Professor of Moral Psychology with expertise in measurement, and the third author is a Professor of Philosophy and originally identified the four theorized components of phronesis we proposed to study) for critique and feedback. If items were identified as problematic, the first author manually created refined versions of these items before generating new item variations. This process was repeated until consensus was reached that we had a list of items that (i) reflected the theorized components of phronesis under study, (ii) were diverse in that they could, in principle, capture how someone generally behaves, and (iii) sufficiently succinct to allow for rapid test administration. The initial questionnaire was designed to be administered online and took approximately 30–40 minutes to complete.

In the item development process, we recognized that some aspects of phronesis could be objectively tested. For example, the ability to discern the moral relevance of a given scenario can have a defined “correct” answer. On the other hand, certain facets of the construct are inherently subjective and best captured through self-report measures. An example of this would be the experience of moral emotions, which is intrinsically subjective and thus most appropriately assessed through self-reporting. A detailed overview of these items is provided in the subsequent sections. Before proceeding further, however, it must be acknowledged that creating items that home in on all the four components (and possible sub-components) of a phronesis construct is a herculean task, not least because of the lack of precedents, apart from our own previous work [9, 18]. The overarching question is always, first, whether we have identified enough sub-functions to be sufficiently reflective of each broader function of the relevant component; and second, whether we have created enough distinct observable variables (survey items) to be sufficiently reflective of those sub-functions. Working on the assumption that “perfect is the enemy of the good,” we embarked on this task seeking to strike a balance between achieving wide construct coverage and having a measure that could be completed relatively quickly and easily.

Moral perception. The ability to accurately perceive morally relevant elements within various scenarios is not a matter of subjective interpretation. Therefore, we incorporated two distinct types of objective assessments to evaluate moral perception. The first test focused on the recognition of whether a situation itself holds any moral relevance. The second test aims to identify the specific virtues that are at stake within those morally relevant situations. These objective tests serve as an alternative to self-report measures for this particular aspect of the broader phronesis construct. The items included under this Moral Perception category were chosen based on the same principles as those in the earlier Phronesis Project [19, 51]. However, the items in this study were designed to be more straightforward to score, as they did not include any open-ended questions. It should be noticed here that the common core in both tests is an assessment of how well the individual homes in on the moral dimension within a complex social situation. While this is not the same as identifying what would be relevant to a fully phronetic person, in order to develop the sensitivity of a phronimos, the moral learner needs to be ablein the first place—to spot and categorize situations of potential moral and characterological relevance. So, the importance here lies in what the respondents identify in the situation, not their assessment of what a phronimos would identify; hence the use of the first-person rather than the third-person within these item sets. Drawing on ideas from Bebeau and colleagues [55], our items focus on identifying and labelling “the moral” and being able to discount other issues. Our items economically attend to the central features of previous moral perception measures, and within our model the component they illuminate behaves much like other treatments of the construct [91].

We initially formulated 20 items aimed at assessing whether participants could discern if a decision in a given scenario would have implications for their character. Participants were guided by the following instruction: “In the upcoming section, you will encounter various scenarios, each requiring a decision. Your task is to identify which scenarios involve decisions that could influence your character. Please ponder the moral or ethical ramifications these situations could have on the individual involved.” Subsequently, participants were presented with scenarios such as: “You have found out a colleague is claiming your work as their own, but confronting them could create team tension” (character-impacting based on a virtue ethical framework) or “You are an avid reader and must choose the next book to read from a stack of equally enticing options” (not character-impacting based on a virtue ethical framework). Participants had two response options: “What I decide to do in this scenario does not affect my character” and “What I decide to do in this scenario affects my character” (note that “affects” was deliberately chosen over “demonstrates”, as in theory, who we are is an aggregate of what we do, in the first instance). Responses were scored as either correct or incorrect.

The subsequent set of items for assessing moral perception consisted of 15 moral dilemmas. In these dilemmas, participants were tasked with correctly identifying the virtues implicated in each scenario. The instruction provided was: “In this section, you will encounter various scenarios along with four character traits that may or may not be pertinent to the situation. Your responsibility is to select the two traits you believe are most relevant.” For instance, participants were presented with a scenario like: “You discover a wallet on the ground containing a substantial amount of money and the owner’s identification. You must decide what action to take. Which of the following traits are most relevant to your decision? (Select two answers from the following options).” Listed below the scenario were two virtues relevant from a virtue ethical standpoint (e.g., honesty and practicality) and two less relevant virtues (e.g., humor and resilience). Participants could select two answers, resulting in scores ranging from 0–2 for each question. We relied on a broad set of virtues (qua positive traits of character) here that could be deemed characterologically relevant, rather than a more narrowly construed set of standard moral virtues.

Regarding the second set of questions, it could be argued that focusing on situations eliciting more than one virtue (1) illicitly limits moral perception to the identification of dilemma-like situations, eschewing single-virtue-eliciting situations, and (2) conflates moral perception with moral adjudication (see below). However, (1) the first set of items homes in on characterologically relevant situations involving (potentially) a single virtue, and (2) the task here is not to adjudicate between or integrate virtues but simply to identify which virtues are at stake in the situation [54].

Moral identity. The issue of whether a participant possesses a conceptual framework for their ideal moral self is best evaluated subjectively. Consequently, for this component, we employed a conventional self-report survey methodology. Participants were presented with statements pertinent to their moral identity. In this case, there was no shortage of previous measures of moral identity, although those have not been related specifically to phronesis, and we drew upon many of those for enlightenment [55]. The items generated aimed to emphasize the relationship between moral identity and moral decisions more concretely than found in other popular measures. For example, one of our items was “When faced with challenging situations I ask myself what a good person would do”. In contrast, popular measures like the Moral Identity Scale [56] include items that are less focused on decision making, such as “It would make me feel good to be a person who has these characteristics”. To be sure, the fact that a decision will (or will not) affect a person’s character is only one way that something could be morally relevant. An alternative approach would be to provide a scenario in which respondents must decide what to do both before and after some event (either morally relevant or not relevant) is introduced. However, as we are interested in moral identity and moral decision-making on a virtue ethical understanding—but not for example a utilitarian one of good moral identity simply as a generator of “prosocial” responses—we opted for the former understanding. Responses were collected using a five-point Likert scale, ranging from “Strongly Disagree” to “Strongly Agree.” Given that we used a single item format for this hypothesized component of phronesis, we initially generated an extensive set of items, totaling 25. These items were modeled after established moral identity questionnaires, maintaining consistency with the format used in previous research, including the earlier Phronesis Project.

