# paper.pdf ## Metadata - PDFFormatVersion=1.6 - Language=en-US - IsLinearized=true - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - CreationDate=D:20260529111433+05'30' - Creator=Adobe InDesign 17.4 (Windows) - ModDate=D:20260529112411+05'30' - Producer=Adobe PDF Library 16.0.7 - Trapped=False - xmpmm:instanceid=uuid:4c8fa4f0-1841-4d18-aec9-a035985111d2 - xmpmm:originaldocumentid=xmp.did:A33916180F2068118083ACD1B377E157 - xmpmm:documentid=xmp.id:c3919c6f-0260-0b47-a537-9e691d2be0e2 - xmpmm:renditionclass=proof:pdf - xmpmm:derivedfrom=xmp.iid:c57fa069-af0a-4247-b095-d8cd1a9590b2xmp.did:c57fa069-af0a-4247-b095-d8cd1a9590b2xmp.did:A33916180F2068118083ACD1B377E157default - xmpmm:history=convertedfrom application/x-indesign to application/pdfAdobe InDesign 17.4 (Windows)/2026-05-29T11:14:33+05:30 - xmp:createdate=2026-05-29T11:14:33+05:30 - xmp:modifydate=2026-05-29T11:24:11+05:30 - xmp:metadatadate=2026-05-29T11:24:11+05:30 - xmp:creatortool=Adobe InDesign 17.4 (Windows) - xmprights:marked=False - photoshop:modifydate=2019-01-31T16:16:25 - photoshop:createdate=2019-01-24T11:43:46 - dc:format=application/pdf - pdf:producer=Adobe PDF Library 16.0.7 - pdf:trapped=False - extensisfontsense:slug=3023672065GraphikOpenType - PS3023672065--0Graphik-Bold1.0011904271029GraphikOpenType - PS1904271029--0Graphik-Regular1.0012428528031News Gothic MT OTOpenType - PS2428528031Monotype Typography0NewsGothicMTOT-Regular1.000543760319News Gothic MT OTOpenType - PS543760319Monotype Typography0NewsGothicMTOT-Bold1.0002785939051KnockoutOpenType - PS2785939051--0Knockout-68FullFeatherwt1.200423856041Graphik X CondOpenType - PS423856041--0GraphikXCond-Super1.0001526035435Graphik X CondOpenType - PS1526035435--0GraphikXCond-Bold1.0003093728306Graphik CondOpenType - PS3093728306--0GraphikCond-Black1.0011341118539Graphik CondOpenType - PS1341118539--0GraphikCond-Regular1.0011385893531GraphikOpenType - PS1385893531--0Graphik-Semibold1.0012986479693Graphik CompactOpenType - PS2986479693--0GraphikCompact-Regular1.0024223138361Graphik CompactOpenType - PS4223138361--0GraphikCompact-Semibold1.0023673047546Harding SevenOpenType - PS3673047546--0HardingSeven-Extrabold0.000103712566Harding SevenOpenType - PS103712566--0HardingSeven-Roman0.0002232452979Harding SevenOpenType - PS2232452979--0HardingSeven-Medium0.000631916388Harding SevenOpenType - PS631916388--0HardingSeven-Bold0.000157629484Publico BannerOpenType - PS157629484--0PublicoBanner-Black1.005 ## Contents ### Page 1 nature human behaviourhttps://doi.org/10.1038/s41562-026-02487-4ArticleStable intuition and the rise of deliberative prosociality in childhoodIn the format provided by the authors and uneditedSupplementary information ### Page 2 Supplementary Information forStable Intuition and the Rise of Deliberative Prosociality in ChildhoodFrancesco Margoni1, Francesco Nava2, Chiara Sotis3, Matthew R Levy2, Valerio Capraro4, Elena Nava41Department of Psychology, University of Stavanger.2Department of Economics, London School of Economics.3Department of Economics, Northwestern University.4Department of Psychology, University of Milano-Bicocca.ContentsA Additional Analyses 2A.1 Balance Checks and Condition Compliance . . . . . . . . . . . . . . . 2 A.2 Main Tables and Results by Task . . . . . . . . . . . . . . . . . . . . . 3 A.3 Non-Linear Age Regressions . . . . . . . . . . . . . . . . . . . . . . . . 4B Robustness Checks 12B.1 Sample Inclusion Criteria . . . . . . . . . . . . . . . . . . . . . . . . . 12 B.2 Task Selection Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 B.3 Time Pressure and Random Choice . . . . . . . . . . . . . . . . . . . . 19 B.4 Estimating Factors by Age . . . . . . . . . . . . . . . . . . . . . . . . . 20C Additional Scripts For Each Task 21C.1 The PGG Task (Explanatory Movie-S1) . . . . . . . . . . . . . . . . . 21 C.2 Additional Scripts and Instructions . . . . . . . . . . . . . . . . . . . . 22 1 ### Page 3 A Additional AnalysesA.1 Balance Checks and Condition ComplianceAs shown in Supplementary Table 1, condition (fast, slow) assignment was roughly equal. There were slightly more children in the fast condition in the 3-4 year age group. This difference arose because 9 children in the 3-4 age group responding under time delay failed at least one comprehension check (and were therefore dropped from the main analysis), while 3 children in this group responding under time pressure failed. No participant failed comprehension in the other age groups.Age Group Age 3-4 Age 5-6 Age 7-8 Age 9-10 Total Percent Slow Condition 56 70 65 66 257 48.95 Fast Condition 63 70 67 68 268 51.05 Total 119 140 132 134 525 100 Percent 22.67 26.67 25.14 25.52 100 –Supplementary Table 1: Sample divided by age group and decision mode. Supplementary Table 2 illustrates compliance with the fast condition across exper- imental tasks, reporting the percentage of non-compliant participants for each task and iteration. Moral dilemmas were the tasks where participants struggled the most to comply with the condition. For the MD, any participant failing to comply with the condition in one of the two frames was recorded as having failed the condition in both frames due to data collection.Task PGG DG UG DEG MD Round 1 2.66 3.42 – 5.34 – Round 2 3.02 3.42 – 1.52 – Round 3 2.66 3.80 – 1.90 – Offer 0 – – 0.38 – – Offer 1 – – 0.00 – – Offer 2 – – 2.66 – – Offer 3 – – 1.90 – – Offer 4 – – 1.52 – – Average 2.78 3.55 1.29 2.92 5.90Supplementary Table 2: Condition non- compliance. 