In the Taboga Forest Reserve in Costa Rica, researchers have introduced a portable testing station equipped with artificial intelligence to study cognitive abilities in wild white-faced capuchin monkeys. The device—a 15-inch touchscreen mounted within a wooden frame—dispenses small rewards, such as dried banana slices, when the monkeys interact with it correctly.
The initiative, detailed in a recent paper published in the American Journal of Primatology, represents an innovative approach to evaluating primate cognition outside laboratory settings. The station utilizes AI-powered facial recognition to detect the presence of capuchins in real time and administer simple learning tasks. This technology aims to provide insights into how cognitive skills relate to survival and decision-making in natural environments, areas difficult to assess through conventional lab studies.
Emory University primatologist Marcela Benitez, who has studied the Taboga capuchins for years, spearheaded the project. Her team trained an artificial intelligence model to identify capuchins based on images collected during previous fieldwork. They then developed a portable apparatus with a touchscreen, webcam, and 3D-printed food dispensers to administer associative learning tests.
During the initial deployment last summer, the researchers encountered some challenges. The first two days yielded no monkey interactions, raising concerns among the team. However, on the third day, an alpha male named Papi began engaging with the device, quickly learning that touching the screen resulted in receiving a food reward. Over multiple sessions, a total of 16 capuchins interacted with the station, with 10 demonstrating they understood the task and at least eight retaining the learned behavior.
Not all capuchins adapted equally; some failed to connect the touchscreen action with the reward, instead exhibiting behaviors like biting or slapping the device without engaging the screen. These variations underscore the complex nature of cognitive testing in wild populations.
Looking ahead, the researchers plan to enhance the system with a more advanced facial recognition model capable of distinguishing individual monkeys with over 97 percent accuracy. Future iterations will introduce a broader range of cognitive tests, including impulse control tasks—rewarding monkeys for touching certain colored squares while ignoring others—and cognitive flexibility assessments where the task rules change dynamically. The system is designed to monitor individual progress across sessions, allowing researchers to explore how cognitive abilities vary among monkeys and how these differences relate to environmental challenges and survival.
Benitez noted the potential to study connections between early-life adversity, such as severe droughts, and cognitive development, as well as how memory skills influence foraging success. The ultimate goal is to link cognition with real-world outcomes and evolutionary fitness.
Despite the promise of the technology, the researchers emphasized that they do not intend to permanently place such devices in the forest, cautioning against disrupting the natural habitat. Presently, the testing stations require human supervision and are not fully autonomous, though the team envisions further advancements over the next decade.