Moral emotion. Moral emotion is also subjective, making self-report measures a suitable method for capturing this component. Additionally, moral emotion is theoretically multi-faceted, encompassing the emotions one experiences when acting morally or immorally, as well as one’s ability to regulate these emotions. Therefore, we incorporated multiple item formats to adequately assess this aspect of phronesis. In this way, we improved upon previous measures of moral emotion in the context of phronesis measurement, which focused more so on empathy and perspective taking [57].

Initially, we developed items to gauge the emotional responses individuals might experience when acting either morally or immorally in various scenarios. Participants received the following instruction: “Below are different scenarios where you decide to take specific actions. If you were to take these actions, how would you feel about yourself?” This was followed by a set of 20 statements, half of which depicted morally upright actions (e.g., “A stranger drops a £100 note without noticing. You pick it up and return it to them”), and the other half illustrated moral failings (e.g., “You exaggerate an issue at work to harm a colleague’s professional reputation”). Participants indicated their emotional response on a five-point Likert scale, ranging from “Extremely Bad” to “Extremely Good.”

While emotional regulation has both subjective (e.g., personal emotional experience) and objective (e.g., observable externalizing behavior) elements, we opted for self-report measures. This decision was made because it would be neither practical nor ethical to objectively assess emotional regulation in stress-inducing situations. Participants were guided by the following instruction: “Below, you will encounter various scenarios that may elicit an emotional response. For each situation, please reflect on how you would typically react and rate your ability to manage your emotions. By ‘manage,’ we mean your capacity to prevent your emotions from overwhelming you and to maintain your composure.” Participants then responded to 20 scenarios, some involving moral transgressions against them (e.g., “A stranger is rude to you in a public place for no apparent reason”) and others involving everyday frustrations (e.g., “You accidentally spill a drink on your clothes just as you are about to leave the house”). The emotional responses that might be regulated in these scenarios ranged from jealousy/envy (e.g., scenarios involving recognition given to someone else for your achievement.), irritation/annoyance (e.g., scenarios involving minor inconveniences, such as someone pushing in a queue or a neighbor playing loud music), anxiety (e.g., waiting for a bus that’s late or facing a cancelled flight), hurt/resentment (e.g., harsh criticism for a small mistake or a co-worker making a joke at your expense). In these scenarios, we avoided specifying a “correct” emotional response, as we were not concerned here with adjudicative capacities, and instead aimed to draw on participants’ self-knowledge about their ability not to make an obviously “incorrect” decision based on emotional impulse. Responses were collected on a five-point Likert scale, ranging from “Very Poor” to “Very Good.” We contend that the items selected under “Emotional Regulation” offer a more nuanced reflection of the construct in question compared to the generic measures of empathy used in the previous Phronesis Project [19, 51]. Specifically, these items target the regulatory aspect of phronesis more directly, rather than focusing solely on the general capacity to experience emotions.

Moral adjudication. Moral adjudication entails the integrative process of arriving at a “correct” moral decision through thoughtful deliberation. This concept comprises two elements: firstly, the selection of the correct moral choice based on some criterion, and secondly, the methodology employed in making that choice.

For assessing “correct” moral choices, participants were instructed as follows: “In the upcoming section, you will encounter a series of scenarios. Each scenario presents two primary considerations that represent different potential responses. Your task is to decide how you would balance these considerations if you were to act in these scenarios. You will have seven boxes to choose from. The first and seventh boxes contain the two main considerations. Selecting either of these boxes indicates that you would focus solely on that consideration, disregarding the other. The intermediate boxes signify varying degrees of balance between the two considerations. Although real-life situations may involve more than two considerations, for the purpose of this exercise, please make your decision based on the two presented.” Participants then responded to 22 items, each accompanied by a 7-point scale. Scoring was conducted in three distinct ways: The first scoring method considered “4” as the “correct” answer, representing a balanced consideration of self and others, in alignment with virtue ethical assumptions. The second method compared the average flourishing scores for each scale point. The point with the highest average was deemed “correct,” and participants were scored based on their distance from this point. This approach allowed for situational variability and was grounded in Aristotelian virtue ethics. The third method simply scored answers on a 1–7 scale, with higher scores representing greater prosociality.

The most practical way to assess moral deliberation was to ask participants to self-report on how they gather and verify information relevant to both general and moral decision-making (e.g., “I evaluate the reliability and credibility of the information sources before making a judgment”). They responded to 23 statements on a five-point Likert scale, ranging from “Strongly Disagree” to “Strongly Agree.”

To assess the extent to which people make an effort to integrate different phronesis components in decision-making, participants were presented with 17 statements and asked to agree or disagree on a five-point Likert scale (Strongly disagree–Strongly agree). For example, the statement “In making a decision, I consider my thoughts, feelings, and the situation at hand” gauges the application of moral emotions in a context-sensitive manner.

The final way we attempted to measure moral adjudication was with a dilemma taken from the Adolescent Intermediate Concept Measure (AD-ICM). The AD-ICM assesses the moral thought processes of young people, in particular, focusing on their transition from self-centered to conventional thinking from a neo-Kohlbergian perspective [58]. Participants use a 5-point scale to rate the morality of potential actions and underlying reasons for those actions that a main character in a story in which a moral dilemma is presented could make. Scoring is based on how closely participants’ choices align with expert judgments. Although the AD-ICM was included as it was part of our previous Long Phronesis Measure (LPM, see [19]; note that the LPM included two moral dilemmas, rather than one), it was simplified for scoring purposes through a Microsoft Excel file with embedded macros developed by the second author, in line with our aim to produce a practical and easy-to-score measure.

