Purpose-bred laboratory animals are underrepresented in the animal-computer interaction (ACI) literature. What the ACI community has published on research with such animals suggests that there are no examples of work utilizing animal-centered approaches. Globally, hundreds of millions of rats and mice are kept for scientific research, representing an overwhelming majority of vertebrate laboratory animals. Standard caging for rats prevents them from exhibiting some natural behaviors, including standing on their hind legs, gnawing to wear down their incisors, and living in social groups. Nonetheless, there is a significant body of laboratory animal welfare research that has been developed since the 1940s that is focused on the effects of enriched housing conditions on rats. An investigatory review of laboratory animal welfare literature reveals, a subset of recent scholarship (n=32) in this area takes interest in transitioning toward an animal-centered ethical framework. That work is happening independently of and parallel to the ACI community's development of its own animal-centered ethics. In this literature review of laboratory animal science journals, half of the articles on rat (and/or mouse) welfare already used technologies that could be leveraged for ACI research. Slightly less than half (47%) of articles were considered to be ethically incompatible with ACI, at least partially. However, only 3 (9%) papers were identified representing work that were wholly incompatible. Of the remaining 53% of articles that were compatible with ACI, 4 (13%) were identified as functionally equivalent to ACI's ethical framework, i.e., animal-centered. This finding stands in opposition to prior ACI work that suggested there were likely no examples of animal-centered research with laboratory animals. Clear opportunities for researchers in the ACI community to collaborate with laboratory animal scientists on research projects improving the well-being of rats and mice, and ultimately to all lab animals.
Continuous monitoring is essential for managing feline diabetes, however, it remains a challenge for both pet owners and cats, often delaying timely care. Current veterinary practices rely on infrequent urine samples, clinical observations, or invasive subdermal glucose monitors, while water intake is typically tracked manually by owners. We present DiabetiCAT, a non-invasive, in-home system that combines a biosensing litter box and a smart water fountain to monitor two key health indicators for diabetic cats: urinary glucose and hydration. The litter box integrates electrochemical biosensors to measure urinary glucose via chronoamperometry, while a scale under the water bowl is used to track hydration. A potentiostat captures the biosensor readings which are transmitted wirelessly to a remote database for storage and analysis. A veterinary-facing app interface paired with veterinary reports was designed in consultation with veterinary professionals. We evaluated the system with a 40-hour simulation using glucose solutions, tested real feline urine samples, and assessed the scale's capacity and precision. In a two-week case study with a diabetic cat, the collected data guided iterative refinements of our interface. Interviews with five veterinarians revealed how continuous, non-invasive biosensing data can support collaborative care among pets, owners, and professionals. DiabetiCAT advances the emerging subfield of Animal Biosensing Computing, demonstrating how IoT-integrated systems can extend the visibility of health-related behaviors and biochemistry in companion animals.
Cat cafes are environments where visitor experiences centre on human-animal interactions. At the same time, the animals' side of the experience has led to widespread ethics concerns, connected with the overstimulation potential of direct, intense animal-human interaction, with concomitant effects on animal welfare. Investigating how technology could bring balance to this relationship, the paper presents MewTube, a tablet-based application designed to engage cats through curated videos while allowing humans to adjust playback speed in response to feline behaviour. From a five-month study with 28 human and 16 feline participants that compared tablet-based play at a cat cafe with traditional toys' use there, analysis revealed that overall enjoyment stayed constant between the two conditions yet toy-focused interactions led to a stronger sense of connection with the cats. Reflecting on our findings - e.g., in the absence of clear responsiveness from the cats, humans in the MewTube condition felt less confident that the animals enjoyed the interaction - reveals gaps in technology for supporting interaction in human-animal relationships. The paper offers recommendations for tackling the challenges of designing systems that respect feline autonomy and sensory preferences.
This paper is intended to help ACI researchers better understand the zoo as a context for our work, especially anyone interested in providing zoo animals with more choice and control. Based on 14 semi-structured interviews with zoo industry experts, we found that working with zoos is a process, influenced by factors such as changes in public perception about what it means to be a healthy, captive animal, advances in animal welfare-related research, the evolution of stakeholder priorities, and how well multiple departments work together. The growing popularity of providing choice and control to zoo animals has powerfully impacted the zoo industry as promoting animals' natural abilities, agency, and autonomy has become a higher priority. Providing choice and control to zoo animals, however, requires practitioners to focus on matters beyond our specific design interventions, especially the challenges of working with and across a variety of stakeholders.
