Wild capuchin monkeys in a Costa Rican forest reserve have learned to use an artificial intelligence touchscreen system to earn banana rewards, offering researchers a new scalable method to study primate cognition in natural environments, according to a study published in The American Journal of Primatology.
CapuchinAI System Architecture and Field Deployment
Developed by researchers at Emory University and the Georgia Institute of Technology, the system is dubbed CapuchinAI. According to the study, the platform integrates a compact, battery-powered computing system into a field research setup to automate cognitive tests in the wild. The equipment uses facial recognition to identify an approaching monkey, presents a customized learning task on a touchscreen, and dispenses a food reward if the animal performs correctly. “The primate brain didn’t evolve in a lab, it evolved in complex, competitive environments,” said Marcela Benítez, Emory assistant professor of anthropology and senior author of the paper. “Yet primate cognition is rarely studied in the wild because the experimental control needed to measure cognition is difficult in unpredictable environments.” Field tests in the Taboga Forest Reserve of Costa Rica demonstrated that the system identifies individual capuchins with 97% accuracy after training on still images and videos.
Bridging Lab Research and Wild Habitats
The project builds on the legacy of Frans de Waal, who pioneered animal cognition studies as director of Emory’s Living Links Centre for Advanced Study of Ape and Human Evolution before dying in 2024, according to Federico Sánchez Vargas, first author of the paper and an Emory Ph.D. student of anthropology. “Our AI method allows us to more deeply understand individuals that we already have data on through field observation,” Sánchez Vargas said. Benítez studies social behaviors in monkeys through a partnership called Capuchinos de Taboga, launched in 2017 in collaboration with the Universidad Nacional Técnica of Costa Rica. To build the technology, Benítez collaborated with Jacob Abernethy, an associate professor of computer science at Georgia Tech, and Sai Rakshith Potluri, a former Georgia Tech graduate research assistant now working as a software engineer at ExtraHop in Seattle. The team utilized open-source software called YOLO (You Only Look Once) to run the model on a laptop.
Building and Testing the Hardware in Costa Rica
Undergraduates at Emory and Georgia Tech manually placed bounding boxes around monkey faces in thousands of images and videos to train the facial recognition system. Sánchez Vargas later adapted the model to recognize any capuchin rather than distinct individuals to trigger a webcam, which rapidly expanded the dataset of face images. Encased in a wooden box built from Home Depot materials and featuring a 3D-printed food dispenser created by Emory TechLab, the apparatus ran for eight hours on a lightweight battery pack. In the forest reserve, a large male capuchin named Trompudo was the first to approach the box, slapping the touchscreen and receiving a banana slice. “It was almost like you could see him realising, ‘Ah, that’s what you have to do, touch the screen!'” Sánchez Vargas said. Observations during the pilot phase showed that 16 individual capuchins interacted with the device differently, with some investigating via their lips to kiss the screen for food.
Did you know? The CapuchinAI system is programmed to limit the number of food rewards an individual monkey can receive during a single session, which stops dominant capuchins from monopolizing the testing platform.
Future Cognitive Testing and Research Goals
The research team is updating its facial-recognition model using recorded videos of the 16 capuchins from the pilot phase to prepare for broader cognitive experiments. According to the study, upcoming tests will target four domains: learning ability, impulse control, cognitive flexibility, and short- and long-term memory. Machine vision allows the setup to switch tests between identified monkeys automatically. “We can explore outstanding questions about how the environment, individual experiences and behaviours connect to cognitive abilities,” Benítez said. The project was launched with a seed grant from Emory’s AI. Humanities program.
Frequently Asked Questions
What is CapuchinAI?
CapuchinAI is a field-research system that uses facial recognition and touchscreens to automate cognitive testing for wild capuchin monkeys, delivering food rewards for correct tasks.
Who developed the CapuchinAI system?
The system was developed by researchers at Emory University and the Georgia Institute of Technology, including Marcela Benítez, Federico Sánchez Vargas, Jacob Abernethy, and Sai Rakshith Potluri.
How accurate is the facial recognition technology?
Field tests in the Taboga Forest Reserve of Costa Rica found that the system identifies individual capuchins with 97% accuracy.
How do the monkeys interact with the touchscreen?
Wild capuchins learn to touch the screen to receive banana slices, with some individuals investigating the box using their lips to kiss the screen for a reward.
Want to stay updated on the latest breakthroughs in animal behavior and technology? Subscribe to our newsletter or explore more research news on our website.
Related reading