Flourishing. The Wellbeing Assessment (WBA; [59]) is an instrument designed to gauge comprehensive wellbeing. It is rooted in a theoretical framework that views human flourishing as a state where all aspects of life are positive. This conceptualization aligns with the World Health Organization’s holistic definition of health, which includes mental, physical, and overall wellbeing, and is also reasonably close to a neo-Aristotelian conception of flourishing [51]. The WBA assesses six domains: emotional health, physical health, meaning and purpose, character strengths, social bonds, and financial stability. Although self-reported, the WBA’s sub-scales have demonstrated predictive validity for more objective criteria. For example, the “physical health” self-report subscale has been shown to predict medical diagnoses and insurance claim data during the validation study [59]. In accordance with neo-Aristotelian virtue theory, we deemed it essential to link the adjudicative function of phronesis directly to flourishing and to explore participants’ deliberative strategies. The items selected under “Moral Adjudication” were more diverse than those in the earlier Phronesis Project, which solely used the AD-ICM.

Procedure.

Data collection for Study 1a involved all 189 items generated for the initial practical wisdom measure’s dimension reduction phase. For Studies 1b and 1c, participants completed only the abbreviated and final 107-item practical wisdom measure following dimension reduction in Study 1a. All participants in all studies completed the WBA. Study 1b participants completed the additional measures for purposes of establishing criterion validity that were not used; this is detailed in Study 3.

Data collection was entirely conducted online. Participants filled out the relevant array of questionnaires for each study within a single session, lasting up to one hour for Studies 1a (large practical wisdom item array and the WBA) and 1b (final practical wisdom measure, the WBA, and criterion validity measures), and 20 minutes for Study 1c (reduced practical wisdom measure and the WBA only). The sequence of the questionnaires was consistent for all participants. However, within each questionnaire, the questions were presented in a randomized order, with the exception of the AD-ICM items in Study 1a, which necessitated sequential completion. Prior to beginning the questionnaires for each study, informed consent was secured from all participants.

Ethics.

This study and other studies hereinafter received ethical approval by the University of Birmingham’s institutional review board in June 2023 (ERN_1043-Jun2023), with a subsequent amendment approved in December 2023 (ERN_1043-Dec2023). Informed consent was sought for all participants for this and subsequent studies: Participants were presented with an information page, and electronically signed a consent page if they wished to proceed with the study. No participant deception was involved. All surveys were completed online, with participants opting in of their own volition after seeing the study advertisement.

Analytical strategy.

Study 1a. We initiated our analysis with EFA to empirically determine the number of factors represented by the 189 items, without imposing our preconceived theoretical structure (APM) on the data. To do this, we employed the psych [60] and gparotation [61] packages within R Studio [62]. Carpenter’s [63] recommendation for the participant:item ratio is 10:1, which we exceeded. The Kaiser-Meyer-Olkin (KMO) measure stood at .94, confirming adequate sampling for the EFA. Additionally, Bartlett’s test was conducted, and a clause was incorporated into the code to halt the EFA if the test value was less than or equal to .05. Depending on the data’s skewness and kurtosis, we employed either Principal Axis Factoring (if skewness was > 2 or kurtosis was > 7) or Maximum Likelihood as the extraction method. The number of factors to retain was ascertained through parallel analysis. Finally, the issue of whether the factors should be correlated was not immediately evident to us. This uncertainty was particularly relevant given the likelihood of a multi-factor solution, a scenario amplified by the large sample size and extensive item pool. To address this, we employed the Promax rotation method. Promax initiates with an orthogonal Varimax rotation, followed by an axis “tilting” to permit obliqueness, rather than pursuing an oblique solution directly as an Oblimin rotation would do. The computational simplicity and speed of Promax make it particularly advantageous when managing a large number of factors. Loadings below .4 for each of these factors were suppressed.

Study 1b. Items that were kept from Study 1a underwent a CFA using the lavaan [64] R package to test the anticipated factor structure within a different yet analogous sample. We used a Diagonally-Weighted Least Squares estimator as our items included non-normal ordinal responses, and robust standard errors to provide a more accurate estimation under these conditions. Consistent with the previous analysis, this CFA was adequately powered, maintaining more than ten participants for each item in the model. Aside from this, we conducted measurement invariance tests in R to understand whether configural, metric, and scalar invariance would be found across the UK vs US data.

Study 1c. We sought to establish test-retest reliability by correlating their scores on the SPM at the time of Study 1c with their scores approximately two months later. This would help us to understand whether scores were stable over time.

Study 1a: Dimension reduction using EFA.

Overall, parallel analysis led to the extraction of 17 factors (RMSEA = .02, TLI = .90; χ²[10716] = 107.16, p < .001). Among these, the initial 14 factors possessed enough items with loadings exceeding .4 that would also allow for the computation of an internal reliability coefficient (see Fig 1). A subset of ten factors demonstrated an acceptable level of internal reliability, with Cronbach’s alpha values surpassing .7. Consequently, these ten factors (107 items, taking 15–20 minutes to complete) were retained for further analyses without further modification (see Table 4). These ten factors explained 30% of the total variance across the total number of items. We hypothesized that these ten factors would load onto the four theoretically derived functions of phronesis as superordinate factors, subject to confirmation in CFA.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Scree plot for the parallel analysis.

https://doi.org/10.1371/journal.pone.0317842.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Eigenvalues, internal reliability, and hypothesized function of the ten-factor phronesis model.

https://doi.org/10.1371/journal.pone.0317842.t004

In naming these ten factors, some item sets contained a mix of general factor items and items that are more directly morally salient. For example, one factor contained items like “I try to consider how my decisions today will reflect on the person I aspire to be in the future”, not specifically mentioning moral aspirations. However, other items in that item set such as “I set personal targets that involve improving my ethical awareness and character” were more directly morally relevant. The fact that all aspirations entail values and conceptions of the good notwithstanding, we named factors based on the aggregate meaning of the items constitutive of that factor in the context of the overall measure (e.g., in the case of the two items mentioned above, these were both aspects of “Aspired Moral Identity”). Later in the measure development process, tests of criterion validity could then be used to further establish whether each factor predicts what it should, given how we have named it (see Study 3).

Study 1b: Testing the identified dimensions in two new samples using CFA.

