Researchers at Emory University and the Georgia Institute of Technology have developed a portable system designed to automate cognitive testing of wild primates. The proof-of-concept platform, called CapuchinAI, combines machine vision, touchscreen-based experiments, video recording, and an automated food dispenser.
The research, published in the American Journal of Primatology, demonstrates how researchers could bring standardized cognitive experiments into natural environments while reducing the need to administer each trial manually.
Bringing controlled experiments into the field
Laboratory studies give researchers control over experimental conditions, but those settings cannot fully reproduce the social and ecological environments in which primate cognition developed. Field research offers greater ecological relevance, but changing weather, animal behavior, and other uncontrolled conditions can make it difficult to administer consistent experiments and collect comparable data.
“The primate brain didn’t evolve in a lab, it evolved in complex, competitive environments,” said Marcela Benítez, an assistant professor of anthropology at Emory and senior author of the study.
The researchers designed CapuchinAI to combine the consistency of laboratory testing with the ecological context of field observation. The compact system includes a webcam, touchscreen, Raspberry Pi computer, motorized food dispenser, and battery pack inside a weather-resistant wooden enclosure.
When the system detects an approaching capuchin, it begins recording, displays a touchscreen stimulus, and dispenses a dried banana reward when the animal completes the task. The battery-powered platform can operate for approximately eight hours before requiring a recharge.
Testing the platform with wild capuchins
The researchers conducted a two-week field trial with two groups of habituated white-faced capuchins at Costa Rica’s Taboga Forest Reserve. Sixteen animals voluntarily interacted with the platform, 10 learned to activate the reward dispenser, and eight formed and retained an association between touching the screen and receiving food.
The team also developed a machine learning model trained to distinguish six individual capuchins using still images, recorded video, and live field footage. The model achieved more than 97 percent precision and recall.
However, the researchers did not have enough high-quality training footage to build an identification model for all approximately 100 capuchins at the field site. During the pilot, the platform therefore detected whether a capuchin was present without consistently identifying each individual. Video collected by the system could help the team expand the recognition model.
Building toward individualized cognitive studies
The researchers plan to use the platform for individualized tests involving learning, impulse control, cognitive flexibility, and short- and long-term memory. A more developed machine vision system could eventually identify each approaching animal, retrieve its previous results, and select a task based on its testing history.
That capability could allow researchers to collect repeated cognitive measurements from individual animals without capturing them or removing them from their social groups. It could also support longer studies examining how cognition changes with age, social status, environmental conditions, or other factors.
The researchers have made computer code and construction guidance available to support adaptations for other field sites and primate species. They emphasized, however, that automated testing should complement rather than replace long-term observation. Researchers still need detailed knowledge of individual animals, social groups, and environmental conditions to interpret the data produced by the platform.
Advanced Lab Management Certificate
The Advanced Lab Management certificate is more than training—it’s a professional advantage.
Gain critical skills and IACET-approved CEUs that make a measurable difference.
Implications for lab management and field operations
Beyond its scientific applications, the system also has implications for how primate research teams organize fieldwork and manage experimental workflows. For lab managers coordinating field sites, a platform like CapuchinAI could reduce the need for continuous on-site staffing during data collection periods, shifting effort toward setup, maintenance, and periodic monitoring rather than constant manual trial administration.
The system’s automated design also introduces a more standardized experimental pipeline, which can improve reproducibility across field sites and reduce variability introduced by different human operators. At the same time, it requires new forms of logistical planning, including equipment calibration schedules, battery and hardware maintenance, software updates, and reliable data storage and transfer systems in remote environments.
For lab operations, the platform could streamline training by reducing the need for extensive behavioral testing expertise in the field, while increasing the importance of technical skills related to device upkeep and troubleshooting. Managers would also need to balance the efficiency gains of automation with oversight responsibilities to ensure data quality, system reliability, and ethical compliance in animal interactions.
The researchers have made computer code and construction guidance available to support adaptations for other field sites and primate species. They emphasized, however, that automated testing should complement rather than replace long-term observation. Researchers still need detailed knowledge of individual animals, social groups, and environmental conditions to interpret the data produced by the platform.
This article was created with the assistance of Generative AI and has undergone editorial review before publishing.
