Research Article

Neuroscience

The self-interest of adolescents overrules cooperation in social dilemmas

State Key Laboratory of Cognitive Neuroscience and Learning, and IDG/McGovernInstitute for Brain Research, Beijing Normal University, China
Beijing Key Laboratory of Artificial Intelligence Safety and Superalignment, China
Department of Adult Psychiatry and Psychotherapy, University of Zurich, Switzerland
Zurich Center for Neuroeconomics, Department of Economics, University of Zurich, Switzerland
School of Psychology, South China Normal University, China
Beijing Institute of AI Safety and Governance, China
Brain-inspired Cognitive AI Lab, Institute of Automation, Chinese Academy of Sciences, China
University of Chinese Academy of Sciences, China
Long-term AI, China

Mar 31, 2026

https://doi.org/10.7554/eLife.106840.4

Open access
Copyright information

Figures
Tables
Additional files

13 figures, 14 tables and 1 additional file

Figures

Figure 1

Download asset Open asset

Experiment setup and behavioral results.

(a) Partner’s cooperation probability: in the first half of the 120 trials, the partner cooperated 78% of the time; in the second half, cooperation alternated between 20% and 80%. (b) Payoff matrix: payoffs are 4 for mutual cooperation, 2 for mutual defection, 0 for cooperation when the other defects, and 6 for defecting when the other cooperates. (c) Trial illustration: after a 0.5 s fixation, participants choose a shape (triangle for cooperation, square for defection) within 4 s and see both players’ choices for 1.5 s. (**d, e**) Post hoc comparisons: d and e show the participants’ cooperation probability on the y-axis. The x-axis represents the consistency of the partner’s actions in previous trials ( $t_{- 1}$ : last trial, $t_{- 1, 2}$ : last two trials, $t_{- 1, 2, 3}$ : last three trials). Large red (adolescents) and blue (adults) dots indicate mean probabilities, with black error bars for standard error (SE). Gray dots represent mean probabilities across trials, and green error bars show predicted cooperation rates with SE. Notes: *n.s*.p>0.05; *p<0.05; **p<0.01; ***p<0.001.

Figure 2

Download asset Open asset

Computational modeling.

(**a, b**) Model comparisons for adolescents and adults, respectively. The y-axis represents model fitness based on the Akaike Information Criterion with a correction for sample size (AICc; Hurvich and Tsai, 1989). For each participant, the model with the lowest AICc served as a reference to compute ΔAICc by subtracting it from the AICc of other models ( $Δ AICc = {AICc}_{x} - {AICc}_{lowest}$ ). A lower ΔAICc indicates a better model fit. Protected exceedance probability (PEP) is a group-level measure that assesses the likelihood of each model’s superiority over the others (Rigoux et al., 2014). (c) Model recovery analysis. Each model was used to generate 100 synthetic datasets, and for each dataset, model fitting and comparison were performed. Each column corresponds to one generative model, and each row corresponds to one fitting model. The color in each cell indicates the probability that the synthetic datasets generated by the model in the column were best fit by the model in the row, with a darker color denoting a higher probability. (**d, e**) Model prediction. Sample illustration of the best-fitting model prediction versus data for adolescents and adults, respectively.

Figure 3

Download asset Open asset

Learning rates and social preferences.

(**a–d**) Comparison between adolescents and adults for positive learning rate ( $α +$ ), negative learning rate ( $α -$ ), social preference ( $ω$ ), and inverse temperature (β), respectively. (**e–h**) Correlation between age and positive learning rate, negative learning rate, social preference, and inverse temperature, respectively. Notes: *p<0.05; **p<0.01.

Figure 4

Download asset Open asset

Analysis of hidden variables from the best-fitting model.