Sample 1 (UK). Each of the ten factors were once again found to be internally reliable, with Cronbach’s Alphas ranging from .72 to .92. We then used a CFA to assess whether the data would fit the hypothesized ten-factor model. The model converged successfully and we observed excellent model fit (CFI = .97, TLI = .97; RMSEA = .04, SRMR = .06; χ²[5414] = 13178.35, p < .001). Next, we used a bifactor structural equation model (SEM) to test whether the ten empirically identified factors could be considered sub-factors of the four theoretical components (Moral Identity, Moral Emotion, Moral Perception, and Moral Adjudication) from which the items were originally derived, and in turn, whether those components loaded onto a superordinate Phronesis factor (in what follows, Phronesis is capitalized when it refers specifically to the factor in question rather than to the general philosophical concept). The model did not converge after 2,436 iterations, and so reliable estimates for our hypothesized bifactor model in a UK sample could not be produced. We reran this model, this time allowing the ten factors to covary within the hypothesized four factors (e.g., Moral Deliberation allowed to covary with Moral Integration as they are both theoretically aspects of Moral Adjudication), but this model did not converge after 3,749 iterations. The failure to converge suggested potential model misspecification or complexity that the data could not adequately support, limiting our ability to confirm the hypothesized structure. The factor covariances are presented in Table 5 (upper).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Factor covariance matrices for US and UK samples.

https://doi.org/10.1371/journal.pone.0317842.t005

Sample 2 (US). Consistent with the previous analysis, this CFA was adequately powered, maintaining more than ten participants for each item in the model. We also tested for configural, metric, and scalar invariance. Once again, internal reliability was strong across the ten sub-factors according to Cronbach’s Alpha values which ranged from .72-.92. The model fit for the ten-factor solution in the US sample was also excellent (CFI = .97, TLI = .97; RMSEA = .04, SRMR = .07; χ²[5414] = 15092.52, p < .001). A bifactor SEM was then fitted to the data with the ten factors loading onto the four theorized components of phronesis, and they in turn onto an overarching Phronesis factor. The model did not fully converge. The factor covariances are presented in Table 5 (lower).

Measurement invariance. We tested whether the measurement models differed across the US vs UK samples for our best fitting models (i.e., the ten latent factors derived empirically through EFA). The initial step in measurement invariance testing, configural invariance, establishes that the factor structure (i.e., the number of factors and their pattern of loadings) is consistent across groups. This configural model served as our baseline for comparison. No constraints are applied to factor loadings or intercepts at this stage. The configural model demonstrated acceptable fit across the two samples (CFI = .96, TLI = 0.96, RMSEA = .03, SRMR = .05), indicating that the factor structure was consistent between the US and UK samples. Next, we tested for metric invariance by constraining the factor loadings to be equal across groups. The chi-squared difference test between the configural and metric models was significant, χ²(96) = 589.57, p < .001, suggesting a change in fit. This change implies that some factor loadings may differ across groups. Subsequently, scalar invariance was assessed by additionally constraining the intercepts to be equal across groups. The chi-squared difference test between the metric and scalar models was also significant, χ²(96) = 292.48, p < .001, indicating that intercepts might also vary across groups. Finally, residual invariance was tested by further constraining the residual variances to be equal across groups. The chi-squared difference test between the scalar and residual models was significant, χ²(106) = 330.88, p < .001, suggesting that the residual variances also differ across the US and UK samples.

Study 1c: Test-retest reliability.

Internal reliabilities for the retested sample were acceptable (.75-.93). The model fit for a ten-factor structure was also excellent (CFI = .99, TLI = .99; RMSEA = .01, SRMR = .06; χ²[5414] = 5482.13, p = .255). Test-retest reliability was established by correlating scores on each of the ten phronesis components with scores two months later. The results are presented in Table 6.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Test-retest reliability of the short phronesis measure in a UK-based sample.

https://doi.org/10.1371/journal.pone.0317842.t006

Participants and design.

In this exploratory study, we aimed to test the network structure of Phronesis (i) in general (USA and UK combined), and (ii) exploring US versus UK differences, using cross-sectional data from Study 1b.

Analysis plan.

We employed network analysis [69] for this study. The analysis hinged on constructing and examining correlation-based networks using qgraph [70] and igraph [71] in R. This approach allowed us to map out how various moral and psychological components, such as Moral Deliberation, Identity Aspirations, and Emotional Regulation, are interconnected. For each analysis we computed key centrality measures: Betweenness, Closeness, Strength (or Degree), and Expected Influence. These are defined as follows:

Betweenness. This is a measure of centrality in a network, reflecting the extent to which a node (i.e., one of the ten phronesis factors, in this case) lies on the shortest path between other nodes. Nodes with high betweenness centrality can be seen as important intermediaries or “bridges” in the network, facilitating communication or interaction between different parts of the network.

Closeness. Closeness centrality measures how close a node is to all other nodes in the network, based on the shortest paths that connect them. A node with high closeness centrality can quickly interact with all others, indicating it has a central position in the network’s overall structure.

Strength. In the context of weighted networks, the strength of a node is the sum of the weights of the edges connected to it. For psychological networks such as these, this can be interpreted as the overall level of direct influence a node has within the network, taking into account the strength of its connections to other nodes.

Expected influence. Expected influence is a measure adapted from strength for networks that include both positive and negative edge weights. It sums up the edge weights while considering the sign of each edge, providing a measure of a node’s total positive or negative influence on the network. A high positive expected influence indicates a node’s strong positive impact on connected nodes, whereas a high negative value suggests a strong negative impact.

All network analyses used the Extended Bayesian Information Criterion (EBIC) using the glasso [72] and bootnet [73] packages to penalize for model complexity. The EBIC helps in selecting a model that is parsimonious yet sufficiently explanatory, thereby preventing overfitting and ensuring the model’s generalizability. This EBICglasso method performs thresholding by penalizing small partial correlations towards zero. This results in a network where only statistically robust and meaningful connections are retained.

Overall.