2 ### Page 4 A.2 Main Tables and Results by TaskSupplementary Table 3 reports the factor loadings for the 3 retained factors, along with the uniqueness coefficients for each. We retained factors with eigenvalues greater than 1. The uniqueness coefficients are relatively high compared to other studies on adult samples, which is consistent with a higher degree of idiosyncratic behaviour in children. The aggregation through factor analysis allows us to extract commonalities across tasks. We show in section B1 and B2 that results are robust to a number of sample and task selection criteria, and in section B4 that results are robust to estimating factor weights separately for younger and older child participants. Supplementary Table 4 presents the linear regression coefficients from our main analysis. In these regressions, the dependent variables are the retained factors with condition (fast, slow), age, and their interaction included as control variables. Supplementary Table 5 presents the linear regression coefficients by task. In these regressions, the dependent variables represent the responses in a specific task, with condition, age, their interaction, and a round/trial indicator included as control vari- ables. Specifically, the regressor round X identifies the fixed effect associated with playing round X relative to round 1. Since participants were required to accept 5 offer levels in the UG, the regressor offer X identifies the fixed effect associated with accept- ing offer X relative to accepting the fair offer (offer 2). In the MD, where participants had to play 2 frames, the regressor footbridge identifies the fixed effect associated with playing the footbridge frame relative to the bystander frame. The independent vari- able beliefs represents a participant’s beliefs about the contributions of others in the PGG.Public Goods GameThe findings from the PGG reveal several key patterns. At age 3, children in the fast condition contribute more than those in the slow condition. However, this effect disap- pears after the preschool years. Older children tend to contribute more than younger ones, and on average, children contribute significantly more than 2 candies, the fair split (see also section B3). Contributions also increase in the second trial. Additionally, children’s expectations are positively correlated with their own contributions.Dictator GameIn the DG, we find no evidence that condition or age are associated with children’s choices. Nonetheless, children, on average, donate significantly more than 2 candies, the fair split (see also section B3). Donations also increase in the second and third trials compared to the first.Ultimatum GameThe findings from the UG indicate that older children are less likely to accept offers than younger ones. Children are most likely to accept fair offers (worth 2 candies), but they are more likely to accept higher offers (worth 3 or 4 candies) than lower offers (worth 0 or 1 candy). Fast decision mode is associated with lower acceptance at the margins, q = 0.049. 3 ### Page 5 Loadings Prosociality Social Optimism Acquiescence UniquenessPGG1 0.3146 0.3613 -0.0759 0.7647 PGG2 0.3075 0.1962 -0.1183 0.8529 PGG3 0.2588 0.2890 -0.0446 0.8475 DG1 0.4740 0.0963 -0.0170 0.7658 DG2 0.3323 -0.0579 -0.0731 0.8809 DG3 0.4401 0.0760 -0.0161 0.8003 UG0 0.2411 -0.0106 0.3298 0.8329 UG1 0.1807 -0.0348 0.4179 0.7915 UG2 0.0426 0.0470 0.1818 0.9629 UG3 -0.1836 -0.0397 0.5989 0.6060 UG4 -0.1001 0.0346 0.5851 0.6465 DEG1 -0.5444 -0.0331 0.1586 0.6774 DEG2 -0.3936 0.0615 -0.0184 0.8410 DEG3 -0.5657 -0.0030 0.0805 0.6735 MDF -0.0097 -0.1556 -0.0300 0.9748 MDB 0.0608 -0.0636 -0.0658 0.9879 BEL1 -0.0397 0.5835 -0.0178 0.6576 BEL2 0.0324 0.5584 0.0349 0.6859 BEL3 0.0555 0.5565 -0.0153 0.6870Supplementary Table 3: Factor loadings.The table reports the factor loadings for the 3 retained factors in the first 3 columns, and the uniqueness coefficient for each task in the fourth column. Loadings with weights above 0.1 appear in blue, while those with weights below -0.1 appear in red.Deception GameIn the DEG, preschoolers in the fast condition are less likely to deceive than those in the slow condition, although this effect diminishes with age. Older children are generally less likely to deceive than younger ones, and deception increases in later rounds.Moral DilemmasChildren are more likely to act in the Bystander dilemma than in the Footbridge dilemma. Older children are more likely to act in both types of dilemmas. Finally, we find no evidence that condition is associated with children’s choices.A.3 Non-Linear Age RegressionsThis section replicates the earlier analysis, but instead of using age as a continuous control, participants are grouped into 4 balanced age categories: 3–4, 5–6, 7–8, and 9–10 years. The insights from these regressions align with those of the main analysis. However, these robustness checks highlight that the association with age is not linear. Rather, it appears to undergo developmental changes at specific ages. These results are reported in Supplementary Table 6, and Supplementary Figure 1. These conclusions are further corroborated by local polynomial regressions in Supplementary Figure 2. The linear regressions reveal several key insights. In the context of Prosociality, children in the fast condition in the 3–4 age group showed higher prosocial tendencies 4 ### Page 6 Factor Prosociality Social Optimism Acquiescence Constant −0.715 −0.153 1.043 (0.192) (0.198) (0.158) Fast 0.997*** 0.215 −0.535* (0.248) (0.257) (0.222) [<0.001] [0.169] [0.017] Age 0.086*** 0.020 −0.144*** (0.024) (0.025) (0.022) [<0.001] [0.171] [<0.001] Age × Fast −0.113*** −0.027 0.069* (0.032) (0.032) (0.030) [<0.001] [0.169] [0.020] R-squared 0.044 0.002 0.111 Observations 522 522 522Supplementary Table 4: Linear regressions by factor.Heteroskedasticity robust standard errors appear in