(**a, b**) Post-hoc comparison of LMM1: interaction of group × previous trial × partner’s choice. The y-axis shows participants’ expectations of partner cooperation probability ( $p$ ) from the best-fitting model. (**c, d**) Self-reported cooperativeness: normalized scores on partner cooperativeness for two orders of partner cooperation probability, with adolescents (orange-red line) and adults (blue line). Scores were assessed on a 0–9 scale and normalized to 0–1. The dotted line indicates the presumed partner’s cooperation probability, with mean values and standard errors shown. (**e, f**) Post-hoc comparison of LMM3: interaction of group × previous trial × partner’s choice. The y-axis shows participants’ intrinsic reward for reciprocity ( $p \times ω$ ) from the best-fitting model. The x-axis represents the consistency of the partner’s actions in previous trials ( $t_{- 1}$ : last trial, $t_{- 1, 2}$ : last two trials, $t_{- 1, 2, 3}$ : last three trials). Colored dots with error bars indicate mean values with standard errors for adolescents (orange-red) and adults (blue), while small gray dots represent individual participants. Notes: *n.s*.p>0.05; *p<0.05; **p<0.01; ***p<0.001.

Appendix 1—figure 1

Download asset Open asset

Model prediction.

This figure compares the empirical cooperation probabilities and the model-predicted values for adults (a) and adolescents (b). The x-axis represents the trial number, and the y-axis represents the mean cooperation probability across participants. The shaded areas indicate the 95% confidence intervals.

Appendix 1—figure 2

Download asset Open asset

Distributions of estimated parameters from the best-fitting model.

Each panel displays one parameter. The histograms and their kernel fits are represented by color bars and curves, respectively. Red indicates participants in the adolescent sample, and blue denotes those in the adult sample. Parameters have been transformed into a log scale for enhanced visualization.

Appendix 1—figure 3

Download asset Open asset

Parameter recovery for the best-fitting model.

Each panel represents one parameter. Each dot corresponds to one virtual participant. The value of r indicates Pearson’s correlation coefficient between the true values (estimated from the participants) and the recovered parameters.

Appendix 1—figure 4

Download asset Open asset

Partial correlation matrices among parameters for the best-fitting model.

The upper-triangular cells show partial correlations for adults, and the lower-triangular cells show partial correlations for adolescents. Each cell shows the partial Pearson correlation coefficient (controlling for the other parameters). Colors range from green (negative) to violet (positive), with the color bar spanning [–1,1]. Notes: *n.s.* p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001.

Appendix 1—figure 5

Download asset Open asset

Group-level posterior distributions from the hierarchical Bayesian estimation for adolescents and adults.

Posterior densities are shown separately for adolescents (red) and adults (blue). $Δ$ values indicate the posterior mean difference (Adult – Adolescent) with 95% credible intervals (CrI) and Bayesian p values. Compared with adolescents, adults exhibited higher positive learning rates ( $α +$ ) and lower negative learning rates ( $α -$ ), suggesting greater differentiation between learning from positive and negative feedback. Adults also showed lower inverse temperature ( $β$ ), indicating more exploratory decision behavior, and higher social reward weight ( $ω$ ), reflecting greater valuation of reciprocity or social outcomes. Notes: *n.s.* p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001.

Appendix 1—figure 6

Download asset Open asset

Model comparison results for (a) adults and (b) adolescents, including the newly added M9 (Social Reward and Pearce–Hall learning).

Lower ΔAICc values indicate better model fits. The dynamic learning rate model (Model 9: Social Reward model with dynamic RL algorithm) did not outperform the best-fitting model (Model 8) in either group.

Appendix 1—figure 7

Download asset Open asset

Convergence diagnostics for the hierarchical Bayesian model.

(a) Distribution of $\hat{R}$ (Rhat) values across all model parameters. The majority of $\hat{R}$ values are below the conservative convergence threshold of 1.01 (red dashed line), indicating stable and well-mixed MCMC chains. The gray shaded area highlights the region where $\hat{R} \leq 1.01$ . (b) Trace plots for the group-level parameters (four chains) in adolescents (left, red box) and adults (right, blue box). Each line represents the sampled posterior values of one chain across iterations, with overlapping traces and stable fluctuations confirming adequate convergence and mixing for all key parameters ( $α +$ , $α -$ , $ω$ , $β$ ).

Author response image 1

Download asset Open asset

Linear and quadratic model fits showing the relationship between age and the ω parameter, with 95% confidence intervals.

Author response image 2

Download asset Open asset

Simulation results showing how variations in each model parameter affect the group difference in mean cooperation probability (Adults – Adolescents).