In Table 7, we present the measures of centrality for each component within the phronesis network, with the network’s overall structure depicted in Fig 2. These centrality statistics help us understand the relative importance or influence of each component within the network. For instance, Betweenness highlights components that act as central connectors or bridges. Moral Self-Relevance, with a high Betweenness score of 26, serves as a key link between otherwise disconnected parts of the network, whereas components like Emotional Regulation and Virtue Identification, both with Betweenness scores of 0, indicate a more isolated position. Closeness reveals how central a component is by measuring its proximity to others in the network. In this context, Moral Self-Relevance (Closeness: 0.0137) is more centrally located compared to Virtue Identification (Closeness: 0.0067), meaning it has a broader reach within the network. Strength reflects the weight of a component’s direct connections. For example, Moral Self-Relevance has the highest Strength score (1.212), indicating strong and numerous connections, while Emotional Regulation and Virtue Identification have weaker connections (Strength: 0.351). Finally, Expected Influence considers the potential impact of a component, accounting for both the strength and direction of its connections. Identity Aspirations and Moral Self-Relevance (Expected Influence: 0.929 and 0.928, respectively) have significant potential to shape the network, whereas Emotional Regulation (Expected Influence: 0.138) plays a more limited role. The range of these statistics—from highly positive (e.g., 1.212 in Moral Self-Relevance’s Strength) to positive but less influential (e.g., 0.138 in Emotional Regulation’s Expected Influence)—helps delineate the spectrum of influence within the phronesis network, from central and pivotal to isolated or less impactful components.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. An overall phronesis network model.

Note. EBIC penalization applied.

https://doi.org/10.1371/journal.pone.0317842.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Centrality statistics for the overall phronesis network model.

https://doi.org/10.1371/journal.pone.0317842.t007

Notably, five components emerge as particularly central within this network: Negative Moral Emotion, Moral Deliberation, Aspired Moral Identity, Moral Self Relevance, and Moral Integration. These central components, highlighted by the bold paths in Fig 2 (blue indicates positive relationships, red indicates negative relationships), play pivotal roles in the network, indicating that their influence is more coordinated and foundational to the overall network conceptualization of phronesis. The centrality of these components signifies not only their individual importance but also their collective contribution to the network’s coherence and functionality. This distinction between central and less-central factors is pivotal, as it underscores the varying degrees of influence different components exert within the phronesis network. Understanding these dynamics offers insights into how individual components interconnect to form a relatively comprehensive model of moral and ethical reasoning.

US vs UK.

Next, we explored whether the network structure looked any different when separated into UK vs US samples. We found that the same network structure broadly held independent of the sample, with the same four components being central overall. The Moral Integration variable appeared to be less central to the Phronesis network in the US compared to the UK, however. Network centrality measures are presented in Table 8, and network plots are presented in Fig 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Network plots for the US vs UK phronesis network models.

Note. EBIC Penalization applied.

https://doi.org/10.1371/journal.pone.0317842.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 8. Centrality statistics for the US vs UK phronesis network models.

https://doi.org/10.1371/journal.pone.0317842.t008

We then compared US vs UK phronesis networks using the NetworkComparisonTest package [74] in R. A network invariance test, encompassing the pattern of connections and the strength of these connections between nodes, suggested a non-significant difference between US and UK network structures (M = .12, p = .102). We also conducted a global strength invariance test. We found that the overall connectivity of the phronesis network was not different in the UK (3.05) compared to the US (3.85); S = 0.80, p = .094, consistent with covariance matrices observed in Study 1b.

Participants.

The two samples first described in Study 1b were included here. As described earlier, one sample was collected in the UK and a second in the USA. The UK sample included 1,000 participants with 997 retained after removing participants who wished to withdraw. Sample 2, collected in the US, included 1,000 participants, with 988 retained after removing participants who wished to withdraw. Furthermore, we tested the degree to which phronesis predicted flourishing over a two-month period with the 300-person sub-sample from Study 1c. The aggregated demographics for the two samples are represented in Table 9 below:

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 9. Demographics for the combined UK/US sample.

https://doi.org/10.1371/journal.pone.0317842.t009

Procedure.

Following an information page, participants were requested to give their consent to participation. Participants in both samples received a battery of assessments (details below). A subset of 300 UK participants were measured two months later using the SPM and a flourishing measure (see below). After completing the questionnaires, participants were given a debriefing and expression of gratitude for their involvement.

Measures.

Phronesis. The full set of phronesis measurement items described in Study 1b was included in these analyses. This measure was administered to both the UK and US samples.

Flourishing. We also included in the analysis the WBA [59]. As described in Study 1a, this instrument is designed to gauge comprehensive wellbeing following the World Health Organization’s holistic definition of health, which includes mental, physical, and overall wellbeing. Specifically, the WBA assesses six domains: emotional health, physical health, meaning and purpose, character strengths, social connectedness, and financial stability. This stands as a reasonable proxy for Aristotle’s notion of flourishing updated for the current era [46, 76]. This measure was administered to both the UK and US samples.

HEXACO-PI-R. Both US and UK participants completed the HEXACO-PI-R [77] to measure the Big Six personality traits, and their various sub-factors:

1. Honesty-Humility: Measures levels of sincerity, fairness, greed-avoidance, and modesty.
2. Emotionality: Captures attributes such as fearfulness, anxiety, dependence, and sentimentality.
3. Extraversion: Focuses on social self-esteem, social boldness, sociability, and liveliness.
4. Agreeableness: Incorporates qualities like forgiveness, gentleness, flexibility, and patience.
5. Conscientiousness: Evaluates organization, diligence, perfectionism, and prudence.
6. Openness to Experience: Explores facets like aesthetic appreciation, inquisitiveness, creativity, and unconventionality.

This measure was administered to both the UK and US samples.

The dark tetrad. We measured the four “dark” personality traits (psychopathy, sadism, Machiavellianism, and narcissism) using the SD4 [78] in both US and UK participants. This was a 28-item measure to which participants responded to statements about themselves on a Likert scale from Strongly disagree–Strongly agree. This measure was administered to both the UK and US samples.

The moral foundations questionnaire. The MFQ2 was used to assess Moral Foundations, qua Moral Foundations Theory (MFT; see [36]). It measures the core moral dimensions according to MFT, namely, Care, Authority, Loyalty, Purity, Proportionality, and Equality. This was administered to the UK-representative sample only.