Animal activity recognition (AAR) using wearable inertial measurement units is widely used to monitor dog posture and behavior in real-world care and training. To provide reliable feedback for veterinarians, trainers, and pet owners, preprocessing must be dependable and consistent because model performance is sensitive to it. However, published pipelines vary greatly and often lack important preprocessing details, making it hard to compare and reproduce results. This study systematically investigates how two preprocessing choices, posture-specific Interquartile Range outlier handling and Butterworth low-pass filtering, affect dog activity classification. Using a controlled experimental design and two publicly available datasets, we tested how posture-specific IQR outlier handling and Butterworth low-pass filtering, evaluated at various cutoff frequencies, influence dog activity classification across multiple algorithms. These findings offer practical guidance for model-aware preprocessing and clear reporting, allowing for reproducible and user-centered deployments of dog activity recognition (DAR).
Animal-Computer Interaction (ACI) research has steadily progressed enrichment methods and engagement technologies for animals, particularly for parrot species. Building on this trajectory, this study explores a low-effort, welfare-conscious, and multimodal approach to engage a captive salmon-crested cockatoo. Leveraging insights from prior studies on parrot-specific interactive technologies, multimodal enrichment, and animal-IoT applications, we implemented a prototype avian-IoT system that emphasizes voluntary engagement, minimal disruption, and remote monitoring. The prototyped system integrates a mobile application, a central server, and interaction devices to coordinate stimuli, manage data, and enable remote observation with low manual maintenance and training efforts. In a pilot experiment conducted at Kyoto City Zoo, the responsive multimodal stimuli effectively engaged the cockatoo, with auditory cues proving especially salient in encouraging interaction and reducing vigilance behaviors. The results demonstrate the feasibility of IoT-mediated multimodal interaction that can balance animal agency with human participation, offering valuable insights for designing animal-centered IoT systems and inclusive cross-species interaction frameworks.
Therapy that incorporates equine movement is a valuable tool for strength and rehabilitation training. While several commercial products exist to monitor the saddle pressure and fit on a horse, therapists utilizing hippotherapy currently do not have an objective means for assessing the progress of their clients' control of weight distribution while engaging in therapy. In this work, we present a device for monitoring the leaning patterns and weight distribution of clients participating in therapy using equine movement. Our prototype uses commercial off-the-shelf parts including barometers that are enclosed in custom-made inflated bladders placed inside a bareback pad to detect changes in air pressure as the bladders are deformed by the weight of the patient. Currently, we are able to show that a client's lean could be detected with an accuracy range of 91.4-98.7%, precision of 48.2%, and an average recall of 42.9% using a simple thresholding algorithm based on rolling sums of pressure changes. Ongoing work seeks to explore methods of designing bladders and processing data that may improve these metrics.
This one-day workshop uses a design fiction approach to explore the multifaceted future of animal-robot interaction in our society. It envisions and critically discusses scenarios in which social robots could foster relationships between humans and companion animals or enhance animal welfare. Based on the curriculum vitae (CVs) submitted during the application process, we will group participants with specific target animals. Following keynote talks by researchers specializing in animal-robot interaction and artificial intelligence (AI) ethics related to animals, participants will discuss the future of animal-robot interaction, including its potential benefits and ethical concerns.
