Fairness takes time: Development of cooperative decision-making in fairness contexts

Abstract

The current study examined the development of fairness behavior and tested whether children’s fair choices are fast and intuitive or slow and deliberate. Reaction times were measured while 4–9-year-olds (N = 94, 49 girls, 84.6% White) completed a novel social decision-making task contrasting fair with selfish choices. Fairness behavior increased during childhood, shifting from predominately selfish choices among young children to fair choices by age 7. Moreover, young children’s fair choices were slow and deliberate, whereas reaction times did not predict older children’s choices. These findings contrast with adults’ intuitive cooperation and point to a protracted development and learning of cooperative decision-making in fairness contexts.

Cooperation is considered a central aspect of human sociality (Silk & House, 2012; Tomasello, 2019). There is considerable evidence to suggest that the capacity to cooperate is deeply rooted in our biology and may thus be considered a core part of human nature. In fact, our closest primate relatives, the great apes, have been shown to exhibit some forms of cooperation (e.g., Horner et al., 2011; Warneken et al., 2007). Moreover, some cooperative behaviors, such as helping others in need, emerge early in human ontogeny (e.g., Svetlova et al., 2010; Warneken & Tomasello, 2006), further supporting the notion of a strong biological contribution to the development of cooperation.

One important question that remains open, however, concerns the underlying mechanisms that lead individuals to make cooperative choices. There currently are two opposing viewpoints with respect to this question. The first is that humans are naturally selfish beings, and that in order to behave cooperatively, we must exert self-control to overcome selfish impulses (Achtziger et al., 2015; Steinbeis, 2018a; Tinghög et al., 2016). The second is that humans are intuitively cooperative, and that we may choose to use reflective control and deliberation to make selfish decisions in order to strategically maximize benefits for ourselves (Isler et al., 2018; Lotito et al., 2013; Rand, 2016; Zaki & Mitchell, 2013). This second viewpoint, also referred to as the ‘intuitive cooperation hypothesis’, has received support from work in behavioral economics (Bear & Rand, 2016; Rand et al., 2014). For example, adults who make faster and thus more intuitive decisions tend to behave more cooperatively than those who make slower and thus more deliberate decisions (Cappelen et al., 2016; Isler et al., 2018; Lotito et al., 2013; Rand, 2016; Rand et al., 2012). These findings suggest that intuitive decision-making processes support cooperative decisions (but see Kvarven et al., 2020, for a meta-analysis that questions the robustness of the effect).

The intuitive cooperation hypothesis provides a valuable framework for thinking about the development of cooperative behaviors. Specifically, it generates two possibilities for how intuitive cooperation initially emerges. The first possibility, in line with an evolutionary or biological basis of cooperation (Warneken & Tomasello, 2006), is that cooperation is the default state from the beginning, suggesting that cooperative decisions may be the intuitive choice even during infancy. The second possibility does not assume a default state but instead assumes that intuitive processes come to favor cooperation over time, particularly when previous experience has shown that choosing to be cooperative is advantageous (Bear & Rand, 2016; Nishi et al., 2016; Rand et al., 2014). One key way to test between these two possibilities is to study intuitive cooperation through a developmental lens, as the question is fundamentally developmental in nature. As such, assessing the intuitiveness of cooperation across multiple age groups is critical to clarifying when and how intuitive cooperation comes about. However, the intuitive cooperation hypothesis has been tested almost entirely with adults, and thus leaves open whether cooperation is indeed the default state and intuitive from early in development or if it develops through socialization and learning over time.

To date, only a few studies have tested the intuitive cooperation hypothesis from a developmental perspective. One of the first studies to explicitly test this hypothesis in early development examined infants’ instrumental helping behavior (Grossmann et al., 2020). This study revealed that by 14 to 18 months of age, infants who were faster to instrumentally help others also helped more frequently, suggesting cooperation may be the intuitive choice even during infancy. Another recent study tested this hypothesis in preschoolers’ sharing behavior by adding a time pressure manipulation to a Dictator Game paradigm (Plötner et al., 2021). This study’s findings also supported the intuitive cooperation hypothesis, with children ages 3, 5, and 7 years all sharing more of their 9 stickers with an absent recipient under time pressure compared to time delay (though note that an odd number of stickers made it impossible for children in this study to distribute fairly between self and other). These recent findings represent important first steps in exploring the intuitive cooperation hypothesis in human development, but critical questions remain open regarding how well they generalize across cooperative behaviors. In particular, different cooperative behaviors vary greatly in the level of cost they impose on the prosocial agent (Svetlova et al., 2010; Warneken & Tomasello, 2007). Moreover, different forms of cooperation are often uncorrelated with one another, emerge at different developmental timepoints, and are likely supported by different motives and mechanisms (Dunfield, 2014; Martin & Olson, 2015; Paulus, 2014; Sommerville et al., 2013; Steinbeis, 2018b). As such, it is critical to continue to expand the test of intuitive cooperation to include a wide range of cooperative behaviors.

