Ready Player One. Valerian. Worlds in chaos that need human video gamers to save their worlds via virtual reality. Similarly, high-tech gadgets utilizing facial recognition or voice activation tools to access superpower suits or weapons are the secret to saving the world in Iron Man or The Incredibles. Even our smart phones today utilize various virtual reality, facial recognition, or voice activation tools to access functions or apps. And automation has also found an entry into behavioral research methodology.
The Flintstone Era
Automating functions is a process that continues to evolve. From Thorndike’s puzzle boxes and Kohler’s insight tasks with bananas, sticks, and boxes to Skinner’s operant conditioning chamber to the world of psychological research into cognitive phenomena continues to strive to remove human influence from experimental studies.
Meet the Jetsons
Today, hand-held laser beams can track the movement of free-ranging animals. Other animals, such as primates, can be trained to use virtual-reality tools to experience certain contexts. Technological advances often occur because running animal subjects is both time-intensive and costly, and any way to reduce training or testing time while also limiting experimenter bias is welcomed.
Many cognitive studies take years of training and testing, especially when working with nonhuman animals. In most cases, nonhuman subjects have to be trained and tested separately, which means less efficient training and testing experiences and potentially, more stress if the subject has to be removed from its home.
With today’s restrictions on nonhuman primate research, maximizing research time is particularly important. In the United States, invasive research with chimpanzees and bonobos has been banned (since 2013) and in the European Union, all great apes (including chimpanzees, bonobos, gorillas, and orangutans) were excluded from invasive research in 2010.
Moreover, a number of other policies and restrictions currently exist for conducting research with nonhuman primates, which limits the types of research that can be performed. David DeGrazia wrote a recent commentary on this issue.
Ready, Player One?
A recent advancement in a wireless touchscreen system that incorporates a face recognition tool provides a non-invasive alternative to automating cognitive training in nonhuman primates.
Created by James Butler and Steve Kennerley of University College London, the Mymou system (Greek for monkey, pronounced “my – moo”) is a low-cost, open source, fully automated wireless touchscreen system that uses facial recognition software, a small tablet, and the animal’s home.
The focus of this article, Butler and Kennerley published their work in a recent issue of the Psychonomic Society’s journal Behavior Research Methods. Although several other automated training systems already exist for non-human primate cognitive training, most of these systems are expensive, require invasive implants, or need subjects to be removed from their home cage. The system developed by Butler and Kennerley addressed all of these issues.
The Nuts and Bolts
As shown in the figure above, the system is relatively simple, especially for those with some electrical skills and programming knowledge. The system is fully wireless and uses Bluetooth to communicate with the system that delivers the reward (i.e., juices of different flavors). In the figure, A represents the sequence of information flow, B refers to the 3D schematic of the holder that was used to secure the tablet to the outside of the cage.
As seen in the image, the entire system includes 1) a tablet, 2) metal locking pin, 3) hole for optional charging cable, 4) tablet holder, 5) sloped funnel, 6) front panel, 7) hole to lock the holder to the cage, 8) reward tubes, 9) hooks to mount holder to cage. As can be seen in the image, C is a photo of the holder depicted in B. (Numbers correspond to the same items in B.) Finally, the fourth image, D, is a photo of the microcontroller (1) connected to a 4-channel peristaltic pump (2).
The really great part of this system is that it can be utilized in the home cage of the animals, and the animal chooses when to use the system. Where previous systems had to be turned on and off at specific times, or animals had to be transported to a training laboratory, Mymou can be programmed to be available or not. Because juice is used as a reward with limited amounts provided, the animals do not become satiated too quickly. The role of choice is critical in the Mymou system, which really enhances welfare of the subjects and meets the requirements of conducting research with non-human primates.
Validation is Sweet
Butler and Kennerey tested their apparatus and the training process with two different cohorts. The first test of the apparatus consisted of two male rhesus monkeys, Subject O and Subject V. The system was available up to 12 hours a day in their home cage. The monkeys were given some training experience with the touchscreen so that they could learn how it worked.
Following the training, the monkeys were tested for 10 months for 182 sessions using 36 increasingly complex cognitive tasks. One task involved learning how certain images were connected to each other so that the monkeys reach a specific goal stimulus. First, the monkeys had to learn which stimuli were related to each other (e. g., movie ticket is linked to orange, but not to key). After the different associations were learned across a grid of 16 images, the monkeys then needed to be able to plan their path to achieve the goal without making an error.
As you can see in the figure below, the monkeys had access to the touchscreen (Panel A) and the juice spouts (Panel B), which were activated when the monkey completed a trial correctly (Panel C).
As Panel D in the figure illustrates, the monkeys were both fairly quick to start using the touchscreen and complete trials. Individual differences in their use and overall performance were observed. But, as shown in Panel E, both monkeys were very engaged in playing the games, completing 100s to 1000s of trials across the 182 sessions.
The system relies on its facial recognition software to accurately detect which animal is currently interacting with the tablet (see example below). Butler and Kennerey indicated that the system was 99% accurate.
Both monkeys showed learning and preferences for rewards. The flexibility of the system was especially helpful in sustaining motivation especially as the tasks became more complex. Ultimately, this automated system enabled the monkeys to learn very difficult and cognitively demanding tasks at their own pace, while learning multiple tasks simultaneously.
As an affordable system with limited technical requirements, the Mymou system seems to be the next advancement in nonhuman primate cognition research. Perhaps in the next few years, they too will be ready for avatars and altered dimensions in which they are tested on their gaming abilities to win the coveted egg that saves the world.
Psychonomics article focused on in this post:
Butler, J. L., & Kennerley, S. W. (2018). Mymou: A low-cost, wireless touchscreen system for automated training of nonhuman primates. Behavior Research Methods. DOI: 10.3758/s13428-018-1109-5.