Recent research has revealed the rich sensory perception of aquatic species, highlighting opportunities to enhance the acoustic experiences of fish in managed care environments where visitors and life support systems contribute to the underwater soundscape. While awareness of underwater noise pollution in natural ecosystems is increasing, its presence in aquarium environments and its potential effects on aquatic animals, remains underexplored. This paper presents a pilot installation, the first stage of a proposed three-part system aimed at fostering public awareness of underwater acoustics and fish communication. The full system envisions: (1) playback of prerecorded fish vocalizations and natural underwater soundscapes, (2) live audio from tanks housing vocal fish species to reveal inter-fish communication, (3) interactive tank with visitor generated sounds allowing visitors to experience the acoustic environment of captive fishes. This paper reports on the deployment and evaluation of the third component, which was implemented without the involvement of any live fish. The installation featured a stylized, uninhabited tank containing a concealed hydrophone, which transmitted real-time audio from visitor actions-such as tapping, splashing, or activating bubbling filters-through headphones. This open-ended, embodied interaction encouraged users to explore how everyday behaviors translate acoustically underwater. Observations of 74 visitors revealed surprise, empathy, and increased awareness of noise pollution. Framed within Animal-Computer Interaction (ACI), this work offers a model for ethically grounded, public-facing systems that promote cross-species understanding.
Sheep temperament, safety, and well-being relies on herd structure and dynamics. Prior works have shown that 1) microclimate can affect sheep flocking behavior and group decisions, 2) and bleating frequency and patterning correspond with individual sheep mood and temperament. We aim to better understand individual sheep and resulting flocking and herd behavior by using electronic headwear that measures microclimate, head movement, eye movement, and bleating patterns to monitor the factors that may affect sheep stress levels and to understand individuals' responses to those factors. In this paper, we present the design and development of a prototype sensing system and our plans to investigate sheep acceptance of the system with habituated petting zoo sheep. The prototype consists of audio, microclimate, and head movement sensors mounted to a similar halter system that is worn by the petting zoo sheep daily. Based on extensive conversations with sheep specialists, we expect that the sheep will grow accustomed to the placement and attachment of the headpiece after acquaintance with the smell of the device. In the future, we will add cameras to the system to analyze field-of-view and ocular movement to determine how the number of conspecifics in view affects individuals' behavior. This work is an effort to develop wearable devices for sheep in the field, whose group behaviors remain poorly understood despite individuals being regularly used in behavioral neuroscience research.
Osteoarthritis (OA) can cause great pain for dogs, limit their daily activities and negatively affect their quality of life. In most veterinary practices, evaluating pain caused by OA is done by veterinary professionals performing examinations and making subjective assessments. In this study, we proposed a novel objective approach using a wearable device equipped with an inertial measurement unit (IMU) to collect high-frequency movement data from dogs and to use machine learning methods to analyze pain status from 28 dogs. Using manifold learning, we were able to generate natural clusters between healthy dogs and those affected by OA pain. Using four simple supervised learning methods and manually extracted movement features as input, we achieved high performance using certain input data streams (100% precision for OA-pain affected dogs, 83% recall and 89% overall classification accuracy). Furthermore, using a custom neural network model, we achieved the best performance of 71% precision, 92% recall and 72% accuracy using raw IMU data as input. As a preliminary study, we also explored the use of a specific task data against using all the data from the entire session to compare the prediction performance. We also compared the prediction capabilities of different input data streams. Although preliminary, the results were promising and indicated that IMU data and machine learning methods could be further leveraged to achieve an owner-conducted early OA pain screening process in an at-home environment where dogs are most comfortable.
The exceptional growth of social media has transformed how people engage with urban wildlife, by providing new media to shape online public discourse, new methods for conservationists, and serve a tool for scientific research. However, little is known about how social media influences these areas. We performed a systematic literature review (SLR) of peer-reviewed publications that included both urban wildlife and a social media component. By doing so, we analysed how social media influences the public perception of urban wildlife, behaviour and its use to facilitate citizen science projects. We analysed (59) papers out of the original 904 identified articles with potential relevance. We found that the literature is mostly regionally focused with Asia Pacific, North America, Europe being the most represented areas. For species diversity, mammals and invertebrates dominate the literature, with a noticeable focus on species that could be deemed as charismatic, like cockatoos or species deemed troublesome, such as coyotes. The role of social media was predominantly for the facilitation of research and recruitment for citizen science. The most commonly used platform within the review was Facebook, leveraging the "groups" feature to identify and engage with relevant communities. While social media shows promise in shaping attitudes, public perception and facilitating coexistence behaviours, its effectiveness in driving real world change remains under researched. The results indicate that while social media offers unparalleled opportunities to spread wildlife awareness and engagement, which are beneficial traits to conservationists, further work is needed to mitigate its capacity to spread misinformation and temper emotionally charged online discourse.