Propensity to morally disengage. The PMDS [79] is a 16-item measure aims to gauge the extent to which participants exhibited a willingness to engage in immoral actions. This scale is comprised of 16 items that span eight distinct factors. These factors are:

Distortion of consequences. For example, the item “Taking personal credit for ideas that were not your own is no big deal” probes this aspect.

Dehumanization. Illustrated by the item “Some people have to be treated roughly because they lack feelings that can be hurt.”

Attribution of blame. Captured by items like “People who get mistreated have usually done something to bring it on themselves.”

Diffusion of responsibility. For instance, the item “It’s okay to tell a lie if the group agrees that it’s the best way to handle the situation.”

Displacement of responsibility. Examined by the item “People shouldn’t be held accountable for doing questionable things when they were just doing what an authority figure told them to do.”

Moral justification. Represented by the item “It is alright to lie to keep your friends out of trouble.”

Advantageous comparisons. For example, the item “Considering the ways people grossly misrepresent themselves, it’s hardly a sin to inflate your own credentials a bit.”

Euphemistic labelling. As per the item “It’s okay to gloss over certain facts to make your point.”

Participants responded to these items using a seven-point Likert scale ranging from “Strongly Disagree” to “Strongly Agree.” This framework allows for a multi-dimensional exploration of an individual’s propensity towards moral disengagement, thereby providing a comprehensive understanding of various aspects that contribute to immoral decision-making. This was administered to the UK-representative sample only.

Analytical strategy.

Simple Spearman correlation matrices were used to explore linear relations between phronesis and (i) flourishing sub-factors, (ii) Big Six personality traits and dark personality traits, (iii) moral disengagement factors, and (iv) Moral Foundations. This was important to show convergent and discriminant validity of the SPM (i.e., that it performs empirically as might be predicted theoretically in relation to other established measures). For these analyses we used the ggplot2 [80], GGally [81], and ggcorrplot [82] packages in R. We also used linear regressions to predict flourishing components using phronesis network components over a two-month interval to establish the predictive validity of the SPM, going above and beyond simple correlations. Finally, we used multiple linear regression to predict various outcomes of interest using phronesis network components, over and above moral foundations. This allowed us to test whether the SPM can explain variance in outcomes of interest over and above one of the most popular measures found within the moral psychology domain, thereby establishing that the SPM measures something important (insofar as the outcome variables are important) that the MFQ2 does not. All analyses were conducted in R Studio [62] using aforementioned packages, or using Jamovi [83] with an R [84] backend. The car package [85] was used to perform hierarchical regression analyses within Jamovi to establish incremental criterion validity of the phronesis network components over and above moral foundations.

Flourishing.

We first explored whether phronesis components correlated with flourishing components (see Table 10). The same general pattern and magnitude of correlations was found within both samples. These findings suggest that all aspects of phronesis aside from the perceptual components (Virtue Identification, Situational Moral Relevance, and Situational Moral Irrelevance) correlated with all aspects of flourishing. The strongest correlations, as might be expected, were between phronesis components and Character Strengths, while the weakest correlations were between phronesis components and Financial Security.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 10. Correlation matrices including WBA flourishing factors alongside ten components of phronesis in the UK (upper) and US (lower) samples.

https://doi.org/10.1371/journal.pone.0317842.t010

Phronesis components were collectively entered as predictors within a linear regression, with flourishing variables measured two months later as the outcome variables. Phronesis components collectively functioned as strong predictors of all aspects of flourishing two months later apart from Financial Security. The findings are found in Table 11. Additionally, baseline phronesis predicted Emotional Health two months later over and above baseline Emotional Health (Model 1 r² = .47, Model 2 r² = .51; F[10, 283] = 2.27, p = .014).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 11. Predicting flourishing using the ten phronesis components two months later.

https://doi.org/10.1371/journal.pone.0317842.t011

Flourishing and the phronesis network.

At the next step, flourishing variables from the WBA were introduced alongside the SPM phronesis network variables for both the UK and US samples. In previous network models, the four most important aspects of the overall Phronesis network were adjudicative variables (namely Moral Deliberation and Moral Integration) and identity variables (namely, Moral Self-relevance and Aspired Moral Identity). Network analyses presented below suggest that Moral Integration is more on the periphery of the phronesis network among US participants compared to UK participants. Moreover, these network models appear to suggest that Social Connectedness is central to overall flourishing, with Financial Security only peripherally related to the phronesis network compared to other flourishing variables. In both the US and the UK, Emotional Regulation sits somewhat separate from the other phronesis and flourishing variables, but it nonetheless supports both. For both samples, Meaning and Purpose as well as Character Strengths were the most closely related aspects of flourishing to the phronesis network, as might have been expected theoretically. A network invariance test suggested a non-significant difference between US and UK network structures (M = .10, p = .408). As before, we also conducted a global strength invariance test. We found that the overall connectivity of the phronesis network with flourishing was no different in the USA (6.34) compared to the UK (5.72); S = .62, p = .249. These relationships between variables in both networks/samples are plotted in Fig 4 and centrality statistics are tabulated in Table 12.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4.

Flourishing in the context of the phronesis network in UK (upper) vs US (lower) samples. Note. EBIC penalization applied.

https://doi.org/10.1371/journal.pone.0317842.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 12. Centrality statistics for the US vs UK phronesis network models in the context of flourishing.

https://doi.org/10.1371/journal.pone.0317842.t012

HEXACO personality and the dark tetrad.

We next sought to explore the personality correlates of phronesis. This included both the Big 6 and the Dark Tetrad (see Table 13). The general pattern of correlations between phronesis components and personality was positive. Notably, the three strongest correlations were between Conscientiousness and (i) Moral Deliberation, (ii) Moral Self Relevance, and Negative Moral Emotion, three of the most central components of the phronesis network. Emotional Regulation was moderately negatively related to trait Negative Emotion, as might be expected. Negative Moral Emotion was only slightly positively related to the general personality trait of Negative Emotion, indicating they are not the same thing. The general pattern of correlations shows negative associations between phronesis components and the Dark Tetrad. The main exception to this was that those with higher Aspired Moral Identity and Moral Integration were slightly more likely to be Narcissistic. Situational Moral Relevance, Situational Moral Irrelevance, and Virtue Identification, our three objective measures of moral perception, were shown to have criterion validity here as they were negatively associated with the Dark Tetrad, on the whole. As might be expected, the strongest negative relationships were between the moral perception variables and the main trait associated with moral perceptual deficits–Psychopathy.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 13. Correlation matrix showing the relationships between phronesis network components and personality (HEXACO and dark tetrad).

https://doi.org/10.1371/journal.pone.0317842.t013

Moral disengagement.

