Adaptation is essential for survival. Across species, it occurs over many generations through evolution and natural selection. Individual animals, however, can also adapt within their own lifetimes – through learning. For many species, humans included, this process is vital.
The challenge arises when adaptation must take place in a hazardous environment, such as when hunting dangerous prey. “In such cases, learning through exploration can pose a serious threat to a juvenile’s survival,” explains Dharanish Rajendra. He is a doctoral researcher working with Professor Chaitanya Gokhale, Chair of Computational and Theoretical Biology at Julius-Maximilians-Universität Würzburg (JMU).
One solution is parental care: adults create protected learning environments in which offspring can gradually approach the challenges of real life.
In a new publication, the two researchers use mathematical and computational models to investigate how such learning environments can be structured to promote successful adaptation.
Step by step towards competence
Many predators specialise in prey that represent a substantial risk for inexperienced hunters. Wolves take on elk and bison many times their size, pythons must subdue porcupines, and meerkats frequently prey on venomous scorpions.
These small mammals, native to southern Africa, live in colonies of up to 45 individuals and are well known for their complex social behaviour. To prepare their young for handling dangerous prey, adults proceed in stages: first offering dead scorpions, then scorpions with the sting removed, and only later presenting fully intact, dangerous prey once the juveniles have developed sufficient skill.
When protection goes too far
In their study, the researchers modelled this developmental transition using a two-phase learning framework, simulating the shift from a protected juvenile stage to an unprotected adult environment.
Their findings highlight a risk reminiscent of so-called “helicopter parenting”: if the learning environment is too safe, or differs too greatly from real conditions, maladaptation can occur. Individuals may reach adulthood insufficiently prepared and struggle to cope with genuine risks.
“What matters most for successful protected learning is that the environment remains sufficiently similar to reality,” says Chaitanya Gokhale. “A gradual increase in risk bridges the gap between a safe developmental space and the demands of the real world.”
The mathematics behind behaviour
To analyse these behavioural strategies, the researchers combined two complementary modelling approaches.
Dynamic programming was used to calculate the theoretically optimal strategy under different environmental conditions. Reinforcement learning was then employed to simulate how individuals acquire such strategies through trial and error.
Together, these methods show how an animal learns to weigh risks against rewards and how early-life experience shapes later performance.
“Our research provides a theoretical foundation for parental care strategies that have long been observed in nature, but whose underlying mathematical logic has remained unclear,” says Dharanish Rajendra. The findings contribute more broadly to our understanding of early-life learning in both animals and humans, as well as species-specific differences in parental care.
Because many species that provide extended parental care are also highly social, the researchers see promising directions for future work. In particular, they aim to investigate how protected developmental environments interact with social learning – learning through observing and drawing on the experiences of others.