The Reasonable, the Rational, and the Good: On Folk Theories of Deliberative Judgment

Igor Grossmann; Niyati Kachhiyapatel; Ethan A Meyers; Hanxiao Zhang; Richard P Eibach

doi:10.1162/opmi.a.24

. 2025 Aug 29;9:1375–1410. doi: 10.1162/opmi.a.24

The Reasonable, the Rational, and the Good: On Folk Theories of Deliberative Judgment

Igor Grossmann ^1,^*, Niyati Kachhiyapatel ¹, Ethan A Meyers ¹, Hanxiao Zhang ¹, Richard P Eibach ¹

PMCID: PMC12435989 PMID: 40958953

Abstract

Judgment is often described in terms of an intuitive (System 1) versus deliberative (System 2) dichotomy, yet sound deliberation itself can take more than one form. Building on philosophical traditions and distinctions in treatment of sound judgment in economics and law, we propose that lay conceptions revolve around two distinct types of deliberate judgment: rational, emphasizing rule-based and utility-focused reasoning for well-defined problems, and reasonable, prioritizing context-sensitive and socially conscious reasoning for ill-defined problems. Across four studies in English-speaking Western samples (Studies 1–4; N = 2,130) and a Mandarin-speaking Chinese sample (Study 4; N = 697), participants described their notions of “sound” and “good” judgment, evaluated social scenarios, chose between candidates with distinct judgmental profiles, and categorized non-social objects. Results consistently showed that people view both rationality and reasonableness as common forms of deliberate sound judgment, while treating them as distinct. Participants preferred rational deliberation for algorithmic social roles linked to well-defined tasks and reasonable deliberation for interpretive roles linked to ill-defined tasks. Moreover, framing decisions as rational vs. reasonable influenced whether participants relied on rule-based vs. overall-similarity strategies in classification tasks. These findings suggest that lay understanding of sound judgment does not rely on a single standard of judgmental competence. Instead, people recognize that both rationality and reasonableness are critical for competent deliberation on different types of problems in life.

Keywords: rationality, decision-making, person perception, folk theories, practical wisdom

INTRODUCTION

Imagine a city official tasked with drafting regulations to address severe housing shortages. She reviews data, runs cost–benefit analyses, and consults urban development models to identify the least biased solution—an approach many would call rational because it emphasizes consistency and formal rules. Yet, when residents protest the proposed construction, citing moral and cultural concerns, the same official meets with them, weighs competing values, and adapts her plans to accommodate local realities. In these actions, she exemplifies a more reasonable approach, aiming to balance multiple stakeholders’ perspectives rather than follow a strictly rule-based, utility-maximizing strategy.

This contrast between formal, rule-based, algorithmic optimization and a more holistic, socially conscious deliberation crops up across countless real-world decisions. From choosing trustworthy advisors to evaluating policy trade-offs, sound judgment might appear to hinge on logical analysis and bias avoidance in some cases, while requiring a context-sensitive approach that attends to the peculiarities of individual cases in others. These examples highlight different forms of deliberate judgment and suggest that the nature of the decision problem shapes which standard people apply.

Beyond the Intuitive–Deliberative Dichotomy: Rational vs. Reasonable Deliberation

The idea that more than one standard defines human judgment is hardly novel. Aristotle distinguished theoretical knowledge (episteme), aimed at universal truths, from practical wisdom (phronesis), concerned with ethical and context-specific considerations (Aristotle, trans. 1984). Renaissance humanist Michel de Montaigne advanced a deliberative experiential ideal, urging reflective engagement with lived particulars over strict rules (de Montaigne, 1580/2007). In contrast, René Descartes upheld systematic deduction and reliance on logical clarity as the gold standard for knowledge (Descartes, 1641/1901). David Hume added yet another voice, portraying much everyday judgment as intuitive, arising automatically from custom and sentiment rather than from either formal logic or reflective experience (Hume, 1748/2007). These diverse perspectives highlight that human judgment cannot be reduced to any single mode, whether analytical, intuitive, or experiential.

Building on these philosophical foundations, psychological scientists have sought to develop empirical models of human judgment and decision-making. One prominent example is the dual-process model, which distinguishes between “System 1” (fast, automatic, affect-driven) and “System 2” (slower, more effortful, rule-based) thinking (Evans & Stanovich, 2013; Kahneman, 2011). While this model has been influential in advancing our understanding of judgment, early accounts often treated deliberation as a unitary phenomenon, emphasizing its role in promoting logical coherence and “rational” decision-making (e.g., Tversky & Kahneman, 1974; cf. Baron, 1993). A range of scholars have since questioned the assumption that all slow reflection forms a single, logically oriented “System 2.” Conceptual critiques point out redundancy in two system labels (Keren & Schul, 2009) and even suggest a continuum model in which the same principles govern both intuitive and deliberate judgments (Kruglanski & Gigerenzer, 2011). Empirically, fast conflict detection studies show that logical monitoring can operate under severe time pressure (Bago & De Neys, 2017; De Neys, 2021), whereas tripartite and motivational models split deliberate thinking into multiple subsystems (Evans & Stanovich, 2013; Melnikoff & Bargh, 2018). These findings suggest that a unitary view of deliberation may not capture the rich insights offered by philosophers such as Aristotle, Montaigne, and Descartes. Recognizing this complexity, other social-cognitive models have proposed that there may be multiple modes of deliberate sound judgment, each guided by a different set of priorities and standards of assessment (e.g., Tetlock, 2002).