One important form of cooperation that has yet to be explored through the intuitive cooperation lens is fairness. The emerging research on the development of fairness behavior in infants and children shows that the foundations of fairness behavior are present remarkably early in development (see Ibbotson, 2014; McAuliffe et al., 2017). By the second year, infants can distinguish between fair and unfair distributions and show a preference for fair distributions and distributors (Geraci & Surian, 2011; Meristo et al., 2016; Sloane et al., 2012; Sommerville et al., 2013). By age 3, children demonstrate some understanding of equality and fairness norms. For instance, around this age, children state that both they and their peers should share resources equally (Smith et al., 2013), and reject unequal distributions that disadvantage them (Blake & McAuliffe, 2011; LoBue et al., 2011; McAuliffe et al., 2013). However, it is not until age 7 or 8 that children reject advantageous inequity as well (Blake & McAuliffe, 2011; Fehr et al., 2008; Kogut, 2012; Smith et al., 2013), and will even choose to throw away a resource rather than receive more than their peer (Shaw & Olson, 2012). Overall, this work indicates that even though the foundations for understanding and expecting fairness are present from early in ontogeny, children’s own fairness behavior develops over many years, making it one of the later emerging forms of cooperation. This protracted development in turn leads to the possibility that young children’s fairness behavior may not initially be intuitive but rather deliberate. Yet no prior developmental work has assessed the intuitiveness of early fairness behavior.

The main purpose of the current study was thus to apply the intuitive cooperation hypothesis to the developmental study of fairness and examine the cognitive processes involved in children’s emerging fairness behavior and decision-making. Specifically, we tested whether children’s choice to behave fairly during a cooperative decision-making task was fast and intuitive or slow and deliberate. We created a one-shot, forced-choice modified Dictator Game in which children were shown two pre-set distributions of a valued resource (stickers) and had to decide between behaving cooperatively and choosing the fair distribution or behaving non-cooperatively and choosing the selfish distribution. Given the evidence supporting a developmental change in children’s fairness behavior from being more self-centered in 4–6-year-olds to being more norm-centered in 7–9-year-olds (e.g., McAuliffe et al., 2017), we focused on 4–9-year-old children’s fairness behavior.

To assess the intuitive or deliberate nature of children’s fair and selfish choices, we measured how quickly children made their choice during the task and used this reaction time as an indicator of how intuitive their decision was, presuming that intuitive decisions occur more quickly than deliberate ones (Kahneman, 2011; Rubinstein, 2007). Reaction time is commonly used to assess the intuitiveness of cooperative decision-making in adults (e.g., Rand et al., 2012), yet only one prior developmental study has reported children’s reaction times in the context of fairness behavior (Blake & McAuliffe, 2011).¹ In that study, 4–8-year-old children participated in the Inequity Game, in which they chose to accept or reject different distributions of candies for themselves and another child. Participants’ choices and reaction times were recorded. Compared to trials with equitable distributions, children of all ages took longer to make their decision on inequitable trials where they received less (disadvantageous inequity), but only 8-year-olds made slower decisions on inequitable trials when they received more (advantageous inequity). These interesting findings show that different cognitive mechanisms underlie children’s responses to disadvantageous versus advantageous inequity. However, Blake and McAuliffe (2011) were primarily interested in children’s reactions to advantageous and disadvantageous inequity rather than in how intuitive or deliberate their decisions were. As such, their paradigm involved accepting or rejecting a single distribution of resources and to complete multiple trials while the recipient watched, differentiating it from the anonymous, one-shot paradigms typically used to test the intuitive cooperation hypothesis in both children and adults (e.g., Plötner et al., 2021; Rand et al., 2012). Thus, no prior developmental work has directly addressed the intuitive or deliberate nature of children’s decisions about fairness.

Finally, given that there is evidence to suggest that differences in behavioral control capabilities are related to children’s cooperative behavior (Aguilar-Pardo et al., 2013; Steinbeis et al., 2012), we also measured children’s inhibitory control in order to examine whether and how it contributes to the development of and individual variability in children’s fairness decision-making.