Given the complex pattern of correlations between phronesis and personality, including the dark personality traits, it was important to explore the relationship between phronesis and morally salient variables in more detail. To that end, we next correlated phronesis components with moral disengagement variables (see Table 14). The results showed a much clearer picture whereby all phronesis components were negatively related to different indices of the propensity to morally disengage, creating a much clearer picture of criterion validity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 14. Correlations between phronesis network components and propensity to morally disengage.

https://doi.org/10.1371/journal.pone.0317842.t014

Moral foundations.

The concept of phronesis emphasizes a deliberate and rational approach to moral decision making, in the service of flourishing. However, other moral theories emphasize the role of foundational emotional intuitions, driven by mostly innate sentiments, in moral judgements. Arguably the most well-studied sentimentalist moral theory is Moral Foundations Theory, which emphasizes Purity, Authority, Loyalty, Proportionality, Equality, and Care as the basic moral intuitions [36]. As a non-sentimentalist moral construct, but a moral construct nonetheless, we would therefore expect some degree of correlation between Moral Foundations and phronesis. When we tested this (see Table 15), we found that several aspects of the phronesis network were positively related to Moral Foundations. Notably though, the Emotional Regulation and Situational Moral Irrelevance aspects of the phronesis network were negatively related to Moral Foundations, reflecting the non-sentimentalist emphasis within phronesis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 15. Correlations between phronesis network components and moral foundations.

https://doi.org/10.1371/journal.pone.0317842.t015

Given the mixed associations between phronesis network components and Moral Foundations, we then decided to test which set of constructs best predicts our key criterion valiables–the flourishing network. Therefore, we used a hierarchical linear regression to test—whether the network of variables constitutive of phronesis would predict aspects of flourishing over and above Moral Foundations. We first entered moral foundations components (Care, Equality, Proportionality, Loyalty, Authority, Purity) as the Model 1 predictors, and phronesis network components as Model 2 predictors. Model comparison results are found in Table 16, and full beta coefficients are found in the supplementary file “MFT_comparisons”. Table 16 shows that moral foundations explain between 5% (Emotional and Physical Health) and 24% (Character Strengths) of the variance in flourishing outcomes. When using phronesis network components to predict these flourishing outcomes over and above Moral Foundations, phronesis network components explain large amounts of additional variance in all flourishing oucomes except Financial Security, ranging between +5% (Physical Health) to +18% (Character Strengths).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 16. Predicting outcomes of interest with phronesis network components, over and above moral foundations.

https://doi.org/10.1371/journal.pone.0317842.t016

We also wanted to explore the incremental criterion validity of phronesis network components over and above Moral Foundations for additional morally salient outcomes. To that end, Table 16 also included morally salient aspects of personality. Moral Foundations explained variance in morally salient aspects of personality, ranging from 5% (Narcissism) to 14% (Honesty/Humility). Phronesis network components explained variance in all of these outcomes over and above Moral Foundations, however, ranging from +9% (Machiavellianism) to +23% (Sadism).

Finally, we tested for the degree to which phronesis network components would predict the propensity to morally disengage over and above Moral Foundations (see Table 16). Moral Foundations explained large amounts of variance in all outcomes of interest, ranging from 6% (Euphemistic Labelling) to 17% (Dehumanization of Others). As before, phronesis network components explained large amounts of variance in these outcomes of interest over and above Moral Foundations for all outcomes, ranging from +12% (Displacement of Responsibility) to +34% (Attribution of Blame).

Phronesis and flourishing.

Our results indicate that the phronesis components collectively predict flourishing outcomes, with particularly strong associations found between phronesis and Character Strengths, as well as Meaning and Purpose. These findings are consistent with the theoretical grounding of phronesis in Aristotelian ethics, which posits that practical wisdom is central to leading a flourishing life. The weaker, yet still significant, relationship with Social Connectedness further supports the potential role of phronesis in facilitating positive social relationships, albeit to a lesser extent than its influence on individual moral strengths.

Interestingly, the perceptual components of phronesis, including Virtue Identification and Situational Moral Relevance and Irrelevance, did not show significant correlations with flourishing outcomes. This suggests that while these perceptual abilities may be important aspects of moral judgment, they are not associated with broader well-being outcomes measured in this study. The peripheral role of these perceptual components within the phronesis network, especially in the US sample, further highlights their distinct function compared to other, more central components like Moral Deliberation and Moral Self-relevance. This is consistent with the idea that knowing the good does not necessarily entail doing the good, and therefore flourishing is not necessarily downstream of moral perception. Additionally, we showed that phronesis is more than a mere correlate of flourishing, but also robustly predicts flourishing two months later.

Phronesis, personality, and moral disengagement.

The phronesis components were also analyzed in relation to personality traits and the propensity to morally disengage. The positive correlations between Moral Deliberation, Moral Self-relevance, and Conscientiousness underscore how these aspects of phronesis align with a conscientious and thoughtful approach to moral decision-making. Notably, the distinction between Negative Moral Emotion and general trait Negative Emotion suggests that these constructs capture different aspects of emotional experience. Negative Moral Emotion is associated with a greater tendency to morally disengage, while also being closely related to Moral Self Relevance in the phronesis network, showing that practically wise people are emotionally invested in being (morally) good people. Additionally, the inverse relationships between phronesis components and Dark Tetrad traits, especially Moral Self Relevance, and the clear pattern of negative associations between phronesis components and moral disengagement, emphasize the potential of phronesis to mitigate tendencies towards morally disengaged behavior. However, the positive correlation between Aspired Moral Identity and Narcissism warrants further investigation, as it suggests that for some individuals, the pursuit of moral aspirations may be driven by self-interest or an inflated sense of self-importance.

