CapuchinAI Brings Automated Cognitive Testing to Wild Monkeys in Their Natural Habitat

CapuchinAI Transforms Wild Primate Cognition Research

Researchers from Emory University and Georgia Institute of Technology have developed CapuchinAI, an open-source AI system that automates cognitive testing of wild monkeys in their natural habitat1

1

. Published in the American Journal of Primatology, this proof-of-concept method provides the first scalable, systematic way to evaluate and monitor the cognitive abilities of wild primates without removing them from their environment2

2

. The battery-powered platform integrates facial recognition, touchscreen tasks, and automated food rewards into a compact field-research system running on a single Raspberry Pi1

1

.

YOLO-Based Computer Vision Achieves 97% Accuracy

The system uses YOLO-based computer vision trained on thousands of GoPro images and videos of white-faced capuchins at Costa Rica's Taboga Forest Reserve3

3

. Emory University and Georgia Tech undergraduates performed labor-intensive digital tagging, placing bounding boxes around monkey faces in thousands of frames2

2

. The facial recognition system identifies individual capuchins with 97% accuracy from static images, video, and live footage in the field1

1

. When tested on previously unseen images, the model detected capuchin faces 98% of the time and correctly identified specific individuals 97% of the time3

3

.

Building the Low-Cost Field Platform

Federico Sánchez Vargas, an Emory PhD student of anthropology and first author of the paper, built the housing in Marcela Benítez's garage using pine planks, deck sealant, rubber insulating strips, and plastic piping from Home Depot3

3

. The finished box stands about 20 inches tall, weighs roughly 35 pounds, and runs for at least eight hours on ordinary travel battery packs3

3

. Inside, a Raspberry Pi computer controls a webcam, touchscreen, and 3D-printed food dispenser4

4

. Sánchez Vargas taught himself Python to adapt the recognition model, stripping it down to flag any capuchin face rather than just the six individuals in the training dataset3

3

.

Wild Capuchins Rapidly Learn Touchscreen Tasks

Field tests revealed that wild capuchins rapidly habituated to the platform and learned touchscreen-reward associations spontaneously without human intervention1

1

. After waiting three days, a large male named Trompudo became the first to interact with the box, slapping the touchscreen and receiving a banana slice3

3

. Over six sessions with two habituated groups, 16 capuchins interacted with the box while 14 more watched from a distance3

3

. Of those 16, ten learned to trigger rewards by touching the screen, eight built a full habit of expecting banana, and five still remembered the association a week after the box was removed3

3

. Learning styles varied significantly—some monkeys caught on immediately while others watched groupmates first, and some even learned to press their mouths to the screen instead of using their hands3

3

.

Personalized Cognitive Tasks Across Four Domains

CapuchinAI delivers personalized cognitive tasks across four main domains: learning speed, impulse control, cognitive flexibility, and working or long-term memory1

1

. The automated software tracks individual participation limits per session, preventing dominant group members from monopolizing the testing box and ensuring balanced data collection across the troop1

1

. The touchscreen only dispenses rewards when its camera confirms a capuchin face is present—when coatis, tree-climbing relatives of raccoons, tried to access the system, they received nothing and lost interest3

3

. The team is now training an upgraded model to recognize all 16 pilot monkeys individually, with plans to test enhanced capabilities for impulse control and cognitive flexibility this summer4

4

.

Bridging Lab Precision with Natural Environments

Marcela Benítez, Emory assistant professor of anthropology and senior author, explained the fundamental challenge: "The primate brain didn't evolve in a lab, it evolved in complex, competitive environments. Yet primate cognition is rarely studied in the wild because the experimental control needed to measure cognition is difficult in unpredictable environments"2

2

. AI-powered cognitive testing now enables researchers to conduct automated cognitive studies with lab-grade precision while preserving the ecological validity of studying wild primates in their natural habitat1

1

. Benítez studies cooperation and social behaviors in both captive tufted capuchins in laboratories and wild white-faced capuchins at Taboga, serving as co-director of Capuchinos de Taboga, a research project launched in 20172

2

.

Building on Frans de Waal's Legacy

Sánchez Vargas noted the project builds on the legacy of Frans de Waal, who pioneered studies of animal cognition as director of Emory's Living Links Center for the Advanced Study of Ape and Human Evolution before passing away in 20242

2

. "Our AI method allows us to more deeply understand individuals that we already have data on through field observation. We can now automate cognitive testing of them and quantify the findings. It's a way of getting into the minds behind the personalities," Sánchez Vargas explained1

1

. Co-authors include Jacob Abernethy, Georgia Tech associate professor of computer science, and Sai Rakshith Potluri, a former Georgia Tech graduate research assistant now at ExtraHop in Seattle2

2

. The open-source paper includes computer coding guides and blueprints for building the low-cost platform, with researchers hoping other scientists will adapt the method to generate cognitive data spanning different species of wild primates across diverse environments2

2

.