From a developmental perspective, based on prior work (Blake & McAuliffe, 2011; Fehr et al., 2008; McAuliffe et al., 2017; Smith et al., 2013), we hypothesized that children’s fairness behavior would increase with age, such that older children would make more fair choices whereas younger children would make more selfish choices. From a cognitive mechanistic perspective, we tested between two competing hypotheses: 1) if fairness behavior is intuitive between ages 4 and 9, then children should make fair choices more quickly than selfish choices, whereas 2) if fairness behavior is deliberate and requires cognitive control, then children should make fair choices more slowly than selfish choices. Testing between these competing hypotheses across age in childhood further affords the opportunity to shed light on the nature of the cognitive processes at play as fairness behavior first emerges in development. In other words, the current approach also allowed us to map out whether the cognitive processes involved in fairness decision-making—intuitive or deliberate—change as a function of development. Finally, we hypothesized that children’s inhibitory control would increase with age (Davidson et al., 2006), and that children with greater inhibitory control would have longer reaction times, particularly when making less intuitive choices (whether fair or selfish).

Methods

Participants

We tested 135 children (64 girls) between the ages of 4 and 9 years from a medium-sized university town in North America. The study took place either in the university lab (M = 101) or through our partnership with the Living Laboratory at a local children’s museum (N = 34). Of the 135 children, 41 (15 girls) were excluded based on our established exclusion criteria, including not passing comprehension checks (N = 6; 5 4-year-olds, 1 5-year-old), not completing the full task (N = 7; 3 4-year-olds, 1 5-year-old, 2 7-year-olds, 1 8-year-old), and exhibiting behavior that made it difficult to calculate their reaction time, such as needing to be prompted during the decision part of the test trial (N = 26; 12 4-year-olds, 5 5-year-olds, 2 6-year-olds, 3 7-year-olds, 1 8-year-old, 3 9-year-olds). Additionally, two more participants were found to be statistical outliers using inter-quartile range in SPSS (reaction time > 3 SD above the mean) and were also excluded (N = 2; 1 4-year-old, 1 8-year-old). Our final sample thus consisted of 94 children (49 girls), including 15 4-year-olds (M = 52 months 16 days, or 52;16, SD = 104.9 days), 16 5-year-olds (M = 66;25, SD = 120.9 days), 16 6-year-olds (M = 77;4, SD = 133.2 days), 16 7-year-olds (M = 90;19, SD = 107.1 days), 16 8-year-olds (M = 102;4, SD = 102.9 days), and 15 9-year-olds (M = 113;8, SD = 91.1 days). In the final sample, 76 were tested in the lab, and 18 were tested at the museum.

Of the families who provided information about race (N = 78), 84.6% of participants were White, 2.6% were Black or African American, 3.8% were Asian, and 9.0% were mixed-race or chose “Other”. Of the families who provided information about parental education (N = 70), 95.7% of participants had at least one parent who attended a 4-year college or beyond. It is worth noting that this was a convenience sample that relied on families who had volunteered to be included in our database or participated while at the museum. The demographics of this sample are fairly representative of the towns from which participants were drawn, however, we acknowledge that this was a WEIRD sample (Henrich et al., 2010), which should be taken into account when considering how the results might generalize across contexts.

Procedure

Modified Dictator Game

We created a forced-choice modified Dictator Game paradigm that would allow us to measure both children’s fairness behavior and their reaction times. Participants were asked to make a forced-choice decision between two different distributions of stickers for themselves and another (absent) child, one that was fair (2:2, or 2 for the participant, 2 for the other child) and one that was selfish (3:1). Similar to previous work with adults (Cappelen et al., 2016; Lotito et al., 2013; Rand et al., 2012), we measured the speed with which children made their decisions (reaction time), starting from the moment the participant was presented with the choice until the moment they indicated their decision by pointing to their choice. Each participant completed a test trial consisting of a single-shot interaction with an absent child recipient. This aligns with much of the adult work on intuitive cooperation, which has relied on one-trial, single-shot interactions. Given that our method was fairly novel, we also included a second, exploratory trial (a second single-shot interaction with a different absent child recipient); however, our primary interest and hypotheses centered on the first test trial. We will thus focus only on the first test trial here; details about the exploratory second trial are provided in the Online Supplemental Material (OSM).

Each phase of the paradigm is described in detail below, and the full script is included in the OSM. The entire procedure was recorded on video so that participants’ behavior could be coded offline.

Setup.

The experiment began with the participant sitting across from the experimenter (E) at a table with two boxes on it. The participant was asked to draw a picture of themselves, which was then placed on one of the boxes to indicate that it was the participant’s box. The other box had a photograph of a neutral-faced, gender- and race-matched child on it (taken from the CAFE set; LoBue & Thrasher, 2015). The participant was told that this other box belonged to the child in the photograph (Figure 1A). Whether the child’s box was on their left or right was counterbalanced across participants.

Figure 1. Modified Dictator Game Paradigm.

Note. Images depicting the modified Dictator Game paradigm. A) The participant’s box with their drawing (left) and the other child’s box with the photograph (right). B) The red and blue plate with a single sticker during the first comprehension check. In this example, the participant is assigned to the red side, and they would need to correctly respond by saying that they would receive one sticker. C) The wall, used to conceal the two plates from the participant until the start of the trial. D) The distribution of stickers used during the two-plate comprehension check, with a fair distribution (1:1) on the left and a selfish distribution (1:0) on the right. E) The distribution of stickers used during the test trial, with a selfish distribution (3:1) on the left and a fair distribution (2:2) on the right. F) The presentation of the two test trial plates after removing the wall. The participant could then choose between the two distributions and indicate their choice by touching one of the plates.

Comprehension Phase.

E explained the rules of the game to the participant by presenting them with a plate that was half red and half blue. E told the participant that they would be shown plates just like this one but with stickers on it, and that any stickers on the red (or blue) side would be the participant’s stickers and would go in their box, and any stickers on the blue (or red) side would be the other child’s stickers and would go in the other child’s box. The color assigned to the participant (red or blue) was counterbalanced across participants, and the plates were always presented to the participant such that their assigned color was on the same side as their box. To check comprehension, E placed a sticker on one side of the plate at a time and asked the participant who would get the sticker (Figure 1B). If the participant answered incorrectly, E repeated the rules and asked again until the participant understood and answered correctly.

Next, E explained that in the game, the participant would be shown two plates at the same time with different amounts of stickers on each plate, and that the participant must touch the one plate that had the number of stickers they wanted for themselves on their side and the number of stickers they wanted for the other child on the other side. The participant was also shown that the plates would first be set up behind a wall that E placed in front of the participant on the table (Figure 1C). Once the plates were set, E would lift up the wall so that the participant could see the two plates and make their decision. The purpose of the wall was to control when the participant first saw the two distribution options, allowing us to have a clear starting point for when the participant began to make their decision.

Again, to check comprehension, E presented the participant with two plates with different distributions of stickers on them, one fair (1:1) and one selfish (1:0) (Figure 1D). E asked the participant how many stickers they and the other child would get if they picked each plate. If the participant answered incorrectly, E repeated the rules and asked again until the participant understood and answered correctly. These distributions were chosen to mimic the fair and selfish options that the participant would see during the test trials but with a different number of stickers.

Test Trial.

E then set up the test trial, during which the participant chose between two plates, one with a selfish distribution of stickers (3:1) and one with a fair distribution (2:2) (Figures 1E and 1F). E set up the two plates behind the wall, repeated the instructions, and then removed the wall and waited for the participant to choose one of the plates by touching or pointing to it. Once the participant had made their choice, E asked them why they made that choice. E then distributed the appropriate number of stickers into the participant’s and the other child’s boxes. Once the experiment was finished, the participant was able to retrieve their stickers from their box to take home.

Parent Questionnaire

We created a 40-question parent questionnaire by combining the three most relevant subscales of the Children’s Behavior Questionnaire (inhibitory control, impulsivity, and empathy), a validated measure of children’s temperament and capacity for self-regulation (Putnam et al., 2001; Putnam & Rothbart, 2006; Rothbart et al., 2001). We asked parents to rate how true each statement was of their child’s behavior using a 7-point Likert scale ranging from “extremely untrue of your child” to “extremely true of your child”, plus a “not applicable” response option. Parents completed the questionnaire on a tablet through Qualtrics (Qualtrics, Provo, UT) during their visit.

Coding

We used video to code children’s choice during the test trial, their reaction time, and their justification for their choice. Participants’ choices during the test trial were recorded as either the fair or the selfish option, and these choices were numerically recoded as the number of stickers children received (2 or 3, respectively). Reaction times were coded using the INTERACT video coding software (Mangold International, Arnstorf, Germany), and were defined as starting from the moment the child saw the two choices (after the wall was lifted) and ending the moment they made their decision by touching one of the plates. If the child only pointed to but did not physically touch a plate, they were coded as having made a decision when they reached the fullest extension of their point. While no child first touched or fully pointed to one of the plates and then changed their mind and selected the other plate, these trials would have been excluded since they fall under the predetermined exclusion criterion of exhibiting behavior that makes it difficult to calculate the participant’s reaction time. A secondary coder coded all videos to establish reliability. Intraclass correlation coefficient (ICC) estimates and their 95% confidence intervals were calculated using SPSS statistical package version 26 (SPSS Inc, Chicago, IL) based on a single rater, consistency, 2-way mixed-effects model. Interrater reliability was determined to be excellent (ICC = 0.999, 95% CI = 0.999–1.000) (Koo & Li, 2016). All reaction time scores were log₁₀-transformed in SPSS to correct for positive skew, and these transformed values were used in all further analyses.

We transcribed and coded children’s justifications of their choices as a form of supporting evidence. Children’s justifications were coded as falling into one of four categories: fairness norm (referencing norms of equality or how the other person would feel about the outcome), explicit desire (referencing a desire to get more stickers), implicit desire (referencing how many stickers each person would get), or other (see Table S1 for more details and examples). All justifications were first categorized by two researchers independently; if there were discrepancies (which occurred for 7.0% of justifications), the researchers discussed and came to an agreement.

Data from the parent questionnaire were used to calculate participants’ mean scale scores for inhibitory control, impulsivity, and empathy (Rothbart et al., 2001). Impulsivity and empathy did not relate to children’s reaction time or choice and will thus not be further discussed.

Results

We first present findings of developmental effects on our main variables of interest, reaction time and selfish vs. fair choice on the test trial, as well as how the relation between these variables changes with age. We then present findings regarding inhibitory control and its relation to our other main variables.

As a preliminary check, we tested for a main effect of gender on our main variables of interest. Only parent-reported inhibitory control showed a significant effect of gender such that girls were reported to have greater inhibitory control than boys (M_Girl = 5.412, SD_Girl = 0.949; M_Boy = 4.953, SD_Boy = 0.849; t(83) = −2.341, p = .022, d = −0.51, 95% CI (−0.85, −0.07)). This result is supported by previous findings of gender differences in inhibitory control measures when using the CBQ (Gagne et al., 2013). However, because this gender effect was not predicted, limited to inhibitory control, and unrelated to our main questions of interest, we did not include gender in our analyses.

Developmental effects on reaction time and choice

We assessed how children’s reaction times and selfish vs. fair choices differed in relation to their age. The correlation between age and reaction time was not significant (r = .067, p = .52, r² = .004). There was a significant negative correlation between age and choice (r = −.410, p < .001, r² = .168), which we explored further in follow-up analyses. Specifically, a chi-square test was used to compare how many children chose either selfishly or fairly at each age. As predicted, children aged 4, 5, and 6 were more likely to choose the selfish option, while children aged 7, 8, and 9 were more likely to choose the fair option (X²[5, N = 94] = 19.514, p = .002, V = .456) (Figure 2). Additionally, children’s justifications also changed with age and aligned closely with their choices. Specifically, younger children mostly justified their common selfish choices with references to desire, and older children often justified their common fair choices with references to fairness norms (see OSM for details).

Figure 2. Developmental Effect on Choice on the Test Trial.

Note. Percentage of children at each age who made fair and selfish choices on the test trial.

Relation between reaction time and choice

We used a binary logistic regression to examine if reaction time was predictive of children’s choice (fair vs. selfish). We first included age in months and reaction time as continuous predictor variables in the model and found that age, but not reaction time, was a significant predictor of children’s choices (Age: B = −0.056, p < .001, OR = 0.946; Reaction Time: B = −1.990, p = .174, OR = 0.137). However, given the developmental effect on choice reported above, as well as evidence from prior work showing a similar developmental effect regarding fairness behavior (Blake & McAuliffe, 2011; Fehr et al., 2008; McAuliffe et al., 2017; Smith et al., 2013), and given that we found no correlation between age and reaction time, we had reason to believe that there may be a difference in behavior between the younger and older children. We thus made the post-hoc decision to split the sample into two age groups: younger (ages 4 to 6) and older (ages 7 to 9), and run additional exploratory analyses to examine the relation between reaction time and choice within the younger and older age groups separately. When split by age, we found that reaction time did predict choice for the younger age group (B = −6.200, p = .045, OR = 0.002) such that children who made faster decisions were more likely to choose selfishly than those who made slower decisions (Figure 3A). There was no association between reaction time and choice in the older age group (B = −0.804, p = .65, OR = 0.448) (Figure 3B). Additionally, when inhibitory control and an age x latency interaction were included in the model as covariates, they were not predictive of children’s choice (all p-values > .28).

Figure 3.

Note. In the younger age group (Figure 3A), children who made faster decisions were more likely to choose the selfish distribution than the fair distribution. In the older age group (Figure 3B), there was no association between children’s reaction time and their choice in the modified Dictator Game.

Developmental effects on inhibitory control

Of the 94 participants, parents of 85 completed the inhibitory control questionnaire. The mean inhibitory control score was 5.190 out of 7 (SD = 0.926). There was a significant positive correlation between age and inhibitory control score (r = .237, p = .029, r² = .056). Additionally, an independent samples t-test revealed that the mean inhibitory control score in the younger age group (M = 4.977, SD = 0.921) was significantly lower than in the older age group (M = 5.409, SD = 0.890, t(83) = −2.195, p = .031, d = −0.48, 95% CI (−0.82, −0.04)). Thus, as predicted, we found that parent-reported inhibitory control increased with age.

Inhibitory control in relation to reaction time and choice

Because mean inhibitory control scores differed for the younger and older age groups, we examined their relation to reaction time and choice in each age group separately. Inhibitory control scores were slightly positively correlated with reaction times for younger, but not older children, although this correlation for the younger age group did not quite reach significance (Younger: r = .295, p = .06, r² = .087; Older: r = .108, p = .5, r² = .012). These findings hint that increased inhibitory control may be associated with taking more time to make a decision, although the relation was inconsistent across age groups. However, for both age groups, inhibitory control did not differ based on whether children made a selfish or fair choice (all p-values > .66). Additionally, an exploratory analysis including an inhibitory control x latency interaction in our regression model revealed that there was no interaction of inhibitory control and reaction time on children’s choice (all p-values > .35). We thus did not find support for our hypothesis that children with greater inhibitory control would take longer to make less intuitive decisions (i.e., fair choices for younger children and selfish choices for older children).

Discussion

The current study tested the intuitive cooperation hypothesis from a developmental perspective by examining the development and cognitive characteristics of children’s emerging fairness behavior. In line with our prediction and previous work (Blake & McAuliffe, 2011; Fehr et al., 2008; McAuliffe et al., 2017; Smith et al., 2013), the present results show that younger children (ages 4 to 6) were more likely to choose the selfish distribution while older children (ages 7 to 9) were more likely to choose the fair distribution. This developmental change in fairness decision-making behavior was reflected in children’s justifications of their choices as well (see OSM for details). As for our central question regarding the intuitive cooperation hypothesis, our results showed that the younger children’s rare fair choices were slow and deliberate, and their predominant selfish choices were fast and intuitive. That is, a shorter reaction time (i.e., making a faster decision) was associated with choosing selfishly among 4–6-year-olds. This suggests that for young children, their faster, and potentially more intuitive decisions are primarily selfish. In contrast, we did not find a relation between reaction time and choice in the older age group (ages 7 to 9). These older children predominantly chose the fair option, yet they did not make these choices more quickly than selfish choices. Together, our findings provide the first systematic evidence that acting fairly rather than selfishly is not the intuitive response but requires time and deliberation among young children. This pattern of behavior may gradually shift during middle childhood, as older children did not differ in the amount of time and deliberation they needed to make either a fair or selfish choice.

This conclusion contrasts with other developmental studies of intuitive cooperation that have tested behaviors other than fairness and found cooperation to be the faster, more intuitive choice. For instance, Grossmann and colleagues (2020) found evidence for intuitive cooperation during infancy when exploring the association between how quickly and how often infants provided instrumental help. In addition, a recent study by Plötner et al. (2021) used a time pressure manipulation to invoke intuitive decision-making during a Dictator Game and found that children as young as 3 were more generous in how much they shared with a recipient when under time pressure, suggesting that like adults, children are more generous when acting under time pressure. In contrast, we did not find this effect in our study, and instead found children’s choices to be in line with what is already known about the protracted development of fairness. These inconsistencies suggest that the development of intuitive cooperation may be more complex than the initial findings in the field indicated, and that differences between the previous findings and our current results are likely due to both methodological differences and the different kinds of cooperative behaviors being examined.