The idea that there may be distinct, valid standards of sound deliberation can be illustrated by contrasting judgmental norms in the domains of economics, epistemology, and the law. In neoclassical economic models, an agent is deemed “rational” if they maximize personal utility under consistent preferences (Simon, 1957; Thaler, 2015). Deviations from this ideal, whether stemming from cognitive shortcuts or social preferences, are typically framed as “biases” or “heuristics” that undermine rationality (Tversky & Kahneman, 1974), though some scholars note that choices diverging from strict utility maximization may be purposeful if moral or contextual factors guide behavior (e.g., Hertwig & Gigerenzer, 2011; Schwartz, 2000). Beyond economics, in the field of epistemology, “rationality” may be defined in terms of Bayesian logic, which specifies formal rules for forming and updating beliefs based on the quality of accumulated evidence (Bovens & Hartmann, 2004). As with economic rationality, deviations from the ideal of Bayesian epistemology are typically framed as “biases” or “heuristics” that undermine rationality (Tversky & Kahneman, 1974), though some scholars note that these deviations may reflect adaptive error management (e.g., Haselton & Nettle, 2006). The economic and Bayesian definitions of rationality share an emphasis on standards of judgment that can be represented with well-defined formulae/algorithms. By contrast, legal scholars often invoke a “reasonable person” standard to judge whether actions and outcomes are acceptable, taking into account situational particulars and broad norms (Sibley, 1953; Tobia, 2018). Far from insisting on internal consistency or self-serving optimization, reasonableness in law is oriented toward fairness, norms, and appropriate contextual sensitivity (Jaeger, 2025, for a review). In political philosophy, Rawls (1999) similarly distinguished the “rational,” aimed at maximizing one’s ends, from the “reasonable,” which involves recognizing and coordinating with others’ interests. Thus, it appears that scholars in different fields treat different forms of deliberation as central to sound judgment: While economists and Bayesian epistemologists emphasize rationality for utility and consistency, lawyers and political philosophers emphasize reasonableness for social norms, moral considerations, and flexible contextual adaptation¹.

Lay Beliefs: Emerging Evidence

How do laypeople parse these concepts? Recent scholarship started to examine when people rely on “head” (reason) versus “heart” (intuition) in their judgments (e.g., Hsee et al., 2003), and whether this division captures the chief ways humans conceive of sound judgment. For instance, Inbar and colleagues (2010) asked Americans how much intuition or reason should guide decisions across various everyday scenarios (e.g., which apartment to rent, which restaurant entrée to order). They systematically varied features such as the complexity or sequentiality of a choice—finding that participants deemed deliberation more suitable for problems that appeared objectively evaluable, multi-stepped, or logically precise. By contrast, participants saw intuition as more suitable for subjective contexts (e.g., personal tastes or “gut” feelings). Relatedly, intuition is preferred when decisions are framed as matters of authentic self-expression (Oktar & Lombrozo, 2022)—an idea rooted in the “ethic of authenticity,” which holds that some personal commitments should occasionally outweigh strict rational calculation (Varga & Guignon, 2020). Intuition is also preferred when one believes oneself to be an expert in the decision domain (Pachur & Spaar, 2015). Meanwhile, Hsee and colleagues (2015) introduced lay rationalism as an individual difference in how much Americans value “hard” facts and cost–benefit reasoning over feelings, showing that those scoring higher on their Lay Rationalism Scale prefer utilitarian (vs. hedonic) products, donate less in high-emotion charity appeals, and save more money. These findings suggest that lay understanding of being “rational” means minimizing emotional influence, at least in the US (also see Li & Hsee, 2019).

Grossmann and colleagues (2020) built on these lines of inquiry by exploring whether rationality and reasonableness constitute distinct lay standards of sound judgment, drawing explicitly on ideas from economics (which treats rational decisions as strict preference-maximization) and legal scholarship (which frequently invokes “reasonableness” to incorporate social norms and contextual fairness). This research combined linguistic/cultural analyses with a series of experiments. In the first set of studies the distinction was tested using spontaneous descriptions, social perceptions, and large-scale analyses of cultural products (newspaper articles, soap operas, legal opinions, Google Books in multiple languages). Across several Indo-European languages, findings showed that “rational” was spontaneously tied to self-focused or instrumental aims, while “reasonable” highlighted social or norm-based considerations. Moreover, laypeople consistently saw “rationality” as preference-maximizing (focusing on logic, personal gain, abstract self-interest), whereas they viewed “reasonableness” as context- and norm-sensitive (highlighting social fairness, contextual adaptation, empathy). In subsequent experiments in Canada, the US, and in Pakistan, participants made decisions in classic economic games, such as the Dictator Game (deciding how to allocate money between oneself and another), the Commons Dilemma (choosing how much to harvest from a shared resource), and the Prisoner’s Dilemma (deciding whether to cooperate or defect). Despite their varying cultural and socioeconomic backgrounds (including Pakistani bankers, street merchants, and North American students) participants consistently described rational actions as those aiming to maximize personal gain, and reasonable actions as balancing personal benefits with fairness. Across these diverse samples, the “rational” choice was viewed as unconstrained by social norms, whereas the “reasonable” choice was seen as attuned to those norms.

Are these findings from Grossmann et al. (2020) simply another way of invoking the “head vs. heart” divide, where rational means unemotional and reasonable means emotionally guided? Or do lay rationality and reasonableness represent two distinct forms of deliberate judgment? It could be that lay understanding of “rationality” simply refers to dispassionate calculation, while lay “reasonableness” relies on emotion-laden intuition. If this is so, the observation of more cooperative behavior in economic games when asked to act reasonably (vs. rationally) in earlier studies (Grossmann et al., 2020) may reflect intuition-driven choice similar to the results in studies by Inbar and colleagues (2010; also see Rand, 2016). Notably, analyses of response times in past studies revealed no significant differences between rational and reasonable conditions (instructions to consider rational or reasonable approaches to choice in economic games; Grossmann et al., 2020; Studies 5–7), suggesting comparable cognitive engagement with the decision scenarios.

