Researchers from Emory University and Georgia Institute of Technology have developed CapuchinAI, an automated facial-recognition touchscreen system used in Costa Rica’s Taboga Forest Reserve. The system allows scientists to conduct scalable cognitive testing on wild white-faced capuchin monkeys by delivering food rewards for correct task performance.
Traditional primate research often forces a choice between the sterile control of a laboratory and the unpredictable chaos of the wild. To bridge this gap, scientists have deployed a battery-powered computing platform that uses AI to identify individual monkeys and present them with tailored learning tasks. According to Indiatimes, capuchin monkeys interacted with the device for banana rewards, with some individuals even learning to kiss the screen.
The CapuchinAI Facial Recognition System
The technical backbone of the project relies on YOLO (You Only Look Once), an open-source software used to build a model capable of running on a laptop. To train the AI, researchers used high-quality GoPro videos of six wild capuchins in the Taboga Forest Reserve. This required a labor-intensive process where undergraduate students from Emory and Georgia Tech placed bounding boxes
around monkey faces in thousands of images and videos to tag their identities.
This training produced a system with high precision. According to Emory, the AI can identify specific monkeys—including individuals named Pele, Pedro, Pio, and Pesto—with 97 percent accuracy from live footage, video, and static images.
Once the system identifies an approaching monkey, it triggers a closed-loop pipeline: the touchscreen presents a task specific to that individual, and the system automatically delivers food if the task is completed correctly.
Bridging the Gap Between Lab and Field
The project’s primary goal is to move cognitive measurement out of artificial settings. While labs offer tight control, researchers note that results can be skewed because animals aren’t interacting with their natural environments.
Benítez, a co-director of the Capuchinos de Taboga project launched in 2017, argues that studying individuals in the wild allows scientists to see how vast variations in life experience shape cognition. By automating the testing, the team can now quantify findings and get into the minds behind the personalities
of monkeys they have already been observing through field data.
Legacy of the Living Links Center
The development of CapuchinAI is framed as a continuation of the work started by Frans de Waal, the former director of Emory’s Living Links Center for the Advanced Study of Ape and Human Evolution, who died in 2024. De Waal was known for treating primates as individuals rather than just subjects of behavioral experiments.
By combining de Waal’s focus on individual primate “portraits” with modern AI, the researchers aim to create a scalable method for monitoring cognitive abilities across different species and environments. To encourage this, the team has released an open-source paper that includes a blueprint for building the low-cost field platform and the associated computer coding.
Environmental Pressures on Capuchin Cognition
The need for such flexible testing tools is highlighted by the volatility of the Costa Rican landscape. Separate research into white-faced capuchins in the tropical dry forests shows that extreme climate events can fundamentally restructure how these monkeys interact. For instance, during severe droughts caused by El Niño between 2014 and 2016, dwindling food sources led capuchins to abandon their usual societal norms.
These ecological stresses create a high-stakes environment where group size becomes a critical factor for survival. While capuchins are generalist feeders, extreme climate events can cause larger groups to go hungry while smaller groups fare better. This suggests that the cognitive and social adaptations the AI is designed to measure are not static, but are constantly reacting to environmental pressures.