parentheses (df = 518). Sharpened FDR q-values appear in square brackets. Significance stars are based on sharpened FDR q-values, with * denoting q < .050, ** denoting q < .010, *** denoting q < .001; no stars are reported for constants as they reflect factor normalization rather than substantive hypotheses.compared to those in the slow condition, b = 0.51, t(514) = 3.25, q = 0.004, 95% CI = [0.20, 0.83]. Moreover, the difference between fast and slow decision modes is significantly larger among 5–6-year-olds than among 9–10-year-olds, b = 0.41, t(514) = 2.17, q = 0.038, 95% CI = [0.04, 0.79], highlighting that the effect of decision mode on Prosociality undergoes meaningful shifts across development. Acquiescence declines markedly at ages 9–10 in both conditions, b = −0.88, t(514) = −6.57, q < 0.001, 95% CI = [−1.14, −0.62], though the decline is attenuated under time pressure, b = 0.42,t(514) = 2.26, q = 0.032, 95% CI = [0.05, 0.79], with similar patterns emerging at ages 7–8. These insights highlight that development is often associated with non-linear shifts in factors, as seen by the changes in coefficients across age groups. 5 ### Page 7 Task PGG DG UG DEG MD Constant 0.718 1.921 1.200 0.726 0.608 (0.294) (0.246) (0.056) (0.085) (0.076) Fast 0.537* 0.263 −0.134* −0.625*** 0.040 (0.287) (0.297) (0.074) (0.107) (0.102) [0.048] [0.166] [0.049] [<0.001] [0.264] Age 0.105*** 0.030 −0.038*** −0.038*** 0.028** (0.029) (0.030) (0.007) (0.011) (0.010) [<0.001] [0.152] [<0.001] [<0.001] [0.006] Age × Fast −0.066* −0.024 0.017 0.071*** −0.004 (0.036) (0.036) (0.010) (0.014) (0.013) [0.048] [0.210] [0.056] [<0.001] [0.297] Round 2 0.328*** 0.297*** 0.069* (0.066) (0.065) (0.027) [<0.001] [<0.001] [0.011] Round 3 −0.034 0.229*** 0.107*** (0.063) (0.059) (0.020) [0.242] [<0.001] [<0.001] Beliefs 0.396*** (0.076) [<0.001] Offer 0 −0.398*** (0.024) [<0.001] Offer 1 −0.232*** (0.022) [<0.001] Offer 3 −0.170*** (0.021) [<0.001] Offer 4 −0.175*** (0.021) [<0.001] Footbridge −0.188*** (0.024) [<0.001] R-squared 0.034 0.013 0.107 0.049 0.061 Observations 1572 1575 2625 1575 1047Supplementary Table 5: Linear regression by task.Heteroskedasticity robust standard errors clustered by participant appear in parentheses. Degrees of free- dom: PGG df = 523; DG df = 524; UG df = 524; DEG df = 524; MD df = 523. Sharpened FDR q-values appear in square brackets. Significance stars are based on sharpened FDR q-values, with * denoting q < .050, ** denoting q < .010, *** denoting q < .001; no stars are reported for constants as they capture baseline levels rather than the treatment effects or developmental patterns of interest.6 ### Page 8 Factor Prosociality Social Optimism Acquiescence Constant −0.186 −0.024 0.453 (0.130) (0.154) (0.097) Fast 0.512** 0.067 −0.297* (0.163) (0.190) (0.137) [0.004] [0.500] [0.037] Age 5-6 −0.086 −0.104 −0.134 (0.167) (0.180) (0.125) [0.443] [0.428] [0.260] Age 7-8 −0.099 0.080 −0.672*** (0.163) (0.182) (0.129) [0.428] [0.464] [<0.001] Age 9-10 0.514** 0.081 −0.881*** (0.158) (0.173) (0.134) [0.003] [0.460] [<0.001] Age 5-6 × Fast −0.274 0.093 0.115 (0.215) (0.230) (0.184) [0.200] [0.477] [0.426] Age 7-8 × Fast −0.264 −0.154 0.469* (0.211) (0.228) (0.180) [0.200] [0.423] [0.016] Age 9-10 × Fast −0.686** −0.125 0.421* (0.208) (0.220) (0.187) [0.003] [0.428] [0.032] R-squared 0.078 0.006 0.138 Observations 522 522 522Supplementary Table 6: Factor regressions with age groups.Heteroskedasticity robust standard errors appear in parentheses (df = 514). Sharpened FDR q-values appear in square brackets. Significance stars are based on sharpened FDR q-values, with * denoting q < .050, ** denoting q < .010, *** denoting q < .001; no stars are reported for constants as they reflect factor normalization rather than substantive hypotheses.7 ### Page 9 -.6 -.3 0 .3 .6 Factor Loading 1 2 3 1 2 3 0 1 2 3 4 1 2 3 F B 1 2 3 PGG DG UG DEG MD BEL Prosociality Optimism Acquiescence(a) Factor Loadings-.6 -.3 0 .3 .6 Predicted Prosociality Age 3-4 Age 5-6 Age 7-8 Age 9-10 Age Groups Slow Fast (b) Prosociality-.6 -.3 0 .3 .6 Predicted Optimism Age 3-4 Age 5-6 Age 7-8 Age 9-10 Age Groups Slow Fast(c) Social Optimism-.6 -.3 0 .3 .6 Predicted Acquiescence Age 3-4 Age 5-6 Age 7-8 Age 9-10 Age Groups Slow Fast (d) AcquiescenceSupplementary Figure 1: Factor loadings and differences in factors by age group and decision mode.Panel (a) depicts the factor loading for the 3 retained factors. Panels (b)–(d) depict the association with age groups estimated from linear regressions with factors as dependent variables and decision mode as control (n = 522). Bars represent mean values; error bars represent 95% CIs. All tests are two-sided and statistical significance is based on sharpened FDR q-values; full test statistics, degrees of freedom, confidence intervals, and q-values are reported in the main text and Supplementary Table 6.8 ### Page 10 -.6 -.3 0 .3 .6 Factor Loading 1 2 3 1 2 3 0 1 2 3 4 1 2 3 F B 1 2 3 PGG DG UG DEG MD BEL Prosociality Optimism Acquiescence(a) Factor Loadings-.6 -.3 0 .3 .6 Predicted Prosociality 3 5 7 9 11 Age Slow Fast (b) Prosociality-.6 -.3 0 .3 .6 Predicted Optimism 3 5 7 9 11 Age Slow Fast(c) Social Optimism-.6 -.3 0 .3 .6 Predicted Acquiescence 3 5 7 9 11 Age Slow Fast (d) AcquiescenceSupplementary Figure 2: Factor loadings and non-linear differences in factors by age and decision mode.Panel (a) depicts the factor loading for the 3 retained factors. Panels (b)–(d) depict the association with age estimated from local polynomial regressions with factors as dependent variables and decision mode as control (n = 522). Lines represent local polynomial estimates; shaded bands represent 95% CIs. Note that local polynomial regressions are presented for descriptive purposes only; all formal tests are two-sided and based on the linear regressions reported in the main text and Supplementary Tables.9 ### Page 11 1.8 2.1 2.4 2.7 3 PGG Contibution 3 5 7 9 11 Age Slow Fast(a) Public Goods Game1.8 2.1 2.4 2.7 3 DG Donation 3 5 7 9 11 Age Slow Fast (b) Dictator Game0 .25 .5 .75 1 Probability of Deception 3 5 7 9 11 Age Slow Fast(c) Deception Game0 .25 .5 .75 1 Probability of Action 3 5 7 9 11 Age Slow Fast (d) Moral DilemmaSupplementary Figure 3: Non-linear age and decision mode differences on PGG, DG, DEG and MD tasks.Panels (a)–(d) depict the association with age estimated from local polynomial regressions with behaviour in PGG (n observations = 1,572), DG (n = 1,575), DEG (n = 1,575), and MD (n = 1,047) respectively as dependent variables and decision mode as control. Lines represent local polynomial estimates; shaded bands represent 95% CIs clustered by participant. Note that local polynomial regressions are presented for descriptive purposes only; all formal tests are two-sided and based on the linear regressions reported in the main text and Supplementary Tables.10 ### Page 12 0 .25 .5 .75 1 Probability of Acceptance 3 5 7 9 11 Age Slow Fast(a) Offer 00 .25 .5 .75 1 Probability of Acceptance 3 5 7 9 11 Age Slow Fast (b) Offer 10 .25 .5 .75 1 Probability of Acceptance 3 5 7 9 11 Age Slow Fast(c) Offer 20 .25 .5 .75 1 Probability of Acceptance 3 5 7 9 11 Age Slow Fast(d) Offer 30 .25 .5 .75 1 Probability of Acceptance 3 5 7 9 11 Age Slow Fast (e) Offer 4Supplementary Figure 4: Non-linear age and decision mode differences on UG acceptance.Panels (a)–(e) depict the association with age estimated from local polynomial regressions with behaviour in the 5 UG trials (4-0, 3-1, 2-2, 1-3, 0-4) respectively as dependent variables and decision mode as control (n= 2,625 observations, 524 participants). Lines represent local polynomial estimates; shaded bands represent 95% CIs clustered by participant. Note that local polynomial regressions are presented for descriptive purposes only; all formal tests are two-sided and based on the linear regressions reported in the main text and Supplementary Tables.11 ### Page 13 B Robustness ChecksB.1 Sample Inclusion CriteriaSupplementary Tables 7 and 8 replicate our main analysis, but include the partic- ipants who failed the comprehension checks. The results remain unaffected by this adjustment. In total, 10 participants failed 12 comprehension checks, with 2 failing more than one check. Supplementary Tables 9 and 10 replicate our main analysis, excluding participants who did not comply with the fast condition. The results remain unaffected by this adjustment. In total, 66 participants were removed from the primary analysis due to non-compliance. Supplementary Tables 11 and 12 replicate our main analysis, excluding partici- pants aged 3–4 years. The results are robust to this sample restriction. In total, 117 participants aged 3–4 were removed from this analysis.Loadings Prosociality Social Optimism Acquiescence UniquenessPGG1 0.3163 0.3570 -0.0616 0.7687 PGG2 0.3129 0.2003 -0.1167 0.8483 PGG3 0.2595 0.2934 -0.0370 0.8452 DG1 0.4705 0.1077 -0.0233 0.7664 DG2 0.3392 -0.0571 -0.0696 0.8768 DG3 0.4499 0.0761 -0.0126 0.7916 UG0 0.2150 -0.0071 0.3500 0.8312 UG1 0.1645 -0.0379 0.4333 0.7838 UG2 0.0391 0.0517 0.1798 0.9635 UG3 -0.1919 -0.0499 0.5922 0.6100 UG4 -0.1102 0.0306 0.5771 0.6539 DEG1 -0.5539 -0.0170 0.1549 0.6689 DEG2 -0.3872 0.0713 -0.0264 0.8443 DEG3 -0.5683 0.0109 0.0687 0.6722 MDF -0.0151 -0.1361 -0.0384 0.9798 MDB 0.0705 -0.0443 -0.0721 0.9879 BEL1 -0.0469 0.5804 -0.0189 0.6605 BEL2 0.0251 0.5454 0.0140 0.7017 BEL3 0.0512 0.5552 -0.0155 0.6888Supplementary Table 7: Factor loadings with participants failing com- prehension.The table reports the factor loadings for the 3 retained factors in the first 3 columns, and the uniqueness coefficient in the fourth column. Loadings with weights above 0.1 appear in blue, while those with weights below -0.1 appear in red.12 ### Page 14 Factor Prosociality Social Optimism Acquiescence Constant −0.731 −0.167 1.026 (0.185) (0.182) (0.148) Fast 1.020*** 0.177 −0.516* (0.240) (0.246) (0.213) [<0.001] [0.251] [0.017] Age 0.089*** 0.023 −0.143*** (0.024) (0.023) (0.021) [<0.001] [0.187] [<0.001] Age × Fast −0.115*** −0.023 0.067* (0.031) (0.031) (0.029) [<0.001] [0.251] [0.021] R-squared 0.048 0.002 0.115 Observations 532 532 532Supplementary Table 8: Regressions with participants failing comprehension.Heteroskedasticity robust standard errors appear in parentheses (df = 528). Sharpened FDR q-values appear in square brackets. Significance stars are based on sharpened FDR q-values, with * denoting q < .050, ** denoting q < .010, *** denoting q < .001; no stars are reported for constants as they reflect factor normalization rather than substantive hypotheses.Loadings Prosociality Social Optimism Acquiescence UniquenessPGG1 0.3120 0.3520 -0.0638 0.7747 PGG2 0.3064 0.2139 -0.0950 0.8513 PGG3 0.2858 0.2799 -0.0280 0.8392 DG1 0.4514 0.0840 -0.0218 0.7887 DG2 0.3153 -0.0667 -0.0897 0.8881 DG3 0.4109 0.0769 -0.0266 0.8246 UG0 0.2263 -0.0289 0.3319 0.8378 UG1 0.2071 -0.0774 0.3993 0.7916 UG2 0.0335 0.0238 0.1240 0.9829 UG3 -0.1725 -0.0382 0.6092 0.5976 UG4 -0.1149 0.0400 0.6029 0.6217 DEG1 -0.5466 -0.0459 0.1733 0.6690 DEG2 -0.3683 0.0720 0.0101 0.8591 DEG3 -0.5735 -0.0108 0.0925 0.6624 MDF -0.0722 -0.1659 -0.0057 0.9672 MDB 0.0526 -0.0681 -0.0446 0.9906 BEL1 -0.0561 0.6098 -0.0025 0.6250 BEL2 0.0606 0.5467 0.0275 0.6967 BEL3 0.0721 0.5593 -0.0393 0.6804Supplementary Table 9: Factor loadings without non-compliants.The table reports the factor loadings for the 3 retained factors in the first 3 columns, and the uniqueness coefficient in the fourth column. Loadings with weights above 0.1 appear in blue, while those with weights below -0.1 appear in red.13 ### Page 15 Factor Prosociality Social Optimism Acquiescence Constant −0.682 −0.148 1.096 (0.199) (0.205) (0.160) Fast 0.723** 0.217 −0.581* (0.264) (0.277) (0.243) [0.009] [0.274] [0.019] Age 0.087** 0.019 −0.151*** (0.025) (0.026) (0.022) [0.001] [0.277] [<0.001] Age × Fast −0.082* −0.026 0.077* (0.034) (0.034) (0.033) [0.019] [0.274] [0.021] R-squared 0.036 0.002 0.123 Observations 457 457 457Supplementary Table 10: Regressions without non- compliants.Heteroskedasticity robust standard errors appear in parentheses (df = 453). Sharpened FDR q-values appear in square brackets. Significance stars are based on sharpened FDR q-values, with * denoting q < .050, ** denoting q < .010, *** denoting q < .001; no stars are reported for constants as they reflect factor normalization rather than substantive hypotheses.Loadings Prosociality Social Optimism Acquiescence UniquenessPGG1 0.3811 0.2692 -0.0933 0.7736 PGG2 0.2782 0.0785 -0.1507 0.8937 PGG3 0.2607 0.1856 -0.0349 0.8964 DG1 0.4764 0.0256 -0.0275 0.7717 DG2 0.2746 -0.1337 -0.1302 0.8898 DG3 0.3966 0.0120 -0.0801 0.8362 UG0 0.3593 0.0088 0.1689 0.8423 UG1 0.2560 -0.0272 0.2546 0.8689 UG2 0.0682 0.0177 0.0509 0.9925 UG3 -0.1335 -0.0490 0.6180 0.5979 UG4 -0.0443 0.0013 0.6014 0.6363 DEG1 -0.4539 -0.0305 0.2875 0.7104 DEG2 -0.3117 0.0545 0.0319 0.8988 DEG3 -0.4986 -0.0397 0.2124 0.7047 MDF -0.0279 -0.1190 -0.0318 0.9840 MDB 0.0485 0.0053 -0.0408 0.9960 BEL1 -0.0086 0.6018 -0.0042 0.6377 BEL2 0.0538 0.5244 -0.0338 0.7210 BEL3 0.0397 0.5790 -0.0308 0.6622Supplementary Table 11: Factor loadings excluding 3–4 year olds.The table reports the factor loadings for the 3 retained factors in the first 3 columns, and the uniqueness coefficient in the fourth column. Loadings with weights above 0.1 appear in blue, while those with weights below -0.1 appear in red. Sample restricted to participants aged 5 years and older.14 ### Page 16 Prosociality Social Optimism Acquiescence Constant −1.084 −0.272 1.308 (0.278) (0.253) (0.238) Fast 0.937* 0.340 −0.491 (0.367) (0.357) (0.346) [0.018] [0.301] [0.162] Age 0.130*** 0.034 −0.166*** (0.033) (0.030) (0.030) [<0.001] [0.242] [<0.001] Age × Fast −0.104* −0.042 0.064 (0.044) (0.043) (0.043) [0.027] [0.297] [0.146] R-squared 0.047 0.003 0.104 Observations 405 405 405Supplementary Table 12: Regressions excluding 3–4 year olds.Heteroskedasticity robust standard errors appear in parentheses (df = 401). Sharpened FDR q-values appear in square brackets. Significance stars are based on sharpened FDR q-values, with * denoting q < .050, ** denoting q < .010, *** denoting q < .001; no stars are reported for constants as they reflect factor normalization rather than substantive hypotheses. Sample restricted to participants aged 5 years and older.15 ### Page 17 B.2 Task Selection CriteriaSupplementary Tables 13 and 14 replicate our main analysis, excluding tasks with uniqueness coefficients exceeding 0.9 from the main factor analysis. In total, 3 trials were removed from the analysis due to high uniqueness values (UG2, MDF, and MDB). The results are robust to this adjustment. Supplementary Tables 15 and 16 replicate our main analysis, excluding all Pub- lic Goods Game trials and their associated belief elicitations. In total, 6 trials were removed from this analysis (PGG1, PGG2, PGG3, BEL1, BEL2, BEL3). The results are robust to this adjustment, confirming that our findings are not driven by the inclusion of the PGG task. Note that this exclusion necessarily eliminates the Social Optimism factor, which was primarily associated with the belief elicitation trials.Loadings Prosociality Social Optimism Acquiescence UniquenessPGG1 0.3175 0.3814 -0.0503 0.7512 PGG2 0.3033 0.2075 -0.1051 0.8539 PGG3 0.2696 0.2866 -0.0225 0.8447 DG1 0.4748 0.1031 -0.0184 0.7636 DG2 0.3375 -0.0560 -0.0730 0.8776 DG3 0.4412 0.0804 -0.0058 0.7988 UG0 0.2377 -0.0102 0.3370 0.8298 UG1 0.1781 -0.0407 0.4074 0.8007 UG3 -0.1799 -0.0439 0.6006 0.6050 UG4 -0.0945 0.0280 0.5834 0.6499 DEG1 -0.5420 -0.0220 0.1751 0.6750 DEG2 -0.3999 0.0624 -0.0236 0.8356 DEG3 -0.5668 0.0023 0.0910 0.6705 BEL1 -0.0381 0.5822 -0.0075 0.6596 BEL2 0.0177 0.5460 0.0103 0.7014 BEL3 0.0387 0.5514 -0.0271 0.6937Supplementary Table 13: Factor loadings excluding high uniqueness tasks.The table reports the factor loadings for the 3 retained factors in the first 3 columns, and the uniqueness coefficient in the fourth column. Loadings with weights above 0.1 appear in blue, while those with weights below -0.1 appear in red. Tasks with uniqueness exceeding 0.9 (UG2, MDF, MDB) are excluded from this analysis.16 ### Page 18 Factor Prosociality Social Optimism Acquiescence Constant −0.681 −0.219 1.055 (0.191) (0.197) (0.150) Fast 0.996*** 0.241 −0.570* (0.247) (0.253) (0.212) [<0.001] [0.301] [0.013] Age 0.082** 0.029 −0.146*** (0.024) (0.024) (0.021) [0.002] [0.216] [<0.001] Age × Fast −0.113** −0.030 0.074* (0.032) (0.032) (0.029) [0.002] [0.302] [0.018] R-squared 0.042 0.004 0.113 Observations 524 524 524Supplementary Table 14: Regressions excluding high uniqueness tasks.Heteroskedasticity robust standard errors appear in parentheses (df = 520). Sharpened FDR q-values appear in square brackets. Significance stars are based on sharpened FDR q-values, with * denoting q < .050, ** denoting q < .010, *** denoting q < .001; no stars are reported for constants as they reflect factor normalization rather than substantive hypotheses.Loadings Prosociality Acquiescence UniquenessDG1 0.4597 -0.0100 0.7885 DG2 0.3210 -0.0657 0.8927 DG3 0.4440 -0.0107 0.8028 UG0 0.2073 0.3348 0.8449 UG1 0.1870 0.4107 0.7964 UG2 0.0645 0.1694 0.9672 UG3 -0.1939 0.5975 0.6054 UG4 -0.0984 0.5738 0.6610 DEG1 -0.5711 0.1578 0.6490 DEG2 -0.3913 -0.0231 0.8464 DEG3 -0.5869 0.0800 0.6492 MDF -0.0336 -0.0197 0.9985 MDB 0.0669 -0.0614 0.9917Supplementary Table 15: Factor loadings excluding PGG and belief tasks.The table reports the factor loadings for the 2 retained factors in the first 2 columns, and the uniqueness coefficient in the third column. Loadings with weights above 0.1 appear in blue, while those with weights below -0.1 appear in red. All PGG trials (PGG1-3) and belief elicitation trials (BEL1-3) are excluded, eliminating the Social Optimism factor.17 ### Page 19 Factor Prosociality Acquiescence Constant −0.624 0.990 (0.191) (0.155) Fast 1.002*** −0.521* (0.242) (0.218) [<0.001] [0.026] Age 0.073** −0.137*** (0.024) (0.021) [0.006] [<0.001] Age × Fast −0.113** 0.068* (0.032) (0.030) [0.001] [0.030] R-squared 0.042 0.102 Observations 523 523Supplementary Table 16: Regressions excluding PGG and belief tasks.Heteroskedasticity robust standard errors appear in parentheses (df = 519). Sharpened FDR q-values appear in square brackets. Significance stars are based on sharpened FDR q-values, with * denoting q < .050, ** denoting q < .010, *** denoting q < .001; no stars are reported for constants as they reflect factor normalization rather than substantive hypotheses.18 ### Page 20 B.3 Time Pressure and Random ChoiceSupplementary Table 17 examines whether time pressure moves younger children’s responses toward random choice. We focus on 3–4 year old participants and test whether choices differ between time pressure and time delay conditions. To avoid distributional assumptions on the residuals, we conduct exact two-sided Wilcoxon rank-sum tests task by task. We find no systematic tendency for time pressure to move participants towards random choice. In 5 of the 9 tasks, behaviour in the fast condi- tion is farther from random choice than behaviour in the slow condition. In the DEG task, choices differ significantly between conditions, with behaviour in the fast condi- tion deviating more from random choice. In UG0, choices also differ significantly, with behaviour in the fast condition closer to random choice, but this does not extend to the other UG tasks (or any other task).Task Mean Choice Mean Choice Random z-statistic Exact Time Pressure Time Delay Choice p-valuePGG 2.44 2.17 2.00 1.58 0.112 DG 2.43 2.36 2.00 0.23 0.819 UG0 0.54 0.75 0.50 -2.37 0.027* UG1 0.73 0.86 0.50 -1.69 0.139 UG2 0.92 0.91 0.50 0.19 1.000 UG3 0.86 0.91 0.50 -0.90 0.538 UG4 0.84 0.89 0.50 -0.82 0.583 DEG 0.21 0.60 0.50 4.07 0.000*** MD 0.61 0.58 0.50 0.51 0.707Supplementary Table 17: Testing whether fast condition moves 3–4 year olds toward random choice.The table reports mean choices for 3–4 year old participants in fast and slow conditions, the expected value under random choice, and two-sided exact Wilcoxon rank-sum tests for whether time pressure and time delay choices differ. Positive z-statistics indicate time pressure choices are farther from random choice than time delay choices, and negative that they are closer. All tests are two-sided. Significance is indicated with stars, with * denoting p < .050, ** denoting p < .010, *** denoting p < .001.19 ### Page 21 B.4 Estimating Factors by AgeThe following linear regressions use the factors estimated by splitting the dataset, where factors are estimated using different weights above and below the median age in the sample. This robustness check demonstrates that the composition of 2 out of the 3 factors is unaffected by age. However, Acquiescence is no longer significant as a factor among older participants, and is consequently dropped from the analysis. As was the case when factors were estimated on the full panel, linear regressions show that condition (fast, slow), age, and their interaction impact on Prosociality, but not on Social Optimism.Factor Prosociality Social Optimism Constant −0.760*** −0.008 (0.194) (0.199) Fast 1.254*** 0.033 (0.253) (0.252) Age 0.091*** 0.000 (0.025) (0.026) Age × Fast −0.145*** −0.003 (0.033) (0.033) R-squared 0.055 0.001 Observations 522 522Supplementary Table 18: Split sample factors.Heteroskedasticity robust standard errors appear in parentheses (df = 518). Significance is based on unad- justed p-values, with * denoting p < .050, ** denoting p < .010, *** denoting p < .001. This analysis serves as a descriptive validation of the factor structure and was not included in the sharpened FDR correction as it does not constitute a test of the treatment or developmental hypotheses of interest.20 ### Page 22 C Additional Scripts For Each TaskC.1 The PGG Task (Explanatory Movie-S1)To explain the PGG to the children, the experimenter first showed them a short movie (Movie S1), which we detail here. The movie shows a male child actor sitting at a table. The child actor is on the left side of the scene. In front of him is an orange bowl (the common pool), with a female adult actor positioned behind the table, on the child’s left, facing the viewer. She explains that the child on her right will play with 3 other players, each having 4 candies. The candies in front of the child actor are his own, and he decides what to do with them. The candies are taped together in pairs, totaling 4 in front of the child actor, between him and the bowl. The 3 partners are not real children but small closed boxes with images of children’s faces (7 x 7 cm drawings) attached to them. Facing the viewer, the adult actor says, “Hello and welcome! Today I’ll explain the rules of a fun game. Listen carefully. Look! He is a child just like you.” The child actor turns to the camera and says, “Hi! I’m Joseph.” The adult continues, “Today he will play this game. Look, he has 4 candies.” She picks up the candies and adds, “They are his own, and he can decide what to do with them.” She then places them back and explains, “But in this game, you don’t play alone. There are 3 other children... this child here, this child here, and this child here.” The adult points to the 3 boxes, saying, “Each one of them, like Joseph, has 4 candies, and each decided how many to put in their box.” The adult actor says, “Now, let me explain how this works! Joseph has 4 candies and can decide to put 2 in the common bowl and keep 2 for himself.” The child actor puts 2 candies in the bowl and keeps 2. The adult continues, “Or, he can put all 4 in the common bowl and keep none.” The child actor places all the candies in the bowl. The adult then adds, “Or he can keep all 4 for himself, putting none in the bowl.” The child actor keeps all 4 candies. The adult actor looks at the camera, “Let’s see an example so you understand. Joseph has 4 candies and can decide how to use them. Let’s look at what Joseph decides to do.” The child actor places 2 candies in the bowl and keeps 2. The adult comments, “Joseph has decided to put 2 candies in the common bowl and keep 2 for himself.” The child actor brings the 2 candies closer. The adult continues, “The other 3 children already decided how many candies to put in the bowl, but we don’t know yet how many. Let’s find out.” The adult opens each box, revealing the candies inside, placing them in front of the corresponding images, saying, “This child put 2 candies, this one put 4, and this one put none.” She adds, “Now, I’ll take them and put them all in the common bowl.” The adult grabs the bowl and shows it to the camera, “Do you know what happens when the candies are put in the common bowl? They double in number!” She then demonstrates, lining up the candies in front of her and adding an equal number from a small bag placed just outside the scene, helping the viewer understand that the total quantity is the same. She continues, “Now I’ll put them all back in the bowl. And I’ll distribute them equally among all the players, so everyone gets the same number of candies.” The adult places 4 candies in front of each player, saying, “4 candies for 21 ### Page 23 Joseph, 4 for this child, 4 for this child, and 4 for this child.” Lastly, the adult engages the viewer: “Did you see what happened? Joseph had 4 candies, now he has 6, and he can take them home. But how many candies he gets depends on how many each child put in the bowl. Are you ready to play now?”C.2 Additional Scripts and InstructionsPublic Goods GameBefore asking the child participant to make her decision on the first trial, the exper- imenter repeated the instructions one last time: “Here are your candies. These are yours, and you can decide what to do with them—whether to put 2, all, or none in the common bowl, and how many to keep for yourself. Remember, if you put them in the common bowl, they will mix with those that the other children decided to give. In fact, now you will play with 3 children who have already played the game. You see, they have already made their choices, but we don’t know how many candies they put in their boxes. Their boxes are closed, and we still don’t know what they decided to put in the common bowl.”Dictator GameIn the fast condition (assuming the DG was administered after another task, where the child was already introduced to the idea of responding within 10 seconds), the experimenter prompted the child by saying, “(As in the previous game,) in this game too, you have to decide right away what to do. You only have 10 seconds to decide how many candies to give to this child, so you have to be really quick! Are you ready to go?” In the slow condition, the experimenter instead said, “(As in the previous game,) in this game too, you have to think very carefully before deciding how many candies to give to this child. You can’t choose right away. Before deciding, you’ll have to wait at least 10 seconds and think very, very carefully! After 10 seconds, you can take all the time you need to decide. I’m in no hurry, think carefully about what to do!”Ultimatum GameIn the fast condition (assuming the first trial was 2-2, and UG was not the first task), to prompt the child’s response, the experimenter said, “(As in the previous game,) in this game too, you have to decide right away whether to say yes or no. Listen, this child suggested keeping 2 candies for herself and giving you 2 candies. Are you okay with that? Tell me yes or no. You only have 10 seconds to decide, so you have to be really quick! Go!” In the slow condition, the experimenter said, “(As in the previous game,) in this game too, you need to think very carefully before deciding whether to say yes or no. You can’t choose right away. Before deciding, you’ll have to wait at least 10 seconds and think very, very carefully! After 10 seconds, you can take as long as you need to decide. I’m in no hurry, think carefully about how to respond!” 