Here, we suggest an alternative interpretation of past scholarship on lay theories of judgment. We propose that lay perceivers believe that both rational and reasonable judgments can be thoughtful and deliberate, but emphasize different aims (cf. Rawls, 1999). Building on earlier scholarship (Grossmann et al., 2020), lay rationality may stress internal consistency and maximizing personal payoff, whereas lay reasonableness may be oriented toward integrating multiple stakeholders’ preferences and civic norms. Consequently, stereotypic attributes of rationality may involve self-serving competences, while stereotypical attributes of reasonableness may involve socially-conscious concerns (Grossmann et al., 2020; Study 3). Moreover, informed by examples in the opening paragraph, we propose that the lay distinction between rationality and reasonableness extends beyond the social dimension, with the reasoning process and judgmental styles favoring rule-based processes and analytic reductionism on the one hand, and dialogical processes and context-sensitive holism, on the other hand. In short, folk theories of rational and reasonable deliberation may encapsulate a rich array of stereotypes, foci, judgmental styles and reasoning processes (see Table 1).

Table 1. .

Proposed framework to distinguish folk theories of rational and reasonable deliberate judgment.

	Rational	Reasonable
Stereotypic attributes	Less sensitive to context/mechanical/consistency-focused	Highly sensitive to context/flexible
Focus	Focuses on commensurable values and seeks to reduce them to a single dimension of valuation	Balances incommensurable values that cannot be reduced to a single dimension of valuation
Judgmental style	Reductionist—i.e., excludes from consideration qualities deemed irrelevant or that cannot be efficiently integrated into categories/rules even at the cost of oversimplification	Holistic—i.e., tries to include as much of the available information as possible even at the cost of precision, clarity, or efficiency
Reasoning process	Deliberative reasoning that uses a logical/syllogistic/rule-based process to proceed in generally linear progression from premises to conclusions	Deliberative reasoning that uses dialogic reasoning that references social conventions/norms and rhetorical strategies such as reasoning by analogy, and narratives
Relevant contexts/domains of application/decision space	Well-defined problems—i.e., closed systems where all relevant variables are fixed and known with clear rules for how to integrate these variables; quality of judgments can be objectively assessed	Ill-defined problems—i.e., open systems where the relevant variables are uncertain and/or unknown and/or one lacks clear rules for how to integrate these variables; quality of judgments must be subjectively assessed

Open in a new tab

If both rationality and reasonableness represent lay standards of deliberate judgment, a question arises about when people switch between them. Building on prior theoretical work (Grossmann & Eibach, 2024; also see Hammond, 2007; Payne et al., 1993; Simon, 1957), we therefore propose that decision ecology informs preferences for rational vs. reasonable judgment (Table 1). Specifically, lay preferences for rationality may concern well-defined problems where self-focused control, algorithmic rules, and bias reduction drive success. Conversely, lay preferences for reasonableness may concern ill-defined problems that require balancing incommensurable values, social input, and context-sensitive adaptation. The hypotheses about the properties (stereotypic attributes, focus, judgmental styles, and reasoning processes) that lay perceivers will attribute to rationality and reasonableness outlined in Table 1 are based on previous theory (Grossmann et al., 2020) and research (Dutta, 2018; Schraw et al., 1995) describing psychological processes and judgmental approaches that are adaptive for coping with well-defined and ill-defined problems, respectively. In other words, we hypothesize that lay-perceivers’ understanding of the decision ecology of each type of problem informs their ideas about what properties of deliberative reasoning are useful for each type of problem, and we further hypothesize that lay perceivers tend to group those adaptive properties together and associate those groupings with the terms “rationality” and “reasonableness.”

The Present Research

We conducted a series of five studies to test several key hypotheses based on the framework we outlined in Table 1. First, we hypothesized that laypeople consider rationality and reasonableness to be distinct standards for evaluating the soundness of judgment. In particular, we tested the framework’s claim that these are both deliberative types of judgment but that they differ in some of their core qualities. To test this hypothesis, we explored if people understand rational and reasonable concepts as deliberate and thoughtful (vs. emotional and intuitive) in the context of sound judgment (in Study 1a and good judgment in Study 1b), but they do not use them interchangeably because they recognize that the concepts are aligned with distinct properties corresponding to the distinctions outlined in Table 1. Although lay definitions of rationality and reasonableness have been differentiated in previous research (Grossmann et al., 2020), that research primarily focused on laypeople’s beliefs about rationality and reasonableness in the context of economic choices that pit self-interest against others’ interests. The framework in Table 1 proposes that laypeople draw broader distinctions between rationality and reasonableness that have implications beyond those tested in previous research. To investigate this, Studies 1a and 1b explore lay definitions of these judgmental properties in greater depth and with more robust methods than in previous research.

Second, we hypothesized that people’s preferences for rational versus reasonable agents should shift systematically across a range of social roles that differ in their demands for analytic generalization versus holistic, context-specific considerations. To test this hypothesis, we examined how laypeople apply rationality vs. reasonableness as standards across a broader range of everyday judgments than the economic transactions tested previously (Studies 2–3). We further hypothesized that people will prefer rationality as the standard of judgment in well-defined contexts but prefer reasonableness as the standard of judgment in ill-defined contexts in people’s lives (Study 3). We conjectured that participants would opt for rational deliberation in well-defined contexts because such contexts involve specific problems, and effective solutions in these cases depend on self-directed control, established rules, and the reduction of bias. Conversely, we expected participants to favor reasonable deliberation for ill-defined problems, because such problems require balancing incommensurable values, social input, and contextual adaptation.

