Temporal dynamics and policy implications in an innovation arms race

Human-modified environments offer novel resources, but their exploitation can be a source of human-wildlife conflict. Residents of Sydney have reported increasing cases of bin-opening behaviour by sulphur-crested cockatoos (Cacatua galerita), with evidence that this behaviour is socially learnt between birds. Households protected their bins, yet cockatoos have learnt to defeat these defences. In response, residents increased their defence level, setting the stage for a behavioural 'arms race'. Here, we investigate this arms race by combining field observations with agent-based modelling. We systematically document protections in a suburban locality over 2 years, revealing spatial clustering of protections indicative of social learning among residents. We find that protections have decreased since 2019, and overly costly protection is dispreferred. With a controlled assay, we characterize the knowledge of local cockatoos, finding differential proficiency in defeating protections. Surprisingly, we identified several cockatoos that can defeat high-efficacy protections, such as locks. Finally, we simulate interactions between two populations of learning agents representing households and cockatoos. We find that social learning accelerates adaptation in both species, while coordination reduces costs associated with defensive escalation. However, policy interventions can have unintended consequences, accelerating cockatoos' skill acquisition and shifting conflict to neighbouring areas. Our findings highlight the dynamic nature of human-wildlife conflict and the importance of considering behavioural feedback loops in urban wildlife management. This article is part of the theme issue 'Transforming cultural evolution research and its application to global futures'.