22 ### Page 24 Deception GameWhile giving the instructions, the experimenter also provided an example of how the game worked: “I’ll show you how this game works. Look, to find out if I’m yellow or green, I’ll draw a card. Oh, I’m yellow! See, the card is yellow! This means I am truly yellow, and this child here is green. Look, now I say “yellow!,” and see, because I told the truth and said the true color of the card I drew, I receive 1 candy and she takes 3. Now, look what happens if I say “green!” Ah, it’s not true that I’m green, but I can say green if I want, and if I say the false color, I take 3 candies and she only takes 1.” Next, participants were instructed on how to respond. In the fast condition, the experimenter said, “(As in the previous game,) in this game too, you have to decide right away how to respond—whether to say yellow or green. The color I’m showing you is the color of your group. As soon as I show it to you, I’ll ask you to tell me the color of your group. You can say the true color and receive 1 candy, giving 3 to this child, or you can say the false color and receive 3 candies, giving 1 to this child. It’s up to you! Remember, you only have 10 seconds to decide, so you have to be really quick!” After drawing a card for the child, the experimenter asked, “In which group are you, the yellows or the greens?” In the slow condition, the experimenter said, “(As in the previous game,) in this game too, you need to think very carefully before telling me if you are in the yellows or the greens. You can’t say the color right away. After I show you the card with the color of your group and ask you the question, you’ll have to wait at least 10 seconds and think very, very carefully! After 10 seconds, you can take all the time you want to decide. I’m in no hurry, think carefully about how to respond!”Moral DilemmasBefore the first dilemma, the experimenter introduced the task: “In this game, I’ll tell you 2 stories. On the screen, you will see a story with some cartoons. It’s important that you listen carefully to what happens in the story because later you will have to decide how the story should end. You decide the ending!” The experimenter then asked, “Who will decide how the story ends?” Once the experimenter ensured the child understood that they had to decide, she proceeded with the first story. For the Bystander Dilemma, the experimenter followed this script: “Mary/Tom is going for a walk. S/he is walking down a hill. [Animation 1: A ball appears at the top of the hill] S/he sees that a giant ball is rolling down the trail. [Animation 2: 5 children appear] The ball is rolling directly toward the 5 children. The children cannot escape because they are looking straight ahead and cannot hear or see the ball.” Participants then received the first probe: “Do the children see the ball rolling down the hill?” After the probe, the experimenter continued: “[Animation 3: The ball rolls down a bit] The ball is rolling down the path and will hurt the 5 children. [Animation 4: The ball is deflected] Mary/Tom knows that if s/he pushes the fence across the trail, the ball will be pushed over to the other trail. [Animation 5: A lone child appears] But on this other trail, there is another child walking down the hill who cannot see or hear the ball coming.” At this point, the second probe was presented: “Does this child see or hear the ball rolling down the hill?” After the probe, the experimenter continued: “If the ball is pushed sideways, this child will get hurt, but the other 5 will be safe.” 23 ### Page 25 Then, the first animation was shown again and the child received the test question: “What should Mary/Tom do? Push the ball over to the side or not push the ball?” For the Footbridge Dilemma, the experimenter said: “Luisa/John is going for a walk. S/he is walking down a hill. [Animation 1: A ball appears at the top of the hill] S/he sees that a giant ball is rolling down the trail. [Animation 2: 5 children appear] The ball is rolling directly toward the 5 children. The children cannot escape because they are looking straight ahead and cannot hear or see the ball.” After receiving the same probe as in the previous dilemma, children were told: “[Animation 3: The ball rolls down a bit] The ball is rolling down the path and will hurt the 5 children. [Animation 4: Bridge and child with backpack appear] Luisa/John is on the bridge between the ball and the 5 children. There is also a child with a big backpack on the bridge. This child cannot hear or see the ball coming because s/he is looking straight ahead.” After the second probe, as in the previous dilemma, children were told: “Luisa/John knows that the only way to stop the ball is to block it with something heavy. [Animation 5: The child is pushed] Luisa/John knows that if s/he pushes the child with the big backpack down onto the trail, the child will get hurt, but the other 5 children will be safe.” The first animation was presented again, and children received the test question: “What should Luisa/John do? Push the child down onto the trail or not push the child?” In the fast condition, right before the test question, the experimenter said, “(As in the previous game,) in this game too, you have to decide right away what Maria/Luisa should do. You only have 10 seconds to give your answer, so you have to be really quick!” In the slow condition, the experimenter said, “(As in the previous game,) in this game too, you need to think very carefully before deciding what Maria/Luisa should do. You can’t answer right away. Before giving me your answer, you’ll have to wait at least 10 seconds and think very, very carefully! After 10 seconds, you can take all the time you want to respond. I’m in no hurry, think carefully about how to respond!”Bystander Dilemma Footbridge DilemmaSupplementary Figure 5: Bystander and Footbridge Dilemmas stimuli.The figure shows the first PowerPoint slide of the animation presented to children to illustrate the Bystander Dilemma (left) and the Footbridge Dilemma (right). In the figure, the female version of the stimuli is shown. A male version was used with male participants. Stimuli were originally designed by Dworazik et al. (2019).24