Third, we hypothesized that the distinction between rationality and reasonableness would extend even to non-social domains. According to the framework in Table 1, laypeople should associate rationality with a reductionist judgmental style—i.e., reduce judgments to the smallest set of relevant considerations while ignoring, downplaying, or trivializing considerations that are seen as non-essential as “noise.” Conversely, laypeople should associate reasonableness with a holistic judgmental style—i.e., include as much of the “good enough” available information as possible even at the cost of precision, consistency, clarity, or efficiency. To the extent that this is true, then it should lead laypeople to expect that even decisions about non-social problems should be made differently when they apply the rational vs. reasonable standard to those problems. To test this hypothesis, we asked participants to indicate their expectations for how a rational person and a reasonable person would categorize non-social objects: by matching a target object by a single feature shared among all objects (a reductionist option) or by focusing on the “good enough” overall similarity even if none of the features were consistently shared among all objects (a holistic option; Studies 4–5)². Assessing whether laypeople extend the distinctions between rationality and reasonableness even to a relatively unfamiliar, non-social task is a strong test of our hypothesis that laypeople recognize these as qualitatively different standards for judgment.

Transparency and Openness

For each study we report how we determined our sample size, data exclusions, all manipulations and measures (https://osf.io/excny/). Study 1 was exploratory and not preregistered. We preregistered the hypotheses, design, and analyses of Study 2 (https://osf.io/fw2ht), Study 3 (https://osf.io/x54j8/), Study 4 (https://osf.io/7x8zk), and Study 5 (https://osf.io/92vpx). All analyses were performed in R Statistical Software (v4.2.3; R Core Team, 2023).

STUDY 1

We started our research by exploring whether people perceive reasonable persons to be more emotional and intuition-driven than rational persons, or whether both concepts reflect chiefly deliberate qualities. Study 1a examined the link between general perceptions of sound judgment and the attributed characteristics of rationality and reasonableness, while Study 1b focused on good judgment when working through a challenging situation, to zero-in on contexts in which judgmental competence is usually required³. In both versions of the study, we took a multi-pronged approach. First, people spontaneously described characteristics of a person showing sound/good judgment. Second, they reported the extent to which they ascribed analytical, moral, and social characteristics to rational and reasonable persons. Methodologically, we relied on natural language processing of word co-occurrences and psychometric factor analyses to examine the semantic understanding of each concept.

If our deliberation account of lay reasonableness and rationality corresponds to folk theories of judgment (Table 1), it will impact attributes of judgmental competence people assign to each concept, revealing distinct semantic and psychometric structures of characteristics associated with each. Specifically, we expected that analytic strengths would be attributed to both, but interpersonal strengths would be uniquely attributed to reasonable people, consistent with prior research (Grossmann et al., 2020). Furthermore, if lay rationality is understood as reductionist judgment in service of a single favored value (and filtering less favored values), psychometric analyses will show distinct clusters of traits attributed to rational people (e.g., greater differentiation between analytical and moral qualities or between agency and communion-oriented characteristics). Conversely, if lay reasonableness is understood as integration-based judgment in service of a plurality of relevant values, the lay concept of reasonableness will show more interconnectedness between traits attributed to reasonable people.

Methods

Participants.

All studies were approved by the University of Waterloo Office of Research Ethics (Protocols 30580 and 40360). We obtained informed consent from all participants. We recruited participants via Amazon’s Mechanical Turk (mTurk) through CloudResearch (Litman et al., 2017) to complete a 5-min survey on the Qualtrics platform in exchange for $0.50. To be eligible for participation, individuals were required to be at least 18 years of age, reside in Canada or the US and possess a HIT approval rating at or above 90%. Study 1a participants were prevented from signing up for Study 1b.

Estimating sample size for advanced psychometric models (up to 30 parameters in Study 1a) and natural language processing studies is complex. Given the exploratory nature of Study 1a, we aimed to ensure content-rich and diverse open-ended content. Therefore, we relied on general rules of thumb rather than formal sample size calculation. To this end, we followed Harrell’s (2015) general guideline of 10–20 observations per parameter in multivariate analyses, while considering Epskamp and colleagues’ (2018) general recommendation for larger samples in psychometric models. Targeting 450 mTurk participants in Study 1a, we oversampled to about 600 (33% increase) to account for expected attrition in mTurk studies (Aguinis et al., 2021). In Study 1b, we mirrored this approach, aiming for an identical sample size. We reviewed open-ended responses to questions concerning definition of sound/good judgment and excluded non-compliant entries (e.g., non-sense, bot-like responses like “VERY NICE”), and those who did not complete the attributes survey. We also included partially completed responses of participants who completed the survey, but possibly stopped before completing the demographics, for our analyses. Table 2 presents the demographic information.

Table 2. .

Demographics of participants in Studies 1–5.

	Study 1a	Study 1b	Study 2	Study 3	Study 4	Study 5
Recruited N	635	608	361	509	599	797
Final N	465	490	284	464	411	697
Age_mean (SD)	36.01 (11.53)	34.72 (11.84)	36.65 (11.57)	35.79 (12.63)	25–34^a	27.24 (6.94)
Sex (% female/male/other)	58.1/40.8/1.1	61.6/37.6/0.8	47.3/52.3/0.3	49.1/50/0.9	54.4/45.3/0	60.0/40.0/0
Ethnicity (%)
Asian-American	6.0	7.8	6.7	^b	4.9
African-American	10.5	11.1	20.1		9.6
White	72.3	71.5	63.0		72.9
Hispanic	5.3	5.5	6.0		5.6
Other	5.9	4.1	4.2		7.0
Income_median (US dollars)	50,001–75,000	50,001–75,000	50,001–75,000	50,001–75,000	N/A
Education (%) / M (SD)^c						11.83 (3.15)
Less than High School	0.9	0.6	0.1	0.9	0.9
High School	7.7	10.2	8.1	14.4	11.0
Some College	35.7	40.5	25.4	28.5	25.0
College	40.2	32.9	47.7	43.5	47.0
Post-graduate	15.5	15.8	18.7	12.7	15.3

Open in a new tab

^{^a}

Participants in Study 4 indicated what age group they belonged to (18–24, 25–34, 35–44, 45–54, 55–64, 65–74, 75–84) rather than providing exact age. The estimate reflects the median age group in the sample.