Providing animals with meaningful 'choice' and 'control' is a central goal in Animal-Computer Interaction (ACI), but whether and how researchers describe these concepts varies across contexts. We reviewed 94 live-animal papers published between 2016-2024 in the Proceedings of the ACM Conference on Animal-Computer Interaction and found that only 37.2% used the words 'choice' or 'control' explicitly. All studies involving zoo and sanctuary animals included some form of reference to these concepts, while just 57.1% of companion-animal papers and 52.8% of working-animal papers made mention of them-either directly or indirectly. We suggest these patterns in ACI publications reflect the broad differences in institutional norms and human-animal relationships that we enact in our research. We further suggest that relational approaches to ACI - like becoming with and dignity-based perspectives - offer useful frameworks for incorporating more detailed reporting about choice and control in our work.
Social media increasingly plays a dominant role in shaping public opinion and attitudes on various topics, including human-wildlife relations. While moderation efforts such as Community Notes on X (formerly Twitter) primarily serve as fact-checking tools to combat misinformation, this study investigates their potential as vehicles for attitudinal change. We examine whether attitudes towards wildlife species exhibiting natural behaviours in urban environments can be neutralised or shifted through a Community Note that provides additional behavioural context. Using a between-groups experimental design, participants were randomly assigned to view a mock tweet either with or without an explanatory Community Note aimed at contextualising the animal's behaviour. Additionally, for the gull species, we tested whether lexical framing influences shifts in attitudes. Validated Likert-scale questions measured immediate attitudes, and qualitative responses were analysed to explore underlying justifications. Our results indicate that the presence of a Community Note significantly reduced negative attitudes. Our results indicate that the presence of a Community Note significantly reduced negative attitudes. For example, the hedgehog group saw the largest percentage reduction in negative responses (75%), followed by rats (38.8%) and foxes (37.8%). A Binomial regression test shows this positive shift in attitude was statistically significant for both the rat (OR = 3.27, p = 0.005) and hedgehog (OR = 4.95, p = 0.019) groups, demonstrating community notes' effectiveness for attitudinal change however this was not the case for the Mann-Whitney test. Our findings highlight the broader implications for detailed moderation tools in depolarizing discourse on socialmedia. We show that the use of Community Notes in our context of urban wildlife can, and does, promote both learning and increases peaceful co-existence with wildlife. We also note that the use of Community Notes could help mitigate harmful messaging, as we continue to see increases in misinformation and disinformation, particularly concerning wildlife conservation and efforts to facilitate human-wildlife coexistence. We discuss the role of social media in shaping perceptions of wildlife, its current limitations, and how interventions like Community Notes could be expanded to promote positive human-wildlife relations in an increasingly urbanized and digitally connected world.
Traditional zoo enrichment methods often lack flexible, non-food-based interactive elements that support animal autonomy, agency, and meaningful engagement. Enrichment that responds dynamically to animal behavior offers promising alternatives to support species-typical activities while giving animals meaningful choices in their experience. We present the design and evaluation protocol for an interactive acoustic enrichment system that transforms a familiar swing into an agency-based interface for zoo-housed black-and-white colobus monkeys (Colobus guereza). Our system maps swing amplitude to progressive soundscape layers-stream, insect, and bird sounds-allowing monkeys to control the complexity of their auditory environment through natural swinging behavior. Unlike passive enrichment, this approach gives animals direct control over their sensory experience, rewarding physical activity with increasingly rich acoustic feedback. The system offers dual audio outputs: an embedded speaker for direct animal enrichment and an external speaker for visitor education, with zookeepers controlling configuration through a mobile application. We propose a three-phase evaluation protocol comparing passive and interactive swing usage through sensor data, behavioral observations, and keeper assessments. This work contributes to Animal-Computer Interaction (ACI) research by illustrating how zoo infrastructure can be augmented into responsive interfaces, providing a methodological framework for developing species-specific, enrichment systems that give animals meaningful control over their environment.
The workshop will explore how digital enrichment, including touchscreen technology, may be used to conduct research into the cognitive capabilities of great and smaller apes, facilitate welfare and enrichment opportunities and support a range of multimodal inputs and outputs, associated with the screen and the environment.
Touchscreen tasks have long been a valuable tool for examining primate cognition and are now also being used at zoos to enhance animal enrichment and guest experience. We introduce a fully automated pipeline for chimpanzee facial recognition that allows for individual identification during touchscreen task engagement. Our system incorporates a YOLOv8 bounding box detector and a fine-tuned ConvNeXt classifier to detect and assign identities to chimpanzees recorded via video streams captured by an embedded camera. We implement a confidence-weighted sliding window approach to mitigate classification errors from low quality input footage. Using these techniques, we were able to achieve a high confidence identification (>80% of video duration) despite noisy video and frequent low-confidence model classifications. This work enhances research, enrichment, and conservation techniques applied to great apes.
New technologies can produce changes in concepts, such as how the concept of work changed with online interfaces, money changed with new currency technologies, and cancer changed with the introduction of microscope technology. Research on animal-computer interactions (ACI) has changed concepts like user, welfare, evaluation, and participatory design, as the design of technology with animals as users, participants, and beneficiaries has required new concepts. We propose that ACI research has the potential to transform an even more fundamental concept: what an "animal" is. We hypothesize that ACI researchers have often been operating with essentialist concepts of animal categories such as species, breeds, or even categories like dog, bird, ape, and wild animal, where those grouping have aimed to reflect "biological" or "natural" groups that are fixed and unchanging. Drawing from the social philosophers Sally Haslanger's and Ian Hacking's methods of analyzing socially constructed categories, we propose that the essentialist concept of animal categories could inhibit ACI research and produce technologies that reinforce the existence of such categories. In turn, research on animal-technology interactions, if conducted with a social constructionist conception, could disrupt conceptions of animal categories, revealing ways in which animal categories are produced by animals' material and social conditions, including their technological realities, and could change with changing technologies, opening new doors for ACI research and beyond.
Due to the recent surge in bear attacks, affected municipalities have taken various measures to drive out the intruders. Political considerations include, in many places, the expansion of shooting permits. More and more regional governmental authorities invest in technologies for efficient bear hunting, such as IP cameras and computer vision. From an ecological perspective, such procedures, which are all solely aimed at decreasing the bear population, are questionable, since ecosystems can benefit from a peaceful coexistence of bears with humans: Bears remove animal carcasses and thus return vital nutrients to the soil and prevent the spread of diseases. Moreover, the bears keep the deer population in check, which is advantageous for the forest vegetation. Conflicts start, once the bears intrude populated areas, which are often rural villages and suburbs with a low population density. Naturally, bears do not see humans as prey and respect their size. When the animal attacks, this happens rather for territorial or protective reasons, e. g., after an unexpected encounter with humans unsettled it. In this work, we analyze means to eradicate the danger: We dynamically apply Weiszfeld's algorithm for finding the shifting geometric median between time-windowed black bear sightings and use the results for implementing an early warning system. In addition, we weigh the spotted locations and predict the expected migration behavior between bear habitats using a simulation with intelligent agents. As a case study, we apply our software to data from Romania and Fukushima and identify the positions where bear repelling facilities should be placed most effectively.
Guide dogs provide vital support to individuals with visual impairments, yet the training process demands significant investment of time and resources. A significant component of this process is mutually identifying the right individual for each guide dog based on factors such as natural walking speed and handle-pull characteristics. We present a novel proof-of-concept wearable sensor system that records pull-force and harness motion data to aid in guide dog training and evaluation. The system integrates a modified harness with force sensors and an inertial measurement unit (IMU), providing a user-friendly and low-cost wireless solution for data collection during training sessions. Preliminary results show a strong correlation between average pull-force and profiling from professionals at a prominent guide dog school. Variations in left-to-right force distribution was found across individual dogs. Additionally, the system enables collection of motion data simultaneously with pull. We present early work validating pace and gait estimation using harness-based IMU data, with commercial gait analysis equipment as ground truth. We apply preliminary machine learning techniques to extract key features from pull and motion data with the goal of enabling advanced visualization and analysis of future gait and pull characteristics. As a quantitative objective tool, the presented system paves the way for improving guide dog training, optimizing dog-handler pairings, and potentially improving the success rate of dog training programs in general.