Moral foundations.

The relationships between phronesis components and Moral Foundations reveal important insights into how different aspects of moral reasoning and emotion interact. As expected, Emotional Regulation was negatively related to Moral Foundations in some cases, which aligns with the idea that Moral Foundations Theory emphasizes the role of basic emotions in moral judgment. This suggests that those who are adept at regulating their emotions may be less influenced by these foundational moral intuitions, potentially reflecting a more rational, rather than emotional, approach to morality. Similarly, Situational Moral Irrelevance was also negatively related to Moral Foundations on the whole, indicating that individuals who excel at identifying when a situation does not require moralization are less likely to rely on sentimentalist moral intuitions. This suggests that emotional moral thinkers might struggle to discern when a situation does not necessitate a moral response, possibly leading to over-moralization of situations.

On the other hand, components of phronesis that are engaged when a situation is deemed morally salient, such as Aspired Moral Identity, Moral Integration, and Moral Deliberation, were positively associated with Moral Foundations. This is perhaps unsurprising, as these aspects of phronesis are likely to be activated in scenarios where moral considerations are at the forefront, and sentimentalist thinkers might assume most situations are morally salient. However, the weaker performance in Situational Moral Irrelevance and Emotional Regulation among these individuals suggests that they might be prone to moralize situations unnecessarily and struggle with managing their emotional responses in morally charged contexts.

Given the mixed associations between phronesis network components and Moral Foundations, we sought to determine which set of constructs better predicts our key criterion variables in the flourishing network. The findings from our regression analyses revealed that while Moral Foundations explained a significant portion of the variance in flourishing outcomes, phronesis network components added substantial predictive power across nearly all outcomes, particularly in areas like Character Strengths and Meaning and Purpose. This suggests that while Moral Foundations provide a useful framework for understanding some aspects of flourishing, the phronesis network offers additional, and often superior, explanatory value. Moreover, when predicting morally salient aspects of personality, phronesis network components consistently explained more variance than Moral Foundations, particularly in traits associated with the Dark Tetrad. This highlights the robustness of phronesis as a construct that not only complements but also extends beyond the explanatory power of Moral Foundations. Finally, in terms of the propensity to morally disengage, phronesis network components again outperformed Moral Foundations in predicting this critical outcome. The ability of phronesis components to account for substantial additional variance in moral disengagement indices underscores their relevance and utility in understanding moral behavior, especially given that moral disengagement is a more proximal moral outcome compared to flourishing. In this way, phronesis has been shown here to predict both proximal (morality) and distal factors (flourishing) over and above Moral Foundations.

Concluding remarks.

The empirical studies undertaken and described in this article have provided a revised model of Aristotelian phronesis (the neo-APM)—originally theoretically motivated only—and yielded a practicable instrument to measure phronesis. Our studies have demonstrated meaningful correlations between phronesis and flourishing and have unearthed various other associations that either confirm previous studies of wisdom in psychology or add to the knowledge basis about the narrower construct of practical wisdom. The data set we gathered is large and some of it still awaits further analysis in subsequent papers. More research in this area is urgently called for, for example research comparing the neo-APM more directly with wisdom as understood in psychology (see 87), although that task is complicated by the fact that there are so many wisdom conceptions available [98], and the internal correlation between them rarely exceeds .3 [99].

We choose to end this discussion by returning to the most “controversial” premise of this article, noted earlier. Why conduct a series of psychological studies to demonstrate the credibility of a construct that is of pure philosophical provenance, and over 2,300 years old at that? This question hides a deeper one: what does it mean to conduct an empirical analysis based on a specific explicit realist moral ontology, such as neo-Aristotelian virtue ethics? This endeavor can be interpreted in many ways, ranging from a weak reading to a strong reading of the extent to which the ontology in question informs the research orientation.

We have rejected the strong reading that a philosophical ontology should constrain the research process, by deciding to follow a bottom-up approach, rather than beginning with CFA to consolidate a pre-existing model as in our previous work [19]. We thus wish to avoid the critique that our approach is revanchist and involves special pleading to a long-gone philosopher [37]. Nevertheless, we acknowledge that how we gathered and interpreted our data has unavoidably been affected by our motivation to establish the credibility and serviceability of the Aristotelian model that we confirmed, in a revised incarnation, in Studies 1a–1c.

We remain unrepentant about our ontological assumptions because we believe no social scientists can avoid them. All research in moral psychology presupposes some implicit moral ontology that steers the analysis, and our uniqueness lies simply in making this ontology explicit rather than implicit. For example, Kohlberg’s [1, 20] enterprise was motivated by his antipathy toward moral relativism, and he gathered evidence that he hoped would rebut such an ontology (and which it did). Phronesis (in a neo-Aristotelian sense) does not become anyone’s chosen area of study unless they are committed to a realist moral ontology, which assumes that there are better or worse answers to moral quandaries, and that there is an essentially intellectual process, which one can cultivate through moral learning, that helps one deliberate and decide upon the better answers and stay away from the worse ones.

All in all, this article has provided evidence of a method through which a philosophical construct can mature into a fully-fledged psychological one. More generally, it has confirmed our assumption about the potential fruitfulness of crossover work between philosophy and psychology. However, such work requires a bridging effort from both sides—rather than philosophical ideas being removed from empirical scrutiny, or psychology ignoring ancient (or even modern) wisdom from adjacent fields. This article has not only demonstrated the viability of transforming a philosophical construct into a psychological one but also highlighted the importance of cross-disciplinary collaboration. More specifically, the neo-APM, as presented in this article, represents a significant stride in the empirical measurement of phronesis. It encapsulates a comprehensive, multi-dimensional approach that integrates theoretical depth with empirical rigor. The measure based on this model is not merely an academic tool; it is a gateway to understanding key dynamics of moral virtue and behavior across various life stages and societal challenges, marking an important contribution to the field of moral psychology. The neo-APM and its associated measurement tool (the SPM) thus offer new avenues for research and application, with the potential to reshape our understanding of moral psychology and its practical implications. We believe that this work marks a significant step forward, not just for the study of phronesis but for the broader field of psychology, which stands to benefit greatly from continued integration with philosophical insights.