^{^b}

We did not include ethnicity in Study 3, because we targeted people from Canada (44.1%), the US (29.4%), and the UK (26.5%), with varied ethnicity nomenclature across these countries.

^{^c}

Chinese participants in Study 5 reported number of years of secondary education.

Procedure.

In Study 1a, participants were asked to list five adjectives, in order of importance, which they believed described “a person who shows sound judgment,” starting with the most important. Next, they provided a brief open-ended response to describe what sound judgment means to them in their own words (explicitly instructed not to copy definitions from websites/dictionaries). Next, participants were asked to sequentially evaluate three agents (a person who exhibits sound judgment, a rational person, a reasonable person) on a list of 28 characteristics provided (see Table S3 and Figures 2 and S3), from the perspective of how society in general views them; the latter phrase was chosen to ensure we target perception of descriptive norms rather than idiosyncratic personal preferences. The ratings for the person who exhibits sound judgment were always elicited first, while ratings for rational and reasonable persons were presented in a counterbalanced order. Participants were also asked to recall and explain the first task they completed in the study as an attention and data quality check. Finally, participants responded to some demographic questions and the study concluded.

Figure 2. — Qualities attributed to rational and reasonable persons in Study 1. Color-coded adjectives reflect Analytical, Moral, and Inner Fortitude items (Study 1a)/Agency and Communion factors (Study 1b). Top panels: Estimates from linear mixed model with responses to all characteristics nested in participants, with target order (rational vs. reasonable) as a covariate and false discovery rate correction for multiple testing. Dashed vertical line delineates effects 1 unit above midpoint of the 1–7 scale in Study 1a / half a unit above the midpoint of the 1–5 scale in Study 1b. Bottom panels: Pearson’s correlations and 95% CIs of average scores across items making up each factor. ***p < .001, **p < .01, *p < .05.

Study 1b followed a nearly identical procedure with three modifications. First, participants listed adjectives and rated an agent who shows good judgment in a challenging situation and described what good judgment in a challenging situation means to them. Second, the agents were rated on common attributes of judgmental competence (see Figure 2 for the characteristics), representing agency (i.e., “getting ahead”) and communion (i.e., “getting along”). Due to a programming error, all participants judged the rational agent last after judging the reasonable agent and the good judgment agent in a random order. Notably, this randomization allowed testing if perceptions of reasonable and rational agents vary when evaluated consecutively versus separated by judgment of a “good person.”

Materials.

In Study 1a, we chose 28 characteristics (along with ‘rational’ and ‘reasonable’ for evaluation of a person showing sound judgment) based on a pilot test where participants described individuals demonstrating sound judgment, and prior research on person and mind perception (Abele et al., 2021; Gray et al., 2007; Wojciszke et al., 1998; see Table S1 in the online supplement). Participants rated each characteristic on a 7-point scale (1 = Not at all, to 7 = Extremely). We incorporated a diverse set of analytical and interpersonal characteristics. To reduce the number of comparisons, in main analyses we followed the psychometric model specifications outlined below to select a subset of 11 items making up three factors for main manuscript analyses (see Figure 2): Analytical Reasoning (analytical, logical, objective, smart; α_rational = .83; α_reasonable = .87), Empathy & Morality (kind, empathetic, caring, moral; α_rational = .90; α_reasonable = .89) and Inner Fortitude (strong, courageous, brave; α_rational = .88; α_reasonable = .88). Analyses on other analytical and interpersonal items were largely consistent with those reported in psychometric analyses below (see Figure 2).

In Study 1b, we adapted a set of 9 attributes representing Agency (competent, confident, independent, competitive, intelligent) and Communion (tolerant, warm, good-natured, sincere) from Fiske et al. (2002), along with ‘trusting’ as an additional communion item to ensure equal set of items for each dimension (Agency: α_rational = .73; α_reasonable = .70; Communion: α_rational = .82; α_reasonable = .77). As in part person perception research, participants rated each characteristic on a 5-point scale (1 = Not at all, to 5 = Extremely). We also assessed ‘selfish,’ to replicate the differential attributions of self-interest in folk theories of rationality and reasonableness (Grossmann et al., 2020).

Natural Language Preprocessing.

After spell check, we excluded responses that did not answer questions (e.g., wrote a paragraph of text when explicitly asked to write adjectives only). For adjective-based analyses, we removed superfluous spaces. This way, we ensured a standardized set of descriptive terms. For adjective-based analyses, we constructed a Document-Feature Matrix (DFM) from the tidytext package (Silge & Robinson, 2016) and applied correlation-based similarity metrics to test for similarity of associations of adjectives with ‘rational’ and ‘reasonable,’ respectively. Furthermore, we created the frequency matrix of pairwise co-occurrences and adjusted for relative frequency of each term (A & B co-occurrence / Σ (A, B) frequencies). We treated these correlation- and cooccurrence frequency-based indices as markers of semantic meaning revealed in conjunction with each term, statistically evaluating the difference in distributions of key-term-specific correlations and frequencies.

For narrative responses of the personal understanding of “sound judgment” (Study 1a) / “good judgment in a challenging situation” (Study 1b), we tokenized the text, excluding numeric words, typical English stopwords obtained from the tidytext package, and custom stopwords reflecting the restatement of the question (e.g., ‘sound,’ ‘judgment,’ ‘means’) and common trivial opening phrases identified upon initial inspection of the texts (e.g., ‘ability’ and ‘decision’ in “it is the ability to make good decisions”).

Psychometric Modelling.

Our first hypothesis was that laypeople mentally differentiate the concepts of rational person and reasonable person. To test this, we needed to know (a) how many latent trait dimensions organise ratings of each target and (b) whether the same or different dimensions underlie the two targets. We therefore treated the trait ratings as manifest indicators, ran an Exploratory Factor Analysis to uncover the minimal factor structure that accounts for the data, and then used multigroup Confirmatory Factor Analysis (CFA) to test measurement invariance across targets. If forcing identical loadings/intercepts degrades model fit, we can conclude that laypeople rely on distinct latent dimensions when thinking about rational versus reasonable agents.

To compare model fit of characteristics attributed to rational and reasonable persons, in Study 1a we first identified the number of factors that best explain the variance in the data. Preliminary Principal Component and Exploratory Factor Analyses (EFA) suggested a three-factor solution (per eigenvalues > 1, scree plot results, parallel analysis). We subsequently evaluated an EFA solution with oblique rotation, to allow for correlation between the 3 factors, in lavaan. We inspected the standardized loadings of each item on respective factors, seeking to identify items with highest loading and minimal cross-loading (i.e., unique loadings β > .350 on a particular factor, and Δ in βs > .350), evaluating subsequent confirmatory model fit. In evaluating model fit for the structural equation models analyzed, we employed several indices: the Comparative Fit Index (CFI) where values above .95 suggest excellent fit and values above .90 indicate an acceptable fit, the Root Mean Square Error of Approximation (RMSEA) with values below .06 indicating good fit and values below .10 indicating an acceptable fit, the Standardized Root Mean Square Residual (SRMR) with values less than .08 reflecting acceptable fit, and the Goodness of Fit Index (GFI) where values above .90 denote satisfactory fit (Hu & Bentler, 1999), thus providing a comprehensive assessment of model adequacy. We selected top four items loading on the first factor (analytical, logical, objective, smart) and the second factor (kind, empathetic, caring, moral) and top three items uniquely loading on the last factor (strong, courageous, brave), resulting in a model with excellent fit, CFI = .960, RMSEA = .081, SRMR = .059, GFI = .946.

Applying the same criteria in Study 1b (eigenvalues > 1, scree plot results, parallel analysis), a two-factor model described the data best, consistent with prior work on Agency and Communion (Bakan, 1966; Wiggins, 1991). The confirmatory 2-factor model with five items each loading on the first factor (intelligent, confident, competent, independent, competitive) and the second factor (good-natured, warm, tolerant, sincere, trusting) showed an acceptable fit, CFI = .936, RMSEA = .076, SRMR = .055, GFI = .955.

To test for measurement invariance of rational and reasonable judgments, in both studies we formally evaluated the similarity of ratings of characteristics for rational and reasonable persons. We contrasted a configural model where factor loadings and intercepts were free to vary, with a constrained model, where factor loadings and intercepts were set to be equal across the rational and reasonable groups. The comparison of these models provided insights into the degree of measurement invariance: if the constrained model did not significantly worsen the fit, it would suggest metric invariance (equivalence in factor loadings) and scalar invariance (equivalence in item intercepts).

Results

Are Reasonable Persons Perceived as Less Deliberate Compared to Rational Persons?

We examined adjective co-occurrences to assess semantic similarity of ‘rational’ and ‘reasonable’ concepts to deliberate vs. intuitive characteristics in participants’ descriptions of a person showing sound judgment (Study 1a) / good judgment in a challenging situation (Study 1b). In both versions of Study 1, participants chiefly relied on deliberate qualities such as ‘thoughtful,’ ‘level-headed,’ ‘attentive,’ or ‘discerning’ (Figure 1, also see Tables S2 and S4). Both ‘reasonable’ and ‘rational’ were more likely to be associated with ‘thoughtful’ than ‘intuitive,’ with notably greater co-occurrence of ‘thoughtful’ with ‘reasonable’ rather than ‘rational’ (Figure 1). Similar results emerged in supplementary analyses (Figures S1 and S4), when estimating co-occurrences via the positive Pointwise Mutual Information to account for the differential frequency of individual terms in the corpus. Overall, these analyses showed little evidence that reasonable persons were perceived as less deliberate compared to rational persons.

Figure 1. — Top ten most frequent co-occurrence of Adjectives with ‘rational’ (blue) and ‘reasonable’ (orange) when asked to describe most important characteristics of a person showing sound judgment (Study 1a) / good judgment in a challenging situation (Study 1b). Adjectives are ordered from those most associated with ‘rational’ to those with ‘reasonable.’ Dumbbell nodes represent the percentage of each adjective’s co-occurrence relative to the sum of independent occurrences of each pair of terms.

Rational Objectivity vs. Reasonable Understanding.

In both versions of Study 1, the Wilcoxon Signed Rank Test across all pairwise correlations of adjectives in a dataset revealed a statistically significant difference in the patterns of associations with ‘rational’ as compared to ‘reasonable,’ Study 1a: V = 24,971, p < .001; Study 1b: V = 21,273, p < .001. As Figure 1 demonstrates, the types of deliberate qualities participants associated with the rational vs. reasonable concepts varied: ‘rational’ was more likely to co-occur with adjectives representing algorithmic qualities—i.e., ‘objective,’ ‘informed,’ and ‘discerning’ (Study 1a)⁴ / ‘analytical,’ ‘unbiased,’ and ‘impartial’ (Study 1b), whereas ‘reasonable’ was more likely to co-occur with adjectives representing holistic, social qualities—i.e., ‘sensible,’ ‘understanding,’ and ‘thoughtful’ (Study 1a) / ‘caring,’ ‘judicious,’ and ‘balanced’ (Study 1b)⁵.

How Common Are Reasonable and Rational Concepts in Descriptions of Sound/Good Judgment?

Across 429 unique adjectives in Study 1a (total n = 2,320 entries), most common terms concerned ‘intelligent,’ ‘smart,’ ‘thoughtful,’ and ‘wise,’ followed by ‘rational.’ Further, top ten terms included ‘trustworthy,’ ‘logical,’ ‘honest,’ ‘calm,’ and ‘reasonable.’ ‘Rational’ and ‘reasonable’ each appeared more often than characteristics such as ‘fair,’ ‘reliable,’ ‘objective,’ and ‘practical.’ Across five adjectives participants provided, ‘rational’ was the fifth most mentioned (61 times), showing more than 10 times the expected probability of mentioning (probability of mention = .026, 95% CI [.020, .034]), significantly higher than the expected probability of .002, p < .001 (see Table S2 in the online supplement). Likewise, ‘reasonable’ was tenth most mentioned (49 times), also demonstrating an effect size that was more than 10 times the expected probability of mentioning (probability of mention = .021, 95% CI [.016, .028]), well above the expected probability of .002, p < .001.

Similarly, across 453 unique adjectives Study 1b participants used to describe a person showing good judgment in a challenging situation (total n = 2,444 entries), top ten most common terms were ‘calm,’ ‘smart,’ ‘intelligent,’ ‘thoughtful,’ ‘decisive,’ ‘wise,’ ‘honest,’ ‘patient,’ ‘fair,’ and ‘level-headed.’ ‘Rational’ and ‘reasonable’ were again significantly more frequent than most other terms, appearing more often than characteristics such as ‘ethical,’ ‘sensible,’ ‘considerate,’ and ‘flexible.’ Across five adjectives participants provided, ‘rational’ was the 11^th most mentioned (39 times; probability of mention = .016, 95% CI [.011, .022]), significantly higher than the expected probability of .002, p < .001. Likewise, ‘reasonable’ was 30^th most mentioned (16 times; probability of mention = .007, 95% CI [.004, .011]), well above the expected frequency, p < .001⁶. Overall, participants spontaneously used ‘rational’ and ‘reasonable’ to describe distinct features of a person demonstrating sound / good judgment, and they did so at the rate well above what would have been expected by chance.

Trait-Ascriptions to Rational and Reasonable Persons in Study 1a.

In the psychometric analyses of trait-ascriptions to rational and reasonable persons, we formally evaluated the comparability of the measurement models for rational vs. reasonable persons. In Study 1a, we compared the fit of the 3-factor measurement model of Analytical, Moral and Inner Fortitude latent trait-ascription factors participants relied on when evaluating rational and reasonable persons (see supplementary exploratory factor analyses indicated the best fit of a 3-factor solution; see items loading on each factor in Figure 2). Testing model fit revealed that a metrically invariant model—i.e., constraining loadings to be equal across ratings of rational and reasonable persons—did not result in a significant worsening of fit compared to unconstrained configural model, Δχ²(8) = 11.864, p = .157. However, moving to the scalar invariant model (constraining both loadings and intercepts) revealed a significantly worse fit, Δχ²(8) = 47.743, p < .001, indicating non-equivalence of item intercepts between groups. In other words, participants showed non-negligible differences in the conceptualization of Analytical, Moral an Inner Fortitude factors for rational as compared to reasonable persons, with the same score on a particular factor might represent different levels of that factor in the rational group compared to the reasonable group.

Given evidence of a metric but not a scalar measurement invariance across rational versus reasonable targets in Study 1a, we focused on item- (rather than factor-) level comparisons of means in Figure 2. The dumbbell plot on the top showcases the estimated means for trait-ascriptions by target group (also see Figure S3 and Table S3 for the 95% CI around the estimates). Because we observed a halo effect for both targets (ratings for all items were on average above scale midpoint; see top left of Figure 2), we used a somewhat more stringent criterion—one unit above scale midpoint—to evaluate our hypothesis concerning ascription of deliberative and interpersonal characteristics to rational and reasonable persons. Consistent with the open-ended results above, participants ascribed deliberate traits to both targets, with characteristics such as ‘analytical,’ ‘level-headed,’ ‘open-minded,’ ‘understanding,’ ‘thoughtful’ and ‘patient’ all rated at least one unit above scale midpoint (dashed vertical line in the dumbbell graph on the top left of Figure 2 and in Figure S3). Also consistent with the open-ended responses, we observed nuanced differences in the type of deliberate qualities participants ascribed to rational as compared to reasonable persons. Rational persons were relatively more analytical, logical, and objective than reasonable persons. Conversely, reasonable persons were relatively more likely open-minded, understanding, patient, or thoughtful than rational persons.

Replicating Grossmann et al. (2020), interpersonal characteristics such as ‘kind,’ ‘caring,’ or ‘empathetic’ were uniquely attributed to reasonable persons—respective estimates were to the right of the vertical dashed line in Figure 2 for reasonable persons, yet to the left of this dashed line for rational persons. Moreover, reasonable persons were attributed significantly higher interpersonal characteristics than rational persons for most interpersonal characteristics we tested—‘empathetic,’ ‘caring,’ ‘kind,’ ‘open-minded,’ ‘understanding,’ ‘patient,’ ‘moral,’ ‘fair,’ ‘thoughtful’ (Figure 2 and Figure S3), while we did not observe significant differences for such traits as ‘wise,’ ‘perceptive,’ ‘responsible,’ ‘brave,’ ‘sensitive,’ ‘calm,’ ‘honest’ or ‘trustworthy.’

If folk theories of reasonableness concern holistic judgment accommodating a plurality of relevant values, then perceivers should expect reasonable persons to integrate analytical and socially conscious characteristics more extensively than rational persons⁷. To test our prediction concerning greater interconnectedness of traits ascribed to reasonable vs. rational persons, we estimated Pearson’s correlation coefficients of average scores across items making up Analytical, Moral, and Inner Fortitude factors (recall, metric invariance allowed for comparisons of factor loadings; we relied on average scores across items to estimate 95% CIs, with CFA-based factor loading correlations yielding a very similar picture). Consistent with our predictions, when participants evaluated reasonable (vs. rational) persons, ratings on Analytical traits were significantly more likely to converge with ratings on Moral and Inner Fortitude traits: t_moral(497.14) = 3.24, p < .001; t_fortitude(498.85) = 3.60, p < .001 (bottom left of Figure 2). We observed no significant differences between target-specific associations between Moral and Inner Fortitude traits, t(437.13) = 1.67, p = .096. In other words, Study 1a participants perceived more interconnectedness between analytical and social-moral traits for the lay concept of reasonableness compared to the lay concept of rationality.

Trait-Ascriptions to Rational and Reasonable Persons in Study 1b.

Comparison of measurement models of Agency- and Communion-related attributions to rational and reasonable persons revealed a significant worse fit of the metrically invariant model, in which we constrained factor loadings across rational and reasonable groups, compared to the unconstrained configural model, Δχ²(8) = 15.847, p = .045, indicating significant structural differences in Agency and Communion factors when evaluating rational and reasonable persons. Following the Study 1a analytical plan, we therefore chiefly focused on item-level analyses.

Like in Study 1a, we observed a halo effect for both targets—all positively loaded items were rated above scale midpoint; see top right graph in Figure 2). Study 1b relied on a 5-point scale, thus we adopted a somewhat more stringent criterion of half-a-unit above scale midpoint to evaluate our hypothesis concerning ascription of interpersonal characteristics to reasonable versus rational persons. The top right graph in Figure 2 indicates that agentic qualities were attributed to both targets, except for ‘competitive’ (also see Table S5 for 95% CIs), with average ratings on all other agentic traits to the right of the vertical dashed line. Additionally, participants rated rational persons relatively more ‘independent’ and ‘intelligent’ than reasonable persons, and rated reasonable persons as relatively more communal on each item than rational persons. Moreover, while participants perceived both rational and reasonable persons as non-selfish, this perception was even stronger for reasonable persons, who were rated as significantly less selfish than their rational counterparts (Figure 2), conceptually replicating earlier work (Grossmann et al., 2020).

Following Study 1a method for evaluating interconnectedness of traits ascribed to reasonable vs. rational persons, we estimated Pearson’s correlation coefficients of average scores across items making up Agentic and Communal factors, along with the Selfishness item. Consistent with our predictions and Study 1a results, when participants evaluated reasonable (vs. rational) persons, ratings on Agentic traits were significantly more likely to converge with ratings on Communal traits: t(586.56) = 2.89, p = .004 (bottom right of Figure 2). We observed no significant differences between target-specific associations between ascription of selfishness and either Agentic or Communal traits, −.05 < ts < 1.05 = 1.67, ps ≥ .294. In other words, Study 1b participants perceived more interconnectedness between agentic and social traits for the lay concept of reasonableness compared to the lay concept of rationality.

Overall, Study 1 demonstrated that laypeople typically associate deliberate qualities with both rational and reasonable persons. Furthermore, when laypeople define sound and good judgment both rational and reasonable qualities frequently come to mind. People perceived both concepts as deliberate and agentic rather than intuitive, yet salient features of rational persons appeared more likely to concern bias reduction and logical analysis, whereas salient features of reasonable persons appeared to focus more on socially conscious, interpersonally sensitive characteristics. Moreover, people were more likely to integrate analytical and socially conscious characteristics when evaluating reasonable as compared to rational persons, suggesting that folk theories of reasonableness (vs. rationality) are more likely to concern holistic judgment accommodating a plurality of social and analytical judgmental competencies. Given these nuances in semantic understanding of lay rationality and reasonableness, in the next study we sought to explore how people apply these folk concepts when evaluating and selecting for distinct social roles where sound judgment is needed.

STUDY 2

When facing challenging decisions in everyday life, sometimes it may be of benefit to apply standardized rules, while other times sound judgment may call for holistic consideration of the social context and interpersonal relationships. Do preferences for rational versus reasonable agents shift systematically across a range of social roles that differ in their demands for analytic generalization versus holistic, context-specific considerations? In Study 2, we assessed preferences for rational or reasonable agents across social roles where sound judgment is needed, selecting roles that a priori emphasize analytic judgment that involves general rules or formulas (e.g., math teacher) or holistic judgment that involves context-specific interpretive work (e.g., English teacher, see Figure 3). We hypothesized that laypeople would prioritize rationality for roles, such as a math teacher, which involve tackling well-defined problems that can typically be solved using algorithmic, rule-based judgment, whereas they would prioritize reasonableness for roles, such as a school principal, that involve tackling ill-defined problems where solutions require context-sensitive balancing.

Figure 3. — Top Panel: Preference proportions for reasonable vs. rational agents in social roles. Displays proportions (with 95% CI) derived from logits in generalized mixed models. The dashed line at .50 indicates parity; above this, preference leans towards reasonable agents, and below, towards rational agents. Bottom Panel: Necessity ratings for rationality and reasonableness by role (1–5 rating scale). Shows estimated means and 95% CI. Ratings indicate moderate to high necessity (3–4) for both rationality and reasonableness across rule-based (on the right) and holistic roles (on the left).

We preregistered that people will be more likely to favor rational agents for positions requiring an algorithmic, rule-like judgment (e.g., CEO of a large company, investment banker, army captain, math teacher) and more likely to favor reasonable agents for positions requiring holistic, context-specific interpretation (e.g., parents, English teacher, school principal, romantic partner, colleague at work). Per pre-registered plan, we probed the preferences for rational and reasonable agents across different social roles by using both binary choices (pre-registered hypothesis 1) and rating-scale-based responses assessing relative preference for rational and reasonable agent for each role (pre-registered hypothesis 2). Moreover, because different roles may differentially favor salience of social and algorithmic considerations, we evaluated the extent to which appraisals of these roles in terms of different interpersonal (warmth, expected interaction frequency, trust) and algorithmic considerations (logic, algorithmic) statistically account for rational vs. reasonable actor preferences.