Based on the bestfitting Model 8 and parameters estimated from all participants, each line represents one parameter (i.e., α+, α-, ω, β) systematically varied within the tested range (α±:0.1–0.9; ω, β:1–9) while other parameters were held constant. Positive values indicate higher cooperation in adults. Smaller ω values most strongly reproduced the observed group difference, suggesting that reduced social reward weighting primarily drives adolescents’ lower cooperation.

Tables

Appendix 1—table 1

Statistical results for cooperation decision (GLMM1).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	0.47	0.22	2.10	p=0.036
Timing	–0.001	0.001	–1.01	p=0.311
Gender	0.14	0.15	0.93	p=0.355
Group	0.79	0.24	3.23	p=0.001
Previous trial	–1.81	0.08	–23.43	p<0.001
Partner’s choice	–0.73	0.10	–7.21	p<0.001
Group × Previous trial	–0.41	0.10	–3.89	p<0.001
Group × Partner’s choice	–0.19	0.11	–1.77	p=0.076
Previous trial × Partner’s choice	1.12	0.04	25.43	p<0.001
Group × Previous trial × Partner’s choice	0.24	0.06	4.05	p<0.001

Coding of variables. Trial number: integer sequence from 2 to 120; gender: female = 0, male = 1; group: adolescents = 0, adults = 1; previous trials: last one trial = 1, last two trials = 2, last three trials = 3; partner’s choice: cooperation = 1, defection = 0.

Appendix 1—table 2

Statistical results for partner cooperation expectation (LMM1).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	0.59	0.03	17.52	p<0.001
Trial number	<0.001	<0.001	0.02	p=0.982
Gender	–0.01	0.03	–0.36	p=0.718
Group	–0.04	0.03	–1.29	p=0.198
Previous trial	–0.25	0.01	–48.97	p<0.001
Partner’s choice	–0.06	0.01	–4.90	p<0.001
Group × Previous trial	–0.04	0.01	–6.10	p<0.001
Group × Partner’s choice	–0.02	0.01	–2.73	p=0.006
Previous trial × Partner’s choice	0.16	0.002	53.89	p<0.001
Group × Previous trial × Partner’s choice	0.03	0.004	7.31	p<0.001

Coding of variables is consistent with Appendix 1—table 1.

Appendix 1—table 3

Statistical results for self-reported perceived partner cooperativeness (LMM2).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/SubjectiveRatingScore.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	5.82	0.27	21.93	p<0.001
Group	0.17	0.35	0.50	p=0.616
Gender	–0.44	0.11	–3.84	p<0.001
Order of sessions	0.09	0.15	0.60	p=0.549
Rating number	–0.13	0.02	–5.55	p<0.001
Group × Order	0.38	0.22	1.72	p=0.085

Coding of variables. Group: adolescents = 0, adults = 1; gender: female = 0, male = 1; order of sessions: stable to volatile = 0, volatile to stable = 1; rating number: integer sequence from 1 to 8.

Appendix 1—table 4

Statistical results for intrinsic reward for reciprocity (LMM3).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	2.43	0.14	17.04	p<0.001
Trial number	–0.001	0.001	–0.60	p=0.551
Gender	0.08	0.13	0.59	p=0.554
Group	0.52	0.15	3.44	p<0.001
Previous trial	–1.23	0.03	–36.18	p<0.001
Partner’s choice	–0.27	0.08	–3.24	p=0.001
Group × Previous trial	–0.57	0.05	–11.95	p<0.001
Group × Partner’s choice	–0.24	0.05	–4.43	p=0.006
Previous trial × Partner’s choice	0.79	0.02	40.64	p<0.001
Group × Previous trial × Partner’s choice	0.37	0.02	13.70	p<0.001

Coding of variables is consistent with Appendix 1—table 1.

Appendix 1—table 5

Statistical results for cooperation decision with age (GLMM_sup1).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	–1.55	0.73	–2.12	p=0.034
Trial number	–0.001	0.001	–0.94	p=0.346
Gender	–0.06	0.18	–0.34	p=0.732
Previous trial	–0.49	0.15	–3.30	p=0.001
Partner’s choice	–0.60	0.33	–1.83	p=0.068
Age	0.10	0.04	2.49	p=0.013
Previous trial × Partner’s choice	0.89	0.17	5.10	p=<0.001
Previous trial × Age	–0.02	0.01	–2.00	p=0.045
Partner’s choice × Age	–0.01	0.02	–0.68	p=0.499
Previous trial × Partner’s choice × Age	0.02	0.01	2.07	p=0.038

Age was treated as a continuous variable, and all other variables were coded as in Appendix 1—table 1.

Appendix 1—table 6

Statistical results for partner cooperation expectation with age (LMM_sup1).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	0.76	0.12	6.52	p<0.001
Trial number	$- 2.53 \times 10^{- 7}$	0.0002	–0.002	p=0.999
Gender	–0.01	0.03	–0.32	p=0.746
Previous trial	–0.10	0.01	–10.51	p<0.001
Partner’s choice	–0.03	0.03	–1.05	p=0.294
Age	–0.014	0.006	–2.27	p=0.023
Previous trial × Partner’s choice	0.15	0.01	13.24	p<0.001
Previous trial × Age	$- 4.76 \times 10^{- 7}$	0.0004	0.001	p=0.999
Partner’s choice × Age	–0.002	0.001	–1.74	p=0.082
Previous trial × Partner’s choice × Age	0.0012	0.0006	1.96	p=0.050

Coding of variables is consistent with Appendix 1—table 5.

Appendix 1—table 7

Statistical results for self-reported perceived partner cooperativeness with age (LMM_sup2).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/SubjectiveRatingScore.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	5.18	1.89	2.74	p=0.006
Gender	–0.37	0.19	–1.96	p=0.050
Order of sessions	2.31	1.21	1.90	p=0.057
Age	0.03	0.10	0.27	p=0.784
Rating number	–0.13	0.02	–6.65	p<0.001
Order × Age	–0.10	0.07	–1.57	p=0.116

Age was treated as a continuous variable, and all other variables were coded as in Appendix 1—table 3.

Appendix 1—table 8

Statistical results for partner cooperation expectation with age (LMM_sup3).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	2.02	0.58	3.49	p<0.001
Trial number	–0.001	0.001	–0.60	p=0.546
Gender	–0.07	0.13	–0.50	p=0.619
Previous trial	–0.22	0.06	–3.59	p<0.001
Partner’s choice	–0.11	0.18	–0.62	p=0.533
Age	0.02	0.03	0.61	p=0.540
Previous trial × Partner’s choice	0.25	0.08	3.28	p=0.001
Previous trial × Age	–0.02	0.003	–5.24	p<0.001
Partner’s choice × Age	–0.02	0.01	–1.85	p=0.065
Previous trial × Partner’s choice × Age	0.04	0.004	9.82	p<0.001

Coding of variables is consistent with Appendix 1—table 5.

Appendix 1—table 9

Statistical results for cooperation decision with phase (GLMM_sup2).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	1.06	0.24	4.34	p<0.001
Trial number	–0.01	0.001	–4.46	p<0.001
Gender	0.17	0.15	1.10	p=0.271
Group	0.60	0.19	3.15	p=0.002
Previous trial	–0.64	0.04	–16.49	p<0.001
Partner’s choice	–0.75	0.10	–7.73	p<0.001
Phase	–0.66	0.10	–6.45	p<0.001
Group × Previous trial	–0.16	0.05	–3.15	p=0.002
Group × Partner’s choice	–0.19	0.11	–1.77	p=0.077
Previous trial × Partner’s choice	1.04	0.04	23.44	p<0.001
Group × Previous trial × Partner’s choice	0.25	0.06	4.07	p<0.001

Phase was dummy-coded (0=stable, 1=changing), and all other variables were coded as in Appendix 1—table 1.

