The Future of Artificial Intelligence in Monitoring Animal Identification, Health, and Behaviour.

With many advancements, technologies are now capable of recording non-human animals' location, heart rate, and movement, often using a device that is physically attached to the monitored animals. However, to our knowledge, there is currently no technology that is able to do this unobtrusively and non-invasively. Here, we review the history of technology for use with animals, recent technological advancements, current limitations, and a brief introduction to our proposed novel software. Canadian tech mogul EAIGLE Inc. has developed an artificial intelligence (AI) software solution capable of determining where people and assets are within public places or attractions for operational intelligence, security, and health and safety applications. The solution also monitors individual temperatures to reduce the potential spread of COVID-19. This technology has been adapted for use at the Toronto Zoo, initiated with a focus on Sumatran orangutans (Pongo abelii) given the close physical similarity between orangutans and humans as great ape species. This technology will be capable of mass data collection, individual identification, pose estimation, behaviour monitoring and tracking orangutans' locations, in real time on a 24/7 basis, benefitting both zookeepers and researchers looking to review this information.

Preference for free or forced choice in Sumatran orangutans (Pongo abelii).

Empirical investigations of humans, pigeons, rats, and monkeys have indicated that these species will select free over forced choice, even when faced with identical outcomes. However, the same has yet to be quantitatively confirmed in nonhuman great apes. This experiment is the first systematic investigation of preference for free or forced choice in great apes using a paradigm in which extraneous variables are highly controlled. Three orangutans were given a choice of one of two virtual routes, one that provided a choice and one that did not via a touchscreen computer program. Choice of either route was rewarded with the same type and quantity of food. Initial results indicated a preference for free choice across all three participants. However, in two control conditions, orangutans' preferences varied, suggesting a weaker tendency to exercise choice than species previously tested. We suggest further investigation of preference for free and forced choice in orangutans and other great apes through alternative experimental paradigms that focus on increasing the fidelity of free and forced choice options.

Borrowing from Peter to pay Paul: managing threatened predators of endangered and declining prey species.

Conservation policy and practice can sometimes run counter to their mutual aims of ensuring species survival. In Kenya, where threatened predators such as lion deplete endangered prey such as Grevy's zebra, conservation practitioners seek to ensure species success through exclusive strategies of protection, population increase and preservation. We found strong selection for the endangered Grevy's zebra by both lion and hyena on two small fenced conservancies in Kenya. Despite abundant diversity of available prey, Grevy's zebra were selected disproportionately more than their availability, while other highly available species such as buffalo were avoided. Lions were therefore not alone in presenting a credible threat to Grevy's zebra survival. Conservation practitioners must consider interlinked characteristics of prey selection, resource availability and quality, the interplay between carnivore guild members and landscape scale population trends performance in wildlife management decisions.

A ghost fence-gap: surprising wildlife usage of an obsolete fence crossing.

Wildlife fencing has become more prevalent throughout Africa, although it has come with a price of increased habitat fragmentation and loss of habitat connectivity. In an effort to increase connectivity, managers of fenced conservancies can place strategic gaps along the fences to allow wildlife access to outside habitat, permitting exploration, dispersal and seasonal migration. Wildlife can become accustomed to certain movement pathways and can show fidelity to these routes over many years, even at the path level. Our study site has three dedicated wildlife crossings (fence-gaps) in its 142 km perimeter fence, and we continuously monitor these fence-gaps with camera-traps. We monitored one fence-gap before and after a 1.49 km fence section was completely removed and 6.8 km was reconfigured to leave only a two-strand electric fence meant to exclude elephant and giraffe, all other species being able to cross under the exclusionary fence. The removal and reconfiguration of the fence effectively rendered this fence-gap (which was left in place structurally) as a "ghost" fence-gap, as wildlife now had many options along the 8.29 km shared border to cross into the neighboring habitat. Although we documented some decline in the number of crossing events at the ghost-gap, surprisingly, 19 months after the total removal of the fence, we continued to document the usage of this crossing location by wildlife including by species that had not been previously detected at this location. We discuss potential drivers of this persistent and counterintuitive behavior as well as management implications.

Colonization and usage of an artificial urban wetland complex by freshwater turtles.

Conservation authorities invest heavily in the restoration and/or creation of wetlands to counteract the destruction of habitat caused by urbanization. Monitoring the colonization of these new wetlands is critical to an adaptive management process. We conducted a turtle mark-recapture survey in a 250 ha artificially created wetland complex in a large North American city (Toronto, Ontario). We found that two of Ontario's eight native turtle species (Snapping turtle (SN), Chelydra serpentina, and Midland Painted (MP) turtle, Chrysemys picta marginata) were abundant and both were confirmed nesting. The Blanding's turtle (Emydoidea blandingii) was present but not well established. Species richness and turtle density were not equally distributed throughout the wetland complex. We noted SN almost exclusively populated one water body, while other areas of the wetland had a varying representation of both species. The sex ratios of both SN and MP turtles were 1:1. We tracked the movement of Snapping and Blanding's turtles and found that most turtles explored at least two water bodies in the park, that females explored more water bodies than males, and that 95% of turtles showed fidelity to individual overwintering wetlands. We performed DNA analysis of two Blanding's turtles found in the created wetlands and could not assign these turtles to any known profiled populations. The genetic data suggest that the turtles probably belong to a remnant local population. We discuss the implications of our results for connectivity of artificial wetlands and the importance of the whole wetland complex to this turtle assemblage.

Testing the effects of perimeter fencing and elephant exclosures on lion predation patterns in a Kenyan wildlife conservancy.

The use of fences to segregate wildlife can change predator and prey behaviour. Predators can learn to incorporate fencing into their hunting strategies and prey can learn to avoid foraging near fences. A twelve-strand electric predator-proof fence surrounds our study site. There are also porous one-strand electric fences used to create exclosures where elephant (and giraffe) cannot enter in order to protect blocs of browse vegetation for two critically endangered species, the black rhinoceros (Diceros bicornis) and the Grevy's zebra (Equus grevyi). The denser vegetation in these exclosures attracts both browsing prey and ambush predators. In this study we examined if lion predation patterns differed near the perimeter fencing and inside the elephant exclosures by mapping the location of kills. We used a spatial analysis to compare the predation patterns near the perimeter fencing and inside the exclosures to predation in the rest of the conservancy. Predation was not over-represented near the perimeter fence but the pattern of predation near the fence suggests that fences may be a contributing factor to predation success. Overall, we found that predation was over-represented inside and within 50 m of the exclosures. However, by examining individual exclosures in greater detail using a hot spot analysis, we found that only a few exclosures contained lion predation hot spots. Although some exclosures provide good hunting grounds for lions, we concluded that exclosures did not necessarily create prey-traps per se and that managers could continue to use this type of exclusionary fencing to protect stands of dense vegetation.

Testing the Prey-Trap Hypothesis at Two Wildlife Conservancies in Kenya.

Protecting an endangered and highly poached species can conflict with providing an open and ecologically connected landscape for coexisting species. In Kenya, about half of the black rhino (Diceros bicornis) live in electrically fenced private conservancies. Purpose-built fence-gaps permit some landscape connectivity for elephant while restricting rhino from escaping. We monitored the usage patterns at these gaps by motion-triggered cameras and found high traffic volumes and predictable patterns of prey movement. The prey-trap hypothesis (PTH) proposes that predators exploit this predictable prey movement. We tested the PTH at two semi-porous reserves using two different methods: a spatial analysis and a temporal analysis. Using spatial analysis, we mapped the location of predation events with GPS and looked for concentration of kill sites near the gaps as well as conducting clustering and hot spot analysis to determine areas of statistically significant predation clustering. Using temporal analysis, we examined the time lapse between the passage of prey and predator and searched for evidence of active prey seeking and/or predator avoidance. We found no support for the PTH and conclude that the design of the fence-gaps is well suited to promoting connectivity in these types of conservancies.

Individual effects of seasonal changes, visitor density, and concurrent bear behavior on stereotypical behaviors in captive polar bears (Ursus maritimus).

Stereotypical behaviors in captive polar bears (Ursus maritimus) can be detrimental to their welfare. These behaviors can be reduced through enrichment programs but are often not completely eliminated, so identifying potential triggers is important. The present study investigated the influences of seasonal changes, visitor density, and concurrent bear activity on stereotypical behaviors exhibited by 3 captive polar bears at the Toronto Zoo. All bears exhibited these behaviors; however, individual differences were found in duration and form. The male exhibited less stereotypical behavior during spring, and the females exhibited less stereotypical behavior during winter. An increase in visitor density was associated with more stereotypical behavior in 1 female but less stereotypical behavior in the other 2 bears. All bears engaged in more stereotypical behaviors when the other bears were inactive, and 1 female engaged in more stereotypical behaviors when the other bears were out of sight. Further, when conspecifics were active, all bears engaged in less stereotypical behaviors. Given the variability among individual bears, future enrichment programs must be tailored to the needs of individuals to maximize efficacy.

Anxiety-related behavior of orphan chimpanzees (Pan troglodytes schweinfurthii) at Gombe National Park, Tanzania.

