There is something about the late afternoon light in April - that peculiar golden hour when the day can't quite decide whether to linger - that reminds one of adolescence itself: caught between two states, committed to neither. It is during these in-between hours that the brain does some of its most interesting developmental work, and a new study has just revealed one of its more inconvenient findings about the teenage mind.

Here's the scene: you do something nice for a teenager. They notice. They understand perfectly well that you've been kind. And then... they pocket the goodwill and look out for number one anyway. Not because they're oblivious to your generosity, but because, computationally speaking, their brains simply don't reward reciprocity the way adult brains do.

The Game That Proved What Every Parent Suspected

Researchers at Beijing Normal University and elsewhere sat down 127 adolescents (aged 14-17) and 134 adults (18-30) and had them play a repeated Prisoner's Dilemma - that classic game theory setup where two players each choose to cooperate or defect, and the best collective outcome requires mutual trust (Wu et al., 2026). Over 120 rounds, participants played against a computer partner whose cooperation rate shifted predictably.

The results were beautifully specific. Adolescents didn't just cooperate less across the board - they cooperated less specifically after their partner had been cooperative. When the partner defected, teens and adults responded similarly (both groups had the good sense to stop being generous to someone who'd just stiffed them). But when the partner extended an olive branch repeatedly, adults ramped up their cooperation in kind. Teenagers? They noticed the olive branch, appreciated the olive branch, and then used it to reach the higher-scoring defection option.

It's Not That They Can't Read the Room

This is where the study takes a genuinely surprising turn. The prevailing wisdom in developmental psychology has been that adolescents cooperate less because their mentalizing abilities - the capacity to model what another person is thinking - are still under construction. The prefrontal cortex, after all, doesn't finish its renovation project until the mid-twenties (Arain et al., 2013).

But the computational modelling told a different story. Using reinforcement learning algorithms, the researchers tracked two things: how well participants predicted their partner's next move, and how much internal reward they derived from reciprocating cooperation. Adolescents were just as accurate as adults at reading their partner's cooperative intentions. Their mental models were perfectly functional, thank you very much.

The difference lay entirely in the reward signal. Adults experienced a growing intrinsic payoff for matching cooperation with cooperation - a kind of neural warm glow that increased the more their partner played nice. In adolescents, that warm glow was essentially a pilot light. Functional, but not exactly heating the room.

The Reward System Is Running a Different Programme

This fits neatly with what neuroscience has been telling us about the adolescent brain for the past two decades. The reward system matures on a faster timeline than the cognitive control regions, creating what researchers call a "dual systems imbalance" (Casey et al., 2008). Recent computational work has shown that adolescents process social information using simpler reinforcement learning strategies, while adults integrate increasingly sophisticated features like counterfactual reasoning - the ability to consider what would have happened if they'd chosen differently (Bolenz et al., 2017).

What Wu and colleagues have added is a crucial piece of the puzzle: it's not that teenagers are bad at social learning. They're quite competent at it. They simply don't find reciprocity as rewarding. The selfish option carries relatively more weight in their internal calculations - not because of some moral failing, but because the neural architecture that makes cooperation feel good is still being wired up.

So What Do We Do With This?

The practical implications are rather pointed. If the problem were poor mentalizing, the solution would be teaching teenagers to better understand others' intentions - more perspective-taking exercises, more empathy training. But since adolescents already understand their partner's cooperative intentions perfectly well, the intervention needs to target something else entirely: making reciprocity feel more rewarding.

This could mean designing social environments where the benefits of cooperation are immediate, tangible, and personally meaningful to teenagers - rather than relying on the abstract satisfaction of having been a decent person. Because, as it turns out, your teenager's brain has done the maths on being decent, and it's currently getting a better rate of return elsewhere.

One suspects evolution will sort it out by their mid-twenties. In the meantime, perhaps take some comfort in knowing that when your teenager fails to reciprocate your kindness, it's nothing personal. Their reward circuitry simply hasn't received the software update yet.

References

1. Wu, X., Fu, H., Aydogan, G., Feng, C., Qin, S., Zeng, Y., & Liu, C. (2026). The self-interest of adolescents overrules cooperation in social dilemmas. eLife, 12, e106840. https://doi.org/10.7554/eLife.106840 | PubMed

2. Arain, M., Haque, M., Johal, L., Mathur, P., Nel, W., Rais, A., Sandhu, R., & Sharma, S. (2013). Maturation of the adolescent brain. Neuropsychiatric Disease and Treatment, 9, 449-461. https://doi.org/10.2147/NDT.S39776 | PMCID: PMC3621648

3. Casey, B. J., Jones, R. M., & Hare, T. A. (2008). The adolescent brain. Annals of the New York Academy of Sciences, 1124(1), 111-126. https://doi.org/10.1196/annals.1440.010 | PMCID: PMC2475802

4. Bolenz, F., Reiter, A. M. F., & Eppinger, B. (2017). Developmental changes in learning: Computational mechanisms and social influences. Developmental Cognitive Neuroscience, 25, 50-59. https://doi.org/10.1016/j.dcn.2017.06.009

5. Zhang, L., & Bhatt, M. A. (2024). Learning to adaptively cooperate through social interactions during childhood and adolescence. npj Science of Learning, 11, 12. https://doi.org/10.1038/s41539-026-00423-9

Disclaimer: The image accompanying this article is for illustrative purposes only and does not depict actual experimental results, data, or biological mechanisms.