Designing Multispecies Worlds for Robots, Cats, and Humans

From cleaning our homes [48], to mowing our lawns [55], to delivering shopping [54] and couriering items around hospitals, [53], robots are finding their place in daily life. As they do so, they will inevitably interact with and be encountered by animals. These might be companion animals, the pets who share our homes or the guide dogs who help us navigate public places, but they might also be wildlife. Often these encounters will be unplanned and secondary to the robot’s intended task, for example cats riding Roombas¹, guide dogs being confused by delivery robots [10], or hedgehogs having to navigate in a world inhabited by lawn mowing robots [42]. However, they could also be intentional. We could design robots to serve animals too. Despite the inevitability of such encounters, planned or otherwise, little is known about how to design robots for animals. Can we even trust them with each other?

We present Cat Royale [8, 44, 45, 46, 52], a creative exploration of designing a domestic robot to enrich the lives of cats through play. A robot, in contrast to non-embodied technology, offers the advantage of physical manipulation of objects in the enclosure. Therefore, it lends itself well to the context of cat play as it is able to, e.g., wave feathers, provide treats, or drag strings around the enclosure for the enrichment of the cats. Cat Royale was an artist-led project  [5] that set out to tackle the wider question of trust in autonomous systems. The idea of creating a playful robot for cats emerged as a means to achieve this goal, while also providing benefits for the cats. Over the course of eighteen months, the artists, in partnership with a team of HCI/HRI researchers, and supported by experts in cat play and welfare, and animal-computer interaction created a bespoke enclosure, a ‘cat utopia’ in their terms. At its centre, a robot arm manipulated a variety of toys to try and entertain a small community of cats who lived there for six hours a day for twelve days. During this time, the robot offered over 500 play activities for the cats. In the background, the artists, roboticist, animal-computer-interaction specialists, machine learning experts, and animal behaviour experts worked tirelessly to ensure their safety, wellbeing and hopefully, pleasure.

While Cat Royale was undoubtedly a unique experience, and not one that can be directly transferred into people’s homes, and while an art project is an unusual approach to designing robots and animal-computer interactions, we propose that Cat Royale offers important insights into the design of robots for companion animals. The artists’ meticulous attention to detail in delivering a safe and entertaining experience for the cats drove them to ‘go beyond’ many of the conventional aspects of robot design. Yes, they had to choose and adapt robot hardware and implement software for motion control, tracking the cats, and recommending games. However, they also had to carefully embed these within a wider environment that allowed the cats to engage appropriately, while humans could oversee safety and enjoy the spectacle. In short, they needed to create an entire ‘robot world’.

We reflect on the design of Cat Royale, and how the experience unfolded with the cats, to draw out wider implications for creating robot worlds (or, more formally ecologies [31]). We distil two key contributions. First, