Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

# Moral judgements of fairness-related actions are flexibly updated to account for contextual information

## Abstract

### Context-present judgements

To characterise context-present moral judgements and identify moral norms latent in these judgements, fPCA was applied to the judgement profiles. Examples of three most prominent patterns in the data can be seen in Fig. 3. Examples of functional fits over the judgement data across a range of Decision-makers’ offers relative to Receivers’ offers can be seen in Fig. 4. The scree plot revealed a three-component solution which accounted for 91.2% of variability in the data. The first component (CP-PC1), accounting for 75.9% of variability in the data, captured inter-individual variability in judgements of the Decision-makers’ offers that were larger than those of the Receivers. Individuals with high CP-PC1 scores endorsed Decision-makers who gave upwards of $2 more than their Receiver partner did in the previous round (Fig. 5d). Most of our participants endorsed these offers, while a minority of 18.21% responded below the scale centre, judging these offers as morally bad. The second component (CP-PC2), accounting for 9.7% of variability across participants, indicated the level of lenience in moral condemnation of Decision-makers’ offers that were smaller than Receivers’ (Fig. 5e). Individuals with high CP-PC2 scores were more lenient in judging Decision-makers’ offers upwards of$3 less than their Receiver partners gave in the previous round. To note, almost all participants qualified these offers as bad, and only a small minority (0.63%) responded just above the scale centre on average. The third component (CP-PC3), accounting for 5.6% of variability, indicated individual variability in judgements of Decision-makers’ offers that were similar to Receivers’ (Fig. 5f). Participants with high CP-PC3 scores endorsed Decision-makers who had given the same amount to the one Receivers gave in the previous round. Most of our participants endorsed these offers, with 77.95% responding above the scale centre on average. Notably, however, 32.06% of participants judged similar offers as morally better than higher offers on average.

These three principle components divided the variability in judgements of Decision-maker offers, considered relative to the Receiver’s previous offers, into three parts—relatively higher, relatively smaller, and similar offers. These parts can be interpreted as reflecting three context-dependent norms, namely relative generosity, relative selfishness and indirect reciprocity. As shown in Fig. 5d–f participants varied in how strongly they endorsed and condemned the Decision-maker’s contextualised actions in line with these norms. To explain, CP-PC1 and CP-PC3 appear to reflect the general tendency to morally endorse relative generosity and indirect reciprocity, respectively, while CP-PC2 appears to reflect general lenience in judging relative selfishness.

On average, participants adjusted their responses after receiving contextual information by 16.72% of the scale length in either direction (SD = 7.28, 95% CI [15.92, 17.53]). Participants changed the valence of their judgement in an average of 22.08% of completed trials (SD = 15.28, 95% CI [20.66, 23.49]).

Similar to those components identified in analyses of context-present judgements, the principle components identified here divide the variability in judgement adjustments into three parts, corresponding to three context-dependent norms: relative generosity, relative selfishness, and indirect reciprocity. The component scores can thus be interpreted as measures of importance people assign to these context-dependent norms. However, this interpretation has an important caveat. In instances in which the initial evaluation strongly relied on context-independent norms, and participants made a response close to the limits of the response scale, there was no option for further adjustment in the same direction. In other words, the extent of adjustments that could be made in a particular direction was somewhat dependent on the initial response, and potentially limited for extreme judgements. For example, if a participant maximally endorsed generosity in their context-absent response, they could not give an even more positive response to relative generosity, due to the bounded limits of the scale. Therefore, stronger expressions of these components might (also) be related to participants’ subjective importance of context-dependent norms over context-independent norms, whereby relevant contextual information is anticipated by withholding extreme positive or negative judgements, leaving room for subsequent adjustment once the context is presented.

### Relations between context-absent and context-present judgement patterns

After characterising the dominant patterns of moral judgements in terms of moral norms, and the importance that people assign to each when making moral judgements, we investigated how the importance of norms correlated across the two judgements. To do so, we correlated the scores of principle components obtained from the first and second judgements. As our principle component scores were not normally distributed, non-parametric Spearman’s rank correlation coefficients were calculated. Overall, there were strong correlations between importance of context-independent norms and their relative context-dependent counterparts. CA-PC1 was positively correlated with CP-PC1 (rs = 0.461, p < 0.001), indicating that those who endorsed high offers in context-absent judgements also endorsed relatively high offers in context-present judgements. CA-PC2 was positively correlated with CP-PC2 (rs = 0.557, p < 0.001), indicating that those who were lenient in judging low offers in context-absent judgements were also lenient in condemning relatively low offers in context-present judgements. CA-PC3 was positively correlated with CP-PC3 (rs = 0.470, p < 0.001), indicating that those who endorsed even split in context-absent judgements also endorsed matching offers in context-present judgements. These correlations between importance of context-independent norms and their relative context-dependent counterparts were stronger than correlations between context-independent norms and other context-dependent norms (e.g., CA-PC1 and CP-PC1 correlation was higher than that between CA-PC1 and CP-PC2, and that between CA-PC1 and CP-PC3). A full matrix of correlations between judgement patterns is included in the Supplementary Table S1.

### Relations between context-absent patterns and updating patterns

In the sections above we report that those who were lenient in judging low offers in context-absent judgements were consistently lenient in condemning relatively low offers in context-present judgements (i.e. a positive relationship between CA-PC2 and CP-PC2). Conversely, we also report that those lenient in judging low offers in their context-absent judgment, also had stronger negative adjustments once they found out that the offer was relatively low (i.e. a negative relationship between CA-PC2 and Adj-PC2). One potential explanation for these opposite effects is that they coexist independently in our sample due to two different trajectories of progression of two judgements that participants took: (1) some of those that were lenient in condemning low offers in context-absent judgement were also overall lenient in condemning relatively low offers; (2) others who were lenient in condemning context-absent judgements, adjusted more negatively to condemn relatively low offers. To test this idea, we used a multiple linear regression analysis to predict CA-PC2 based on CP-PC2 and Adj-PC2. This regression model was statistically significant F(2,310) = 1299, p < 0.001, with an R2 = 0.893. There were significant effects of both CP-PC2 (β = 0.736, SE = 0.017, t = 39.07, p < 0.001), and CP-Adj2 (β =  − 0.738, SE = 0.019, t = 39.15, p < 0.001), providing evidence that indeed both judgement tendencies co-existed in our sample.

## Discussion

We developed a novel task in which participants made moral judgements of others’ actions. In this task participants observed the outcomes of a Dictator game, in which a Decision-maker shared a proportion of 10 dollars with a Receiver. Participants judged different Decision-makers’ offers in absence of any context and made subsequent judgements after they were presented with contextual information related to this action, i.e. the Receivers’ past offers when acting as a Decision-maker in a previous round of the game. We found that people adjusted their moral judgements after learning contextual information (by 16.62% of the scale length on average across all trials), and they did so in a systematic way. Participants flexibly transitioned from relying on context-independent norms (e.g., generosity, equality) to relying on context-dependent norms (e.g., relative generosity, indirect reciprocity) after learning this contextual information. Our findings form the basis of a novel framework for understanding how people update their moral judgements of sharing behaviour as they learn new contextual information.

### Judgement profiles before and after learning relevant contextual information

Participants were remarkably consistent in their judgement patterns across the testing session, yet there were stark differences in judgement styles among participants. The principle components derived from these judgement patterns captured the bulk of the inter-individual variability, allowing us to identify judgement styles that were influenced by several distinct fairness-related norms. For judgements made before contextual information had been presented, three separate components characterised how individuals judged low (< $4), approximately even-split, and high (>$6) offers by the Decision-maker. These three components can be seen as an organizing framework for people’s moral preferences in simple distributive justice problems and can be interpreted as the extent to which participants’ judgements relied on selfishness, equality and generosity norms. Critically, our design and analysis strategy allowed us to estimate the relative importance of these norms for each individual. A majority of participants found the generosity norm important and judged high offers as good. However, a smaller portion of participants found equality to be more important than generosity. These individual differences are compatible with previous research showing that multiple norms can be used to distribute resources, and in particular that both generosity and equality norms influence people to act against their economic interest and give up money38,40,44,50, as well as with previous research showing that norms influence people’s moral decisions when their economic interest is not at stake30,53. Our findings extend this to moral judgements, showing that in situations where economic interest is not at stake, generosity and equality norms have varying degrees of influence across individuals in their third-party moral judgements. We also found that in addition to generosity and equality, selfishness permits multiple interpretations. Previous research examining sharing behaviour has not discussed norms for selfishness as separate from equality and generosity, and sharing low portions has been interpreted as behaviour motivated by self-interest44,50. In our task, participants’ judgements posed no monetary cost to them, in contrast to the first-person dictator game, yet still some participants were more lenient in judging low offers than others, and this variability was orthogonal to variability associated with high and medium offers. This indicates that, in addition to norms endorsing generosity and equality, norms condemning selfishness might also shape individuals’ preferences for distributive justice. These norms condemning selfishness may impact people’s sharing behaviours and judgements independently from self-interest, and future studies should explore this possibility by correlating the importance of selfishness norms with measures of sharing behaviour.

For judgements made after learning relevant contextual information, we identified three components which corresponded to the influence of context-dependent norms. Individuals showed substantial variation along principle components corresponding to judgements of relatively higher, relatively lower, and similar offers relative to what the Receiver had previously given. These principle components can be seen as an organizing framework for people’s moral preferences in distributive justice problems in which information about the deservingness of each individual is salient. These principle components can be interpreted as reflecting the importance of three context-dependent norms (relative generosity, relative selfishness and indirect reciprocity). Our results suggest that, while most participants relied on each of these norms, the relative importance of each norm was not equally distributed and varied across individuals. Relative generosity was more important than indirect reciprocity to most people, but a notable portion of people had the opposite preferences (32%). Some of these people also condemned relatively higher offers, which may reflect the importance that people assign to retribution (e.g., that someone who behaved poorly in the past should not receive generosity but should be punished). Our principle component findings are in line with previous research showing that people account for deservingness of moral actors when sharing resources by either reciprocating45,46,47, or exceeding the past favour of their partner57,58. Moreover, our findings that a portion of people found rewarding selfish actions to be morally inappropriate, are in line with previous research showing strong individual differences in punishment of selfishness42,59,60. Variation in punishment across individuals, with only small number of punishers has been shown to be beneficial for reinforcing moral standards and sustaining cooperation within larger groups61,62.

### Transitions from context-independent to context-dependent norms

Our results show that participants who endorsed high offers in context-absent judgements also endorsed relatively high offers in context-present judgements. Similarly, those who endorsed even-split offers in context-absent judgements also endorsed similar offers in context-present judgements, and adjusted more strongly to endorse similar offers. These results are in line with our hypothesis that people flexibly switch from relying on context-independent norms (generosity, selfishness, equality) to relying on related context-dependent norms (relative generosity, relative selfishness, indirect reciprocity) depending on the availability of the context, in order to maintain coherence in judgements with an overarching moral principle, or virtue. This finding supports the idea that the set of norms on which people rely is susceptible to change depending on the availability of contextual information. This finding is also in agreement with previous research showing that fairness norms change within minutes if circumstances are altered51. Overall, the norms that exert the greatest influence over moral judgements appear to be highly dependent on information available, and that to understand moral norms of various contextual contingencies, we need to study how contextual information modulates which norms we place importance on, adding one piece of information at a time. Nuanced manipulations of contextual information could be highly valuable for advancing our understanding of other context-dependent norms both in distributive justice (e.g., equity, reciprocity) as well as other moral domains (e.g., revenge, defence).

Our findings outline systematic relationships between the importance of specific context-independent norms and specific context-dependent norms to individuals. While much of the existing literature conceptualises context-dependent and context-independent norms as opposites48,49, the strong relations in our study contrarily may reflect general virtues or values. For example, the observed relationship between the endorsement of equality and endorsement of indirect reciprocity may reflect a general tendency to endorse balanced (reciprocal and equal) distributions of resources. Endorsement of generosity and endorsement of relative generosity may reflect general moral endorsement of the generosity virtue. These coherent relationships within virtues were stronger than relationships between unpaired norms. This interpretation introduces the idea of a latent order (i.e. a taxonomy) of context-dependent and context-independent moral norms. So far moral psychology has systematically studied individual differences in the importance people assign to broader domains (e.g., moral foundations), with fairness being only one domain compared with harm, purity, loyalty, and authority63. Within the fairness domain, there have been studies showing strong differences in sharing behaviour across individuals48,49,50. However, individual differences in fairness virtues have not been studied systematically. One possible interpretation of our results is that more abstract concepts such as a generosity virtue, or a symmetry virtue (equality, indirect reciprocity) guide our moral cognition of sharing, and that appropriate norms are used in situations where contextual information is or is not available. Future research may investigate whether the importance people assign to virtues in judgement of simple sharing scenarios generalise to other and more complex distributional justice situations in experiments as well outside the lab.

### Individual preferences for context-dependent norms

Participants’ responses to low offers (< $4) showed a different updating pattern to that described in the previous section. Our results suggest that there were two updating trajectories that participants took: (1) some of those that were lenient in condemning selfishness in context-absent judgement were also lenient in condemning relatively low offers, which may reflect the underlying context-invariant virtue for condemning selfishness, and (2) others who were lenient in condemning context-absent low offers adjusted more negatively to condemn relatively low offers after learning the context. These findings support the view that some participants withheld their judgements when faced with a low offer and adjusted more strongly once they learned the context, indicating that these individuals found context-dependent norms (relative-selfishness) more important than context-independent norms (selfishness) even before the context was revealed. These findings are in line with previous research showing that people rely on context-dependent norms when additional contextual information exists48. However, since withholding of judgements primarily occurred in trials with low offers, this raises a question why low offers are special in this regard. One possibility is that this is related to the negative valence of these judgements. Speculatively, judgements that someone’s action is “morally bad” may require more evidence and a higher degree of certainty, as condemnation can be socially costly (as is the case for blame judgements), and can damage relationships21,64. Because it is important for people to avoid incorrect or premature blaming, people may adopt context-dependent norms immediately in their judgement of low offers when they are aware that contextual information exists. When it comes to judgements of high offers, mistaking actions as morally good may not be as important, and context-independent norms favouring positive judgement can be adopted with less concern for context that is still unavailable. ### Limitations and future directions Our findings regarding relationships between initial judgement styles and the extent of judgement adjustments should be interpreted with caution, as the level of noise in the data (e.g., due to imprecisions in mouse clicking) could potentially boost or mask these relationships. We explained this potential issue in the “Results” section. However, we also provided evidence that the observed relationships were not strongly affected by this noise by demonstrating that correlation coefficients remained very similar when we excluded participants for which we suspected this noise to have the highest influence on ratings. Future studies may attempt to resolve this issue by altering the design and incorporating a measure of participants’ desire to alter their previous judgement positively or negatively after contextual information is presented (for example see ref.65). This would give participants the opportunity to signal the importance of contextual factors in trials where they had initially made judgements at the extreme ends of the good/bad spectrum. Another limitation of this study was that some participants might have had suspicions around whether they are judging actions of real people. Although we found that this was not the case in a small piloting sample, there is a chance that some people in our much larger study sample had doubts. Participants’ trust of the cover story is a very important factor for typical economic games paradigms in which people make decisions with economic implications. However, this might be less important for the study of third-party moral judgements, as morality researchers typically present imaginary scenarios to people28. Moreover, any judgement patterns that may hypothetically result from a realisation that the cover story is not true (e.g., random or invariant responding, change in strategy), were included as exclusion criteria in this study. Future studies could closely investigate potential differences between third party moral judgements of real and imagined scenarios. Another limitation relating to generalisability of our findings is that, while making initial judgements, our participants had expectations that they were going to learn contextual information, but moral judgement updating in everyday life may often occur without such expectancy. This expectancy may modulate both initial judgements as well as contextual updates. For example, this expectancy may make individuals perceive the information available as incomplete, initial judgements as imperfect, and result in more cautious decision-making. Additionally, this expectancy may prepare participants to adjust their judgements, inducing demand effects. Ideally, moral judgement updating should also be studied in scenarios in which information changes are not expected; however, this is difficult to study in multi-trial tasks, as repeated introduction of contextual information across trials inevitably gives rise to expectations. Future studies could also consider manipulating these expectancies parametrically in blocked or between-subject designs. Another limitation of this study was that we could not reliably measure response times, due to multiple sources of noise associated with responding using a computer mouse. Response times may offer additional insight into the temporal dynamics of the moral judgement process and improve our understanding of the processes involved in judgement updating (e.g., anticipation and integration of contextual information). Future studies may use more controlled mouse-click designs or button presses to measure response times. Finally, since our study methods were not publicly preregistered before the experiment was conducted, it would be beneficial for future research to replicate our findings using alternative methods and stimuli. Our findings suggest that people substantially adjust their judgements in information-dynamic situations, which is important to consider for the future development of moral judgement theories. Extant theories of moral judgement4,5,6 implicitly assume that judgements occur in information-static environments, in which all the necessary information is already available to the decision-maker. To note an exception, Social Intuitionist Model (SIM) specifies how moral judgements change, however only on a longer timescale, and in response to social influence rather than learning new information6. Moral judgement theories would benefit from explicitly considering how different morally relevant information is dynamically integrated into judgements. For example, in relation to dual process theory, future research should consider whether consequence-relevant and rule-relevant information is differently integrated in real time. In relation to SIM, it would be interesting to consider whether harm and fairness information is more or less likely to lead to context-specific judgment adjustments than information related to binding foundations (ingroup loyalty, authority, and purity). ### Conclusion Our findings show that most participants flexibly switch from relying on context-independent norms (generosity, selfishness, and equality), to relying on context-dependent norms (relative generosity, relative selfishness, and indirect reciprocity), as they successfully integrate contextual information. Participants do this in a consistent manner, which suggests that the organizing framework for distributive justice norms has three dimensions reflecting broader virtues: generosity, condemnation of selfishness, and balance. When it comes to judgement of low offers, we present evidence that some participants were particularly careful not to prematurely judge actions as bad without knowing contextual information and adjusted their judgements more strongly when they learned the context. These findings demonstrate that people adjust their moral judgements when they find out new contextual information, switching between different fairness norms that underpin their decisions. ## Methods ### Participants The sample consisted of 371 University of Melbourne undergraduate students (263 female, 108 male, Mage = 20.57, SD = 2.36, range: 18–36 years), who participated in an online study for course credit. The sample size was determined using sensitivity analyses66 for two-tailed hypothesis testing (α = 0.05) with 80% power67 to detect correlation effect sizes larger than r = 0.15, accounting for an expectation that a part of the sample (around 10%) will be excluded. The minimum effect size was chosen as to detect small effect sizes, which are common in individual differences research68. Fifty eight people were excluded based on predefined data quality criteria: 7 participants failed an attention-check (had given wrong answers in more than 30% of catch trials—see below), 27 participants had responded to over 30% of attention-check questions in the questionnaires incorrectly, 2 participants placed their response in the middle of the scale throughout the task (i.e., did not show any variation in their responding at all, suggesting disengagement with the task), 1 participant did not show any coherence in their responding across identical trials (suggesting random responding), 3 participants drastically changed their context-absent judgement patterns partway through the task (suggesting misunderstanding the task for a substantial period of time), 1 participant consistently rated giving larger amounts as bad and smaller amounts as good (suggesting confusion about scale endpoints, or deliberately providing unexpected behaviour inconsistent with anyone else), and 17 participants misinterpreted the task and judged the Receiver’s as opposed to Dictator’s action (their response correlated positively with the amount that Receiver had given with r > 0.4). Given that we did not preregister the exclusion criteria, we repeated all relevant analyses with the full sample. All results reported below replicate (with slightly weaker, but nevertheless significant coefficients; data not shown), demonstrating that our exclusion criteria did not bias the results. The final sample consisted of 313 participants (222 female, 91 male, Mage = 20.66, SD = 2.51, range 18–36 years). All participants in this study provided a written informed consent via an online form. The study was approved by the Human Research Ethics Committee of the Melbourne School of Psychological Sciences (Ethics ID 1750046), and all methods were performed in accordance with the relevant guidelines and regulations. ### Experimental paradigm The experiment consisted of three stages. First, participants read a cover story (described below) and a description of the task, and then completed a comprehension test. These instructional materials are available in the supplementary materials (Supplementary Methods S2). Once they passed the comprehension test, they proceeded to complete the experimental task. Finally, they completed a set of questionnaires. #### Cover story and overview Participants were presented with a cover story in which they learned about a recently conducted experiment investigating people’s economic decisions. This experiment was entirely fictional, however participants were not informed of this. Participants were informed that their task in the current study would be to morally evaluate these people’s actions in two consecutive steps. In this fictional experiment, a group of people completed a variant of the dictator game consisting of two rounds (see Fig. 1). In the first round, people participating in the experiment were randomly assigned roles of the Dictator and the Receiver, and the Dictator was given$10. The Dictator was termed “the Decision-maker”. The Decision-maker decided on the proportion of the $10 that they would give to the Receiver ($x). In the second round, the Round 1 Decision-maker became the Round 2 Receiver and was assigned a new partner, who was then also given $10. The Round 2 Decision-maker decided how much to give to the Receiver ($y). Importantly, the Round 2 Decision-maker was able to see how much their partner, the current Receiver, had given in the first round as the previous Decision-maker.

In the current (real) study, participants were instructed to judge only the action of the Decision-makers in Round 2 (i.e. the sharing behaviour of Decision-makers, who themselves had a priori knowledge of the previous sharing behaviours of their respective assigned Receivers). Importantly, they were required to make two consecutive judgements of the same sharing behaviour: For the first judgment, they were only shown how much the Decision-maker shared with the Receiver ($x). To give their moral judgement, participants used a continuous scale with endpoints labelled as “bad” and “good”, but no further visual cues. Importantly, participants knew at this stage that the Decision-maker made this decision knowing how much the Receiver had given to their respective partner ($y) in the previous round playing the same game. Following this initial, context-absent judgement, the amount that the Round 2 Receiver had given in Round 1 (when playing as the Decision-maker) was revealed, and participants were again asked to make the same judgement regarding the Decision-maker’s action as before, again using the continuous scale. Participants judged the same behaviour (i.e. the morality of the current Decision-maker’s sharing behaviour), and the only difference was that participants knew what the Decision-maker “knew” all along: the context in which the Decision-maker’s action occurred.

We piloted this cover story (and the entire experimental protocol) on a separate sample of 18 participants. The piloting session ended with a verbal interview in which the experimenter asked questions concerning believability of the cover story: “Did you have any doubts about the experiment?” and “Was there anything that you found suspicious about the study and the instructions?”. These participants did not report any suspicion that the cover story was not real.

Participants were asked to observe a series of independent transactions that various Decision-makers made towards various Receivers, always first without knowledge of the context, and then again after being provided with this context (i.e. the amount $y that the Receiver gave in the previous round, that would have informed the current Decision-maker’s sharing decision). Each trial started with a fixation cross. Next, participants saw a screen displaying how much the current Decision-maker (in Round 2 of the fictional cover study) had given to the current Receiver: “Decision-maker gives:$y”, where y was a number ranging from 0 to 10. The amount that the Decision-maker gave was drawn from a uniform distribution across trials, such that all possible dollar values (no cent amounts) were covered. The response scale with “good” and “bad” labels on each end of the scale was simultaneously presented below the transaction description. Participants indicated their (context-absent) judgement on the scale by clicking at a location between the endpoints using a mouse cursor, and these judgements were mapped onto a numerical scale ranging from − 150 to 150. Next, participants saw the second evaluation screen, providing the context. This screen was identical to the first, but added the crucial information from the previous fictional round: “Receiver gave: \$x”, where x was a number ranging from 0 to 10. All possible combinations of x and y values were presented once during the experiment in a randomised order. Participants made a second judgement using the continuous scale, after which the screen was cleared, and a fixation cross marked the start of a new trial.

There were 121 trials in the experiment. Additional attention-check trials were dispersed throughout the task in which participants were required to report the dollar amounts seen in the preceding trial. Participants knew that these trials could occur at random times during the experiment. The attention check trial screen contained the words: “Decision-maker” and “Receiver”, as well as sequences of numbers from ranging from 0 to 10 under each word, referring to the dollar amounts from the previous trial, and participants were instructed to select their answer using their mouse cursor.

Before starting the main task, each participant read the cover story and learned about the task. Participants’ understanding of the task was tested at the end of the instruction session by a comprehensive quiz. The quiz questions specifically tested participants’ knowledge of the rules of the (fictional) two-round dictator game, as well as their comprehension of the two-stage trial structure (for details see the Supplementary Material S2). To ensure full understanding of the task, participants who did not answer all the questions correctly were asked to read the instructions again. Individuals who failed the quiz three times were not invited to complete the experiment.

#### Questionnaires

Various questionnaires were administered after the task procedures. We will analyse and report the questionnaire results in a separate publication, given that this is beyond the scope of the present paper. For completeness, the measures are listed below: We administered the Big Five Inventory 2 (BFI2)69, the agreeableness section of the HEXACO Personality Inventory-Revised (HEXACO)70, the Questionnaire of Cognitive and Affective Empathy (QCAE)71, the Moral Foundations Questionnaire (MFQ), a brief set of self-report measures for political orientation72, the Social Dominance Orientation scale (SDO)73, and the Global Belief in a Just World Questionnaire (GBJWQ)74, and basic demographic measures.

### Analyses of moral judgement patterns

#### Functional principle component analysis

With the goal of characterizing moral judgement patterns across individuals and identifying moral norms (i.e. latent variables giving rise to these patterns of judgements), we used Functional Principle Component Analysis (fPCA)75,76. We used the ‘fda’ toolbox v2.2.677 implemented in R v3.5.0. This approach allowed us to fit smooth but flexible spline functions to the patterns of judgements for each participant using a fixed functional form, and to identify dominant modes of variation in the data. There are several benefits of using fPCA: this approach treats the entire spline fit of a participant as a single entity (i.e. a function producing the data), circumventing the problems of data sparsity in high dimensional space present in traditional principle component analyses (PCA) methods78,79. The fPCA approach also imposes smoothness constraints and penalties to the data fitting procedure76,78. This is necessary because PCA approaches as more commonly used are vulnerable to outliers and may result in principle components that are driven by these outliers, often masking meaningful variability in the data. The constraints and penalties implemented in fPCA minimise the impact of these outliers by accounting for neighbouring data points, and constitute the mathematical implementation of the assumption that true effects of moral norms on judgements across parametrically modulated stimuli will be smooth. This approach has so far been applied in a wide range of fields including medicine, biology, econometrics, and climate studies80. In psychology, it has been used to study sources of variation in emotion intensity profiles81, and the sources of variation in judgements of tension in music82. The current study extends the use of fPCA to judgements across parametrically modulated moral stimuli, with the aim of studying fairness norms latent in moral judgement.

##### Context-absent judgement

To characterise patterns in context-absent judgements across parametrically modulated actions (i.e. sharing different amounts of money), and to identify main components of variability in these patterns, we performed a fPCA. We specified a B-spline basis system with 5 interior knots on points equally spaced over the interval [0, 10] (i.e. the range of values that Decision-maker had given). Each segment in the system was a 4th order polynomial function, resulting in a 9th order basis. The number of knots in the basis system was chosen to be smaller than the number of values that Decision-maker had given in order to impose mild smoothness constraints to the function over the range of values. For each participant, we computed the median judgement across trials (ranging from bad to good) for each value given by the Decision-maker. We fitted the specified spline function system to the data of each participant and performed PCA using these functions as input data. We determined the appropriate number of components by inspection of the scree plot and applying the elbow-criterion (similar to ref.81), which means identifying the component solution after which including additional components does not lead to explaining any more significant proportions of the variability. The resulting component solution was rotated using the varimax rotation technique.

##### Context-present judgement

To characterise patterns in context-present judgements across parametrically modulated contextualised actions, and to identify main components of variability in these patterns, we again performed fPCA. To prepare the data for the analysis, we computed median judgements for each magnitude of Decision-maker’s sharing relative to the amount that the Receiver had previously given. The relative sharing of the Decision-maker ranged from − 10 (i.e. Decision-maker gave 0, Receiver gave 10), to 10 (i.e. Decision-maker gave 10, Receiver gave 0). We specified a 9th order B-spline system with 5 interior knots equally spaced over the interval [− 10, 10) (i.e. the relative sharing range). We fitted the defined spline system to the median data (for examples see Fig. 4) and performed fPCA with varimax rotation of the final component solution after applying the same criteria as described above.

To characterise patterns in judgement adjustments (i.e. the extent to which people shift their judgement towards “good” or “bad” after learning the contextual information) across parametrically modulated contextualised actions, and to identify main components of variability in these patterns, we again performed fPCA. Similar to analyses of context-present judgements, we specified a 9th order B-spline system with 5 interior knots equally spaced over the interval [− 10, 10]. For each participant we computed judgement adjustment by subtracting the initial context-absent judgement from the final context-present judgement and computed the median judgement adjustment for each magnitude of Decision-makers’ sharing relative to the Receiver (− 10 to 10). We fitted the defined spline system to the median data, and the resulting splines depicted the shift in judgements (ranging from bad to good) for the context-present judgement relative to the context-absent judgement. We performed fPCA with varimax rotation, as described above.

### Quantifying the extent of decision adjustments

Our first aim was to assess whether and to what degree individuals adjusted their judgements. To do so, we derived two simple measures of adjustment for each participant. The first measure was the average absolute adjustment. This measure was calculated as the absolute difference between context-absent and context-present judgements for every trial. This difference was expressed as a percentage of the total scale length, averaged across trials, resulting in the measure of average absolute adjustment for each participant. The second measure was the percentage of trials in which adjustment resulted in a change in judgement valence—i.e., changes in judgements from the left to the right side of the scale, or vice versa. We calculated the means of these two measures within our sample and estimated the corresponding 95% confidence intervals.

### Correlation analyses

The second aim was to examine whether updating is characterised by systematic shifts between pairs of context-independent and context-dependent norms. First, to investigate the relation between context-independent and context-dependent norms, we correlated our measures of importance of specific norms across context-absent and context-present judgements. These analyses captured the relationships between individual differences in context-absent and context-present judgements, treating these two judgements as independent from each-other. As such, these analyses tested whether people who used one norm in context-absent judgements tended to employ another specific norm in context-present judgements. To study individual differences in trial-wise adjustment dynamics, we also calculated the level of adjustment by subtracting the context-absent judgement from the context-present judgement on each trial. As described above, we characterised the individual variation in patterns of context-present adjustment across a continuous range of actions and contexts, using fPCA. To investigate norm shifts across trials—i.e. whether people who used one norm in context-absent judgment tended to adjust their decisions in a particular way in response to contextual information, we correlated the respective principle component scores of context-absent judgements and adjustments.

## Data availability

The datasets generated and analysed during the current study are available in the OSF repository, https://doi.org/10.17605/OSF.IO/XCBUH.

## Code availability

The code used for the analyses presented in this paper are available in the OSF repository, https://doi.org/10.17605/OSF.IO/XCBUH.

## References

1. 1.

Kohlberg, L. Essays on Moral Development, Volume 1: The Philosophy of Moral Development (Harper & Row, San Fancisco, 1981).

2. 2.

Greene, J. D. & Haidt, J. How (and where) does moral judgement work?. Trends Cogn. Sci. 6, 517–523 (2002).

3. 3.

Greene, J. D. et al. Pushing moral buttons: The interaction between personal force and intention in moral judgment. Cognition 111, 364–371 (2009).

4. 4.

Greene, J. D. An fMRI investigation of emotional engagement in moral judgment. Science 293, 2105–2108 (2001).

5. 5.

Greene, J. D. The rat-a-gorical imperative: Moral intuition and the limits of affective learning. Cognition 167, 66–77 (2017).

6. 6.

Haidt, J. The emotional dog and its rational tail: A social intuitionist approach to moral judgment. Psychol. Rev. 108, 814–834 (2001).

7. 7.

Haidt, J. & Baron, J. Social roles and the moral judgement of acts and omissions. Eur. J. Soc. Psychol. 26, 201–218 (1996).

8. 8.

Simpson, A., Laham, S. M. & Fiske, A. P. Wrongness in different relationships: Relational context effects on moral judgment. J. Soc. Psychol. 156, 594–609 (2016).

9. 9.

Miron, A. M., Warner, R. H. & Branscombe, N. R. Accounting for group differences in appraisals of social inequality: Differential injustice standards. Br. J. Soc. Psychol. 50, 342–353 (2011).

10. 10.

Sawaoka, T., Newheiser, A.-K. & Dovidio, J. F. Group-based biases in moral judgment: The role of shifting moral standards. Soc. Cogn. 32, 360–380 (2014).

11. 11.

Olson, J. G., McFerran, B., Morales, A. C. & Dahl, D. W. Wealth and welfare: Divergent moral reactions to ethical consumer choices. J. Consum. Res. 42, 879–896 (2016).

12. 12.

Feather, N. T. Judgments of deservingness: Studies in the psychology of justice and achievement. Personal. Soc. Psychol. Rev. 3, 86–107 (1999).

13. 13.

Falk, A., Fehr, E. & Fischbacher, U. Testing theories of fairness—Intentions matter. Games Econ. Behav. 62, 287–303 (2008).

14. 14.

Feather, N. T. & Deverson, N. H. Reactions to a motor-vehicle accident in relation to mitigating circumstances and the gender and moral worth of the driver. J. Appl. Soc. Psychol. 30, 77–95 (2000).

15. 15.

Samuelson, W. & Zeckhauser, R. Status quo bias in decision making. J. Risk Uncertain. 1, 7–59 (1988).

16. 16.

Voigt, K., Murawski, C. & Bode, S. Endogenous formation of preferences: Choices systematically change willingness-to-pay for goods. J. Exp. Psychol. Learn. Mem. Cogn. 43, 1872–1882 (2017).

17. 17.

Voigt, K., Murawski, C., Speer, S. & Bode, S. Hard decisions shape the neural coding of preferences. J. Neurosci. 39, 718–726 (2019).

18. 18.

Jonas, E., Schulz-Hardt, S., Frey, D. & Thelen, N. Confirmation bias in sequential information search after preliminary decisions: An expansion of dissonance theoretical research on selective exposure to information. J. Pers. Soc. Psychol. 80, 557–571 (2001).

19. 19.

Tversky, A. & Kahneman, D. Judgment under uncertainty: Heuristics and biases. Science 185, 1124–1131 (1974).

20. 20.

Hogarth, R. M. & Einhorn, H. J. Order effects in belief updating: The belief-adjustment model. Cogn. Psychol. 24, 1–55 (1992).

21. 21.

Monroe, A. E. & Malle, B. F. People systematically update moral judgments of blame. J. Pers. Soc. Psychol. 116, 215–236 (2019).

22. 22.

Gürcay, B. & Baron, J. Challenges for the sequential two-systems model of moral judgment. Think. Reason. 23, 49–80 (2017).

23. 23.

Baron, J. & Gürçay, B. A meta-analysis of response-time tests of the sequential two-systems model of moral judgment. Mem. Cognit. 45, 566–575 (2017).

24. 24.

Koop, G. J. An assessment of the temporal dynamics of moral decisions. Judgm. Decis. Mak. 8, 527–539 (2013).

25. 25.

Stanley, M. L., Dougherty, A. M., Yang, B. W., Henne, P. & De Brigard, F. Reasons probably won’t change your mind: The role of reasons in revising moral decisions. J. Exp. Psychol. Gen. 147, 962–987 (2018).

26. 26.

Byrne, R. M. J. & Timmons, S. Moral hindsight for good actions and the effects of imagined alternatives to reality. Cognition 178, 82–91 (2018).

27. 27.

Christensen, J. F. & Gomila, A. Moral dilemmas in cognitive neuroscience of moral decision-making: A principled review. Neurosci. Biobehav. Rev. 36, 1249–1264 (2012).

28. 28.

Hofmann, W., Wisneski, D. C., Brandt, M. J. & Skitka, L. J. Morality in everyday life. Science 345, 1340–1343 (2014).

29. 29.

Clifford, S., Iyengar, V., Cabeza, R. & Sinnott-Armstrong, W. Moral foundations vignettes: A standardized stimulus database of scenarios based on moral foundations theory. Behav. Res. Methods 47, 1178–1198 (2015).

30. 30.

Konow, J. Is fairness in the eye of the beholder? An impartial spectator analysis of justice. Soc. Choice Welfare 33, 101–127 (2009).

31. 31.

Fehr, E. & Schmidt, K. M. A theory of fairness, competition, and cooperation. Q. J. Econ. 114, 817–868 (1999).

32. 32.

Crockett, M. J., Kurth-Nelson, Z., Siegel, J. Z., Dayan, P. & Dolan, R. J. Harm to others outweighs harm to self in moral decision making. Proc. Natl. Acad. Sci. U.S.A. 111, 17320–17325 (2014).

33. 33.

Benz, M. & Meier, S. Do people behave in experiments as in the field?-Evidence from donations. Exp. Econ. 11, 268–281 (2008).

34. 34.

Kvaran, T. & Sanfey, A. G. Toward an integrated neuroscience of morality: The contribution of neuroeconomics to moral cognition. Top. Cogn. Sci. 2, 579–595 (2010).

35. 35.

Baumard, N., André, J.-B. & Sperber, D. A mutualistic approach to morality: The evolution of fairness by partner choice. Behav. Brain Sci. 36, 59–78 (2013).

36. 36.

Adams, J. S. Inequity in social exchange. In Advances in Experimental Social Psychology (ed. Berkowitz, L.) 267–299 (Academic Press, Cambridge, 1965).

37. 37.

Messick, D. M. & Schell, T. Evidence for an equality heuristic in social decision making. Acta Psychol. (Amst.) 80, 311–323 (1992).

38. 38.

Shapiro, E. G. Effect of expectations of future interaction on reward allocations in dyads: Equity or equality. J. Pers. Soc. Psychol. 31, 873–880 (1975).

39. 39.

Deutsch, M. Equity, equality, and need: What determines which value will be used as the basis of distributive justice?. J. Soc. Issues 31, 137–149 (1975).

40. 40.

Hysom, S. J. & Fişek, M. H. Situational determinants of reward allocation: The equity-equality equilibrium model. Soc. Sci. Res. 40, 1263–1285 (2011).

41. 41.

Diekmann, A. The power of reciprocity: Fairness, reciprocity, and stakes in variants of the dictator game. J. Confl. Resolut. 48, 487–505 (2004).

42. 42.

Fehr, E. & Fischbacher, U. Third-party punishment and social norms. Evol. Hum. Behav. 25, 63–87 (2004).

43. 43.

Mandel, D. R. Economic transactions among friends: Asymmetric generosity but not agreement in buyers’ and sellers’ offers. J. Confl. Resolut. 50, 584–606 (2006).

44. 44.

Shaw, A. Beyond ‘to share or not to share’: The impartiality account of fairness. Curr. Dir. Psychol. Sci. 22, 413–417 (2013).

45. 45.

Nowak, M. A. & Sigmund, K. Evolution of indirect reciprocity by image scoring. Nature 393, 573–577 (1998).

46. 46.

Meristo, M. & Surian, L. Do infants detect indirect reciprocity?. Cognition 129, 102–113 (2013).

47. 47.

Surian, L. & Franchin, L. Infants reason about deserving agents: A test with distributive actions. Cogn. Dev. 44, 49–56 (2017).

48. 48.

Konow, J., Saijo, T. & Akai, K. Equity versus equality: Spectators, stakeholders and groups. J. Econ. Psychol. https://doi.org/10.1016/j.joep.2019.05.001 (2019).

49. 49.

Xiao, E. & Bicchieri, C. When equality trumps reciprocity. J. Econ. Psychol. 31, 456–470 (2010).

50. 50.

Gächter, S. & Riedl, A. Dividing justly in bargaining problems with claims: Normative judgments and actual negotiations. Soc. Choice Welfare 27, 571–594 (2006).

51. 51.

DeScioli, P., Massenkoff, M., Shaw, A., Petersen, M. B. & Kurzban, R. Equity or equality? Moral judgments follow the money. Proc. R. Soc. B Biol. Sci. https://doi.org/10.1098/rspb.2014.2112 (2014).

52. 52.

Rasinski, K. A. What’s fair is fair—Or is it? Value differences underlying public views about social justice. J. Pers. Soc. Psychol. 53, 201–211 (1987).

53. 53.

Cappelen, A. W., Moene, K. O., Sørensen, E. Ø. & Tungodden, B. Needs versus entitlements—An international fairness experiment. J. Eur. Econ. Assoc. 11, 574–598 (2013).

54. 54.

Bicchieri, C. The Grammar of Society: The Nature and Dynamics of Social Norms (Cambridge University Press, Cambridge, 2005).

55. 55.

Krupka, E. L. & Weber, R. A. Identifying social norms using coordination games: Why does dictator game sharing vary?. J. Eur. Econ. Assoc. 11, 495–524 (2013).

56. 56.

Forsythe, R., Horowitz, J. L., Savin, N. E. & Sefton, M. Fairness in simple bargaining experiments. Games Econ. Behav. 6, 347–369 (1994).

57. 57.

Greenberg, J. Countering inequity with inequity: Over-rewarding generosity and under-rewarding greed. Eur. J. Soc. Psychol. 12, 181–185 (1982).

58. 58.

Ule, A., Schram, A., Riedl, A. & Cason, T. N. Indirect punishment and generosity toward strangers. Science 326, 1701–1704 (2009).

59. 59.

Crockett, M. J., Clark, L., Lieberman, M. D., Tabibnia, G. & Robbins, T. W. Impulsive choice and altruistic punishment are correlated and increase in tandem with serotonin depletion. Emotion 10, 855–862 (2010).

60. 60.

Fetchenhauer, D., Schlösser, T., Lotz, S., Baumert, A. & Gresser, F. Individual differences in third-party interventions: How justice sensitivity shapes altruistic punishment. Negot. Confl. Manage. Res. 4, 297–313 (2011).

61. 61.

O’Gorman, R., Henrich, J. & Van Vugt, M. Constraining free riding in public goods games: Designated solitary punishers can sustain human cooperation. Proc. R. Soc. B Biol. Sci. 276, 323–329 (2009).

62. 62.

Sigmund, K. Punish or perish? Retaliation and collaboration among humans. Trends Ecol. Evol. 22, 593–600 (2007).

63. 63.

Graham, J. et al. Mapping the moral domain. J. Pers. Soc. Psychol. 101, 366–385 (2011).

64. 64.

Siegel, J. Z., Mathys, C., Rutledge, R. B. & Crockett, M. J. Beliefs about bad people are volatile. Nat. Hum. Behav. 2, 750–756 (2018).

65. 65.

Fleming, S. M., van der Putten, E. J. & Daw, N. D. Neural mediators of changes of mind about perceptual decisions. Nat. Neurosci. 21, 617–624 (2018).

66. 66.

Faul, F., Erdfelder, E., Buchner, A. & Lang, A.-G. Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behav. Res. Methods 41, 1149–1160 (2009).

67. 67.

Cohen, J. Statistical Power Analysis for the Behavioural Science 2nd edn. (Academic Press, New York, 1988).

68. 68.

Gignac, G. E. & Szodorai, E. T. Effect size guidelines for individual differences researchers. Pers. Individ. Differ. 102, 74–78 (2016).

69. 69.

Soto, C. J. & John, O. P. The next Big Five Inventory (BFI-2): Developing and assessing a hierarchical model with 15 facets to enhance bandwidth, fidelity, and predictive power. J. Pers. Soc. Psychol. 110, 1–27 (2016).

70. 70.

Lee, K. & Ashton, M. C. Psychometric properties of the HEXACO-100. Assessment 25, 543–556 (2018).

71. 71.

Reniers, R. L. E. P., Corcoran, R., Drake, R., Shryane, N. M. & Völlm, B. A. The QCAE: A questionnaire of cognitive and affective empathy. J. Pers. Assess. 93, 84–95 (2011).

72. 72.

Graham, J., Haidt, J. & Nosek, B. A. Liberals and conservatives rely on different sets of moral foundations. J. Pers. Soc. Psychol. 96, 1029–1046 (2009).

73. 73.

Ho, A. K. et al. The nature of social dominance orientation: Theorizing and measuring preferences for intergroup inequality using the new SDO7 scale. J. Pers. Soc. Psychol. 109, 1003–1028 (2015).

74. 74.

Lipkus, I. The construction and preliminary validation of a global belief in a just world scale and the exploratory analysis of the multidimensional belief in a just world scale. Pers. Individ. Differ. 12, 1171–1178 (1991).

75. 75.

Ramsay, J. O., Hooker, G. & Graves, S. Functional Data Analysis with R and MATLAB (Springer, New York, 2009).

76. 76.

Ramsay, J. O. & Silverman, B. W. Functional Data Analysis (Springer, New York, 2005).

77. 77.

Ramsay, J. O., Wickham, H., Graves, S. & Hooker, G. Package ‘fda’: Functional data analysis. CRAN (2018).

78. 78.

Shang, H. L. A survey of functional principal component analysis. AStA Adv. Stat. Anal. 98, 121–142 (2014).

79. 79.

Geenens, G. Curse of dimensionality and related issues in nonparametric functional regression. Stat. Surv. 5, 30–43 (2011).

80. 80.

Ullah, S. & Finch, C. F. Applications of functional data analysis: A systematic review. BMC Med. Res. Methodol. https://doi.org/10.1186/1471-2288-13-43 (2013).

81. 81.

Verduyn, P., Van Mechelen, I. & Frederix, E. Determinants of the shape of emotion intensity profiles. Cogn. Emot. 26, 1486–1495 (2012).

82. 82.

Vines, B., Krumhansl, C. L., Wanderley, M. M. & Levitin, D. J. Cross-modal interactions in the perception of musical performance. Cognition 101, 80–113 (2006).

## Acknowledgements

We thank Kun Zhao, Damien Crone, and Anne Löffler for helpful discussions.

## Funding

This research was supported by an Australian Research Council grant (ARC DP160103353) to S.B.

## Author information

Authors

### Contributions

All authors contributed to conception and design. M.A. programmed the experiment, collected and analysed data, and drafted the article. All authors reviewed and revised the manuscript.

### Corresponding author

Correspondence to Milan Andrejević.

## Ethics declarations

### Competing interests

The authors declare no competing interests.

### Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

## Rights and permissions

Reprints and Permissions

Andrejević, M., Feuerriegel, D., Turner, W. et al. Moral judgements of fairness-related actions are flexibly updated to account for contextual information. Sci Rep 10, 17828 (2020). https://doi.org/10.1038/s41598-020-74975-0

• Accepted:

• Published: