Optimising play for learning risky behaviour
Rajendra, D.; Gokhale, C. S.
Show abstract
Animals adapt their behaviour to current environmental conditions to enhance survival and reproductive success. While longterm adaptation occurs through evolutionary processes acting on heritable variation, individuals can also adapt within their lifetime via learning. Learning is particularly advantageous in environments that are uncertain or fluctuate across a lifespan or a few generations. However, reliance on individual learning entails a critical risk. Juveniles may begin life poorly adapted to their surroundings, requiring exploration to learn. Such an approach can be costly and dangerous, especially for species engaging in risky activities such as hunting dangerous prey. We explore how early-life learning in a protected environment, such as one buffered by parental care, can facilitate effective behavioural adaptation in later, riskier contexts. As a representative case, we model the decision-making process of a predator hunting both safe and dangerous prey. We analyse decisionmaking dynamics through reinforcement learning, extending beyond classical dynamic programming approaches. Our results show that experiences in a juveniles early environment can generalise to a distinct adult environment, provided there is sufficient structural similarity between them. Our findings demonstrate that incorporating structured play or safe exploration in early life can significantly enhance the performance of learningbased adaptation in dangerous environments.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Social inhibition maintains adaptivity and consensus of foraging honeybee swarms in dynamic environments 96%
- Adaptation to DNA damage as a bet-hedging mechanism in a fluctuating environment 96%
- Cost and social distancing dynamics in a mathematical model of COVID-19 with application to Ontario, Canada 95%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.