Experimental investigation of domestic chicks' spatial navigation strategies

The purpose of this research was to assess the strategies used by baby chickens to navigate an environment while foraging, to confront such strategies with machine learning (ML) models, and finally to apply the data to an AI environment, which mimics a real ecosystem. Specifically we wanted to determine whether chickens' decisions followed the Nearest Neighbour and Optimal strategies. And to analyse to what extent animals' strategies matched the decisions made by the ML model. Four-five day-old domestic chicks (N=35) were placed, one by one, in three versions of a maze. These were randomly generated configurations of 7x7 (71.4cm x 71.4cm) squares, each square measured 10.2cm x 10.2cm. The dependent variables consisted in the timing and the order in which the chicks pecked and ate each mealworm in the maze. The independent variables were the 3 maze configurations, and the order in which these were experienced by each chick. The experiment took place over a period of 3 months, each week the order of mazes was altered. We found that the chicks exhibited decision-making behaviour that aligned with machine learning models. Specifically, the chickens predominantly followed the Nearest Neighbour strategy in 68% of their decisions and the Optimal strategy in 62%. This result suggests that chicks’ behaviour could be simulated by AI models.