Appendix 1—table 10

Statistical results for partner cooperation expectation with phase (LMM_sup4).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	0.77	0.03	28.09	p<0.001
Trial number	$- 5.86 \times 10^{- 6}$	$9.25 \times 10^{- 5}$	–0.06	p=0.949
Gender	–0.02	0.03	–0.96	p=0.337
Group	–0.06	0.03	–2.36	p=0.018
Previous trial	–0.08	0.002	–33.62	p<0.001
Partner’s choice	–0.07	0.01	–6.65	p<0.001
Phase	–0.14	0.01	–19.16	p<0.001
Group × Previous trial	–0.013	0.003	–4.02	p<0.001
Group × Partner’s choice	–0.025	0.007	–3.11	p=0.002
Previous trial × Partner’s choice	0.15	0.003	50.67	p<0.001
Group × Previous trial × Partner’s choice	0.03	0.004	7.40	p<0.001

Coding of variables is consistent with Appendix 1—table 9.

Appendix 1—table 11

Statistical results for intrinsic reward for reciprocity with phase (LMM_sup5).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	3.57	0.14	25.57	p<0.001
Trial number	–0.0011	0.0006	–2.01	p=0.045
Gender	0.02	0.13	0.18	p=0.860
Group	0.25	0.13	1.96	p=0.050
Previous trial	–0.38	0.02	–22.73	p<0.001
Partner’s choice	–0.33	0.07	–4.59	p<0.001
Phase	–0.83	0.06	–14.76	p<0.001
Group × Previous trial	–0.19	0.02	–8.79	p<0.001
Group × Partner’s choice	–0.27	0.05	–5.18	p<0.001
Previous trial × Partner’s choice	0.70	0.02	37.41	p<0.001
Group × Previous trial × Partner’s choice	0.37	0.03	13.99	p<0.001

Coding of variables is consistent with Appendix 1—table 9.

Appendix 1—table 12

Statistical results for cooperation decision with SVO (GLMM_sup3).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	–0.63	0.22	–2.88	p=0.004
Trial number	–0.002	0.001	–1.16	p=0.247
Gender	0.18	0.14	1.27	p=0.202
Group	0.56	0.18	3.11	p=0.002
Previous trial	–0.71	0.04	–18.17	p<0.001
Partner’s choice	–0.77	0.10	–7.41	p<0.001
SVO	0.02	0.005	3.35	p<0.001
Group ×Previous trial	–0.14	0.05	–2.84	p=0.005
Group ×Partner’s choice ×Age	–0.16	0.11	–1.44	p=0.150
Previous trial ×Partner’s choice	1.13	0.04	25.18	p<0.001
Group ×Previous trial ×Partner’s choice	0.25	0.06	4.06	p<0.001

SVO was treated as a continuous variable, and all other variables were coded as in Appendix 1—table 1.

Appendix 1—table 13

Statistical results for intrinsic reward for reciprocity with SVO (LMM_sup6).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	2.03	0.15	13.73	p<0.001
Trial number	–0.001	0.00 1	–0.71	p=0.479
Gender	0.02	0.12	0.18	p=0.855
Group	0.26	0.13	2.07	p=0.039
Previous trial	–0.45	0.02	–25.75	p<0.001
Partner’s choice	–0.31	0.08	–3.65	p<0.001
Phase	0.006	0.004	1.49	p=0.137
Group × Previous trial	–0.18	0.02	–8.02	p<0.001
Group × Partner’s choice	–0.17	0.06	–3.03	p=0.002
Previous trial × Partner’s choice	0.80	0.02	40.26	p<0.001
Group × Previous trial × Partner’s choice	0.37	0.03	13.38	p<0.001

Coding of variables is consistent with Appendix 1—table 12.

Appendix 1—table 14

Statistical results for cooperation decision predicted by cooperation expectation (GLMM_sup4).

Source data available at https://raw.githubusercontent.com/xiaoyanwu2024/Adolescents_SelfInterest_Cooperation/main/data/dataset.csv.

Fixed effects	Estimated beta value	SE	t value	p value
(Intercept)	–4.57	0.35	–12.92	p<0.001
Trial number	–0.002	0.001	–2.25	p=0.024
Gender	0.20	0.28	0.71	p=0.475
Group	1.37	0.34	4.06	p<0.001
Cooperation expectation	7.90	0.33	24.05	p<0.001
Group × Cooperation expectation	0.01	0.19	0.08	p=0.938

Coding of variables. Trial number: integer sequence from 2 to 120; gender: female = 0, male = 1; group: adolescents = 0, adults = 1. Cooperation expectation was treated as a continuous variable.