This study examined the anxiety levels and social interactions of two orphan and four mother-reared adolescent chimpanzees (Pan troglodytes schweinfurthii) in the Kasekela community at Gombe National Park, Tanzania. We used focal sampling in the field at Gombe to observe these adolescent individuals. Their social interactions and anxious behavior, measured as rough scratching, were recorded. The two orphans differed from others of a similar age by exhibiting higher levels of anxiety and lower levels of play. These results suggest that a mother's absence, even in naturalistic conditions in which other members of the community are available to the orphan, may have long-lasting impact on an adolescent's anxiety and its ability to engage in complex social interactions, such as play.

Feature versus spatial strategies by orangutans (Pongo abelii) and human children (Homo sapiens) in a cross-dimensional task.

Despite the fact that photographic stimuli are used across experimental contexts with both human and nonhuman subjects, the nature of individuals' perceptions of these stimuli is still not well understood. In the present experiments, we tested whether three orangutans and 36 human children could use photographic information presented on a computer screen to solve a perceptually corresponding problem in the physical domain. Furthermore, we tested the cues that aided in this process by pitting featural information against spatial position in a series of probe trials. We found that many of the children and one orangutan were successfully able to use the information cross-dimensionally; however, the other two orangutans and almost a quarter of the children failed to acquire the task. Species differences emerged with respect to ease of task acquisition. More striking, however, were the differences in cues that participants used to solve the task: Whereas the orangutan used a spatial strategy, the majority of children used a feature one. Possible reasons for these differences are discussed from both evolutionary and developmental perspectives. The novel results found here underscore the need for further testing in this area to design appropriate experimental paradigms in future comparative research settings.

Information seeking by orangutans: a generalized search strategy?

Recent empirical work has suggested that some species of non-human primates may be aware of their knowledge states. One finding to support this claim is that they seek information about the location of a hidden food item when they are unsure of its location, but not when they already know where it is, which purportedly demonstrates metacognition. However, this behaviour may instead reflect a generalized search strategy, in which subjects reach for food when they see it, and search for it when they do not. In this experiment, this possibility was addressed by testing orangutans in three conditions in which the location of a food item was sometimes known to subjects, and other times required subjects to visually seek the missing information. All subjects exhibited behaviour consistent with a metacognitive interpretation in at least two of the three conditions. Critically, in two of the conditions, subjects refrained from seeking visual information, and correctly found the hidden food item without ever seeing it, using inference by exclusion. The results suggest that animals that succeed in this information-seeking task are not merely acting according to a generalized search strategy, and instead seek information adaptively according to their knowledge states.

Orangutans (Pongo abelii) "play the odds": information-seeking strategies in relation to cost, risk, and benefit.

Recent research has examined whether animals possess metacognition, or the ability to monitor their knowledge states. However, the extent to which animals actively control their knowledge states is still not well delineated. Although organisms might be capable of seeking information when it is lacking, it does not mean that it is always adaptive to do so. In the present set of experiments, we examined the flexibility of this behavior in captive orangutans (Pongo abelii; two adults and one juvenile) in a foraging task, by varying the necessity of information-seeking, the cost associated with it, the likelihood of error, and the value of the reward. In Experiment 1, subjects searched for information most often when it was "cheapest" energetically. In Experiment 2, subjects searched for information most often when the odds of making an error were the greatest. In Experiment 3, subjects searched for information more when the reward was doubled in value. In Experiment 4, adult subjects adapted to risk/benefit trade-offs in their searching behavior. In every experiment, subjects sought information more often when they needed it than when they already knew the solution to the problem. Therefore, the current research suggests that information-seeking behavior in orangutans shows a sophisticated level of flexibility, comparable to that seen in human children, as they appear to "play the odds" when making the decision to seek information or not.

Strategies in landmark use by orangutans and human children.

Landmark use has been demonstrated in a variety of organisms, yet the manner in which landmarks are encoded and subsequently used appears to vary between and sometimes within species, even when faced with identical landmark arrays. In the present experiments, orangutans and human children were shown a square array of identical landmarks and were trained to locate a hidden goal in the centre of the array. In Experiments 1 and 2, the search space appeared to be discrete, with white gridlines dividing up the space, and in Experiments 3a and 3b, the search space was uniformly coloured, making it appear continuous. In all experiments, following training, subjects were given a single expansion test, to determine their landmark strategy use, based on peak search activity. The orangutans appeared to use absolute directional vectors from individual landmarks, with peak search activities on the inner corners of the square array, and they used this strategy persistently. In contrast, human children showed two landmark-based strategies, absolute directional vectors and a relational or "middle" strategy, with the majority of children starting their search in the middle region. Although some children, especially young children, persistently used one strategy like the orangutans, many changed strategies when the original one failed to yield the hidden goal.

The use of perceptual features in categorization by orangutans (Pongo abelli).

The extent to which categorization of natural classes in animals reflects a generalization based on perceptual similarity versus an abstract conceptual representation remains unclear. Here, two experiments were conducted to identify the perceptual features used by orangutans when categorizing pictures. In Experiment 1, subjects were trained and tested for transfer on a concrete discrimination (gorillas or orangutans vs. other primates). Analysis of performance on both positive and negative exemplars revealed that performance was best on photos with faces, particularly close-ups. Moreover, error trials did not seem to reflect instances of mistaken identity, but rather, exemplars that may have been distracting for other reasons, such as novel coloration or morphology. In Experiment 2, photos were modified to test the effects of various features. Color modifications caused the biggest decrease in performance, and eye modifications also affected performance deleteriously. Therefore, two perceptual features, namely eye regions and color, played a key role in subjects' ability to categorize. However, performance based on an underlying concept cannot be ruled out, because both of these features are highly relevant in terms of defining category membership. Although a subset of features was identified as playing a key role in categorization, these features differed depending on whether feature-use was studied indirectly, as in Experiment 1, or directly, as in Experiment 2.

Levels of abstraction in orangutan (Pongo abelii) categorization.

Levels of abstraction have rarely been manipulated in studies of natural concept formation in nonhumans. Isolated examples have indicated that animals, relative to humans, may learn concepts at varying levels of abstraction with differential ease. The ability of 6 orangutans (Pongo abelii) of various ages to make natural concept discriminations at 3 levels of abstraction was therefore investigated. The orangutans were rewarded for selecting photos of orangutans instead of humans and other primates (concrete level), primates instead of other animals (intermediate level), and animals instead of nonanimals (abstract level) in a 2-choice touch screen procedure. The results suggest that, like a gorilla (Gorilla gorilla gorilla) tested previously (Vonk & MacDonald, 2002), orangutans can learn concepts at each level of abstraction, and unlike other nonhumans, most of these subjects rapidly learned the intermediate level discrimination.

Natural concepts in a juvenile gorilla (gorilla gorilla gorilla) at three levels of abstraction.

The extent to which nonhumans are able to form conceptual versus perceptual discriminations remains a matter of debate. Among the great apes, only chimpanzees have been tested for conceptual understanding, defined as the ability to form discriminations not based solely on simple perceptual features of stimuli, and to transfer this learning to novel stimuli. In the present investigation, a young captive female gorilla was trained at three levels of abstraction (concrete, intermediate, and abstract) involving sets of photographs representing natural categories (e.g., orangutans vs. humans, primates vs. nonprimate animals, animals vs. foods). Within each level of abstraction, when the gorilla had learned to discriminate positive from negative exemplars in one set of photographs, a novel set was introduced. Transfer was defined in terms of high accuracy during the first two sessions with the new stimuli. The gorilla acquired discriminations at all three levels of abstraction but showed unambiguous transfer only with the concrete and abstract stimulus sets. Detailed analyses of response patterns revealed little evidence of control by simple stimulus features. Acquisition and transfer involving abstract stimulus sets suggest a conceptual basis for gorilla categorization. The gorilla's relatively poor performance with intermediate-level discriminations parallels findings with pigeons, and suggests a need to reconsider the role of perceptual information in discriminations thought to indicate conceptual behavior in nonhumans.

Spatial memory and foraging competition in captive western lowland gorillas (Gorilla gorilla gorilla).

Spatial memory and foraging competition were investigated in three mother/offspring pairs of western lowland gorillas,Gorilla gorilla gorilla, using a naturalistic foraging task at the Toronto Zoo. Sixteen permanent food sites were placed throughout the animals' enclosures. All of the sites were baited and a pair of animals was free to visit the sites and collect the food. Five of the subjects collected the food with accuracy better than chance. Most of the subjects visited the sites using a pattern, and for half the subjects this was one of adjacency. The high accuracy of five of the subjects and the lack of a consistent adjacency pattern suggest that the animals did in fact use spatial memory. Furthermore, the gorillas tended to avoid visiting food sites that had been previously depleted by their partner. They also appeared to split their search of the enclosures, each visiting only a proportion of the food sites. This indicated that the animals were competing and altering their foraging behaviour based on the behaviour of their partner. Therefore, accuracy was recalculated to take this into account. When the sites depleted by either animal in a pair during a given trial were worked into the accuracy calculations for individual animals, three of the animals still maintained accuracy above chance. This suggests that the animals were not only able to remember which sites they had depleted, but those sites depleted by their foraging partner as well.