It is well established that different prosocial behaviors develop along distinct trajectories and are likely supported by different mechanisms and brain systems (Dunfield, 2014; Knafo-Noam et al., 2018; Paulus, 2018; Sommerville et al., 2013; Steinbeis, 2018b). Unlike instrumental helping and sharing, which emerge during infancy and toddlerhood (Svetlova et al., 2010; Warneken & Tomasello, 2007), overt fairness behavior emerges substantially later in ontogeny (Dunfield, 2014; McAuliffe et al., 2017; though note that an implicit understanding of fairness may be present in infancy; see Sloane et al., 2012; Sommerville et al., 2013). Our developmental findings with regard to children’s choices (i.e., that children between 4 to 6 years were more likely to choose the selfish option whereas children 7 years and older were more likely to choose the fair option) are consistent with the literature on the development of fair behavior (e.g., Fehr et al., 2008; McAuliffe et al., 2017; Smith et al., 2013) but not with the literature on the development of other prosocial behaviors such as helping and sharing which shows that even toddlers and preschool-age children behave generously by helping others and sharing valuable resources (Callaghan et al., 2011; Warneken, 2015). Given that the pattern of children’s choices we observed align closely with the developmental timeline of fairness behavior rather than the more precocious development of prosocial behaviors such as helping and sharing, we believe our findings are most parsimoniously understood as reflecting children’s fairness decisions (though we acknowledge that we cannot conclusively rule out alternative interpretations based on our study).

One explanation for the unique developmental trajectory of fairness behavior is that fairness requires one to not only understand but also to abide by the social norm of equality, and the alignment between understanding and behaving in accordance with these social norms is a skill that may still be emerging during middle childhood (Blake, 2018; House & Tomasello, 2018; Smith et al., 2013; Steinbeis, 2018b). Further, the transition to norm-centered fairness in middle childhood is influenced by and learned through experience with culture-specific fairness norms (Blake et al., 2015; Callaghan & Corbit, 2018; House et al., 2013; Ibbotson, 2014). Previous cross-cultural studies have shown that children’s costly prosocial behavior looks similar across cultures until middle childhood, at which point it starts to diverge and align with the culture-specific normative behaviors exemplified by adults (Blake et al., 2015; Callaghan & Corbit, 2018; House et al., 2013; Ibbotson, 2014). It is possible that the younger children in the current study had yet to reach these developmental milestones and had difficulty following fairness norms and controlling selfish tendencies, whereas the older children’s improved behavioral control and greater socialization experience allowed them to better align their behavior with fairness norms. There are thus multiple plausible explanations for why, unlike helping and sharing, fairness behavior emerges later and is not intuitive in early childhood. It will be fascinating for future work to explore and tease apart these possibilities.

In the present study, we measured children’s natural reaction times (rather than creating a time pressure as in some prior work, e.g., Plötner et al., 2021) to assess the intuitiveness of children’s fairness decisions. To our knowledge, only one prior study has assessed the relation between fairness behavior and un-manipulated reaction time in early childhood (Blake & McAuliffe, 2011). This study measured children’s fairness behavior and reaction time in the Inequity Game and found that, when compared to equitable trials, 4–8-year-olds took longer to make a decision on disadvantageous inequitable trials, while only 8-year-olds took longer on advantageous inequitable trials. In contrast with the current findings, Blake and McAuliffe (2011) did not find any differences in 4–6-year-old children’s reaction times between equitable and advantageous inequitable trials, the distribution most similar to the current study. However, note that when conducting within-subject comparisons of reaction time across all advantageous inequitable trials, Blake and McAuliffe found that children ages 4 to 7 were faster to accept an advantageous offer than to reject it. This result aligns with our finding that younger children were faster to choose a selfish distribution.

Work with adults on the intuitive cooperation hypothesis has found that adults are faster to make cooperative, rather than selfish choices (Cappelen et al., 2016; Isler et al., 2018; Lotito et al., 2013; Rand, 2016; Rand et al., 2012; Zaki & Mitchell, 2013), although note that these studies examined sharing rather than fairness behavior specifically. In contrast, we found that young children (ages 4 to 6) were faster to behave selfishly, and we found no relation between reaction time and choice in older children (ages 7 to 9). While our findings may appear to contradict the claims of the intuitive cooperation hypothesis, we propose instead that they provide insights into how intuitive cooperation initially emerges and how its development may be more nuanced. A simple version of the hypothesis might predict that cooperation is the default state beginning in infancy, whereas a more nuanced version is that cooperation becomes the intuitive choice over time through socialization and experience. Our results support the second possibility and indicate that, at least one form of cooperation – fairness – is initially deliberate rather than intuitive in early childhood. Middle childhood may then be a transitional period during which children’s fairness behavior is just beginning to align with a more robust understanding of fairness and equality, but choosing to behave fairly may still not be an intuitive and quick decision for children. It is also important to note that cooperative behaviors are often driven by different motivations, including sympathetic concern for others, a sense of obligation or respect, or strategic self-interest, and different motives may result in faster or slower decision-making (Engelmann & Tomasello, 2019; Martin & Olson, 2015; Tomasello, 2020). Seen this way, the current findings suggest that what is considered the intuitive choice can evolve over time, likely through socialization and more experiences with cooperative decision-making and equal resource distribution (Brownell & The Early Social Development Research Lab, 2016; Capraro & Cococcioni, 2015; Henrich et al., 2001; McAuliffe et al., 2018). This explanation fits nicely with the more nuanced version of the intuitive cooperation hypothesis, which accounts for the role of learning and previous experience in the link between intuitive decision-making processes and cooperative choices (Bear & Rand, 2016; Nishi et al., 2016; Rand et al., 2014). This proposal is tentative, however; testing it will require future work to include a broader age range (older children, adolescents, and adults) and to examine the generalizability of the current results to other cultural and socio-economic contexts.

Finally, we were interested in the potential influence of children’s inhibitory control on both their reaction time and fairness behavior. We found that parent-reported inhibitory control was positively correlated with age, and that the younger children in our sample scored significantly lower than the older children, which aligns with the current understanding of inhibitory control development during this age (Davidson et al., 2006; Simpson & Carroll, 2019). We found weak evidence that inhibitory control scores may be positively related with younger children’s reaction time, with those scoring higher on inhibitory control generally taking longer to make a decision. However, we did not find a relation between inhibitory control and children’s fairness behavior in the modified Dictator Game. In relation to sharing behavior, some studies have found a positive association with children’s behavioral control (Steinbeis, 2018a; Steinbeis & Over, 2017), while others have failed to find such an effect (Liu et al., 2016; Smith et al., 2013). Our findings contribute to the ongoing discussion about the relation between behavioral control and fairness, but additional research is still needed. In particular, it will be important to incorporate behavioral measures of inhibitory control rather than relying solely on parental report, since there is evidence to suggest that behavioral and scale measure of behavioral control are only weakly correlated and may be tapping into different constructs (Toplak et al., 2013).

Beyond relying on a parental report measure of inhibitory control, there are several other limitations of the current study that are worth future exploration. For example, we only included one fair-selfish distribution pattern in our paradigm, but additional kinds of distributions should be tested. Including a disadvantageous inequity distribution or varying the degree of inequality between the distributions to modulate the level of decision conflict will be important follow-up manipulations. It will also be valuable to use multiple test trials beyond one additional exploratory test trial to assess how children’s behavior may change across multiple trials. A strength of the current study is that we measured the participant’s own behavior in a first-person fairness decision-making task, yet it will be useful to examine how children’s reaction times change in a lower conflict setup like a third-party fairness task. Lastly, there is a general lack of developmental cooperation decision-making studies that report reaction time measures, making it difficult to interpret and contextualize our results. We hope that more developmental researchers will include measurements of reaction time when studying children’s cooperative decision-making, as this will provide further insight into the nature and nuances of this kind of measure when examining children’s decision-making behavior.

In conclusion, the current study demonstrates the developmental changes in fairness behavior during early and middle childhood while simultaneously assessing potential cognitive mechanisms through the lens of the intuitive cooperation hypothesis. The findings inform our understanding of the development of fairness, suggest that fairness behavior is initially deliberate rather than intuitive while highlighting the complex nature of the emergence of intuitive cooperation, and support accounts that distinguish between different forms of cooperative behavior as guided by different underlying mechanisms (Dunfield, 2014; McAuliffe et al., 2017; Steinbeis, 2018b). The protracted development of fairness choices suggests that extensive learning, presumably from observing and practicing equal resource distributions, as well as continuous exposure to culture-specific fairness norms likely plays a role in the development of this skill. This highly socialized pattern of development has widespread implications for informing how we teach children about fairness, both in formal and informal settings, and how we manage our expectations of what we expect from children at different stages of development. It will be important to continue to study intuitive cooperation from a developmental perspective and use a variety of measures to study different cooperative behaviors to help elucidate how the cognitive mechanisms underlying cooperation change over time.

• Explored the intuitive cooperation hypothesis in young children’s fairness behavior

• Measured reaction time of cooperative decisions with a novel modified Dictator Game

• Children’s fairness behavior shifts from self-centric to norm-centric around age 7

• Younger, but not older children made selfish choices more quickly than fair choices

• Intuitive cooperation in fairness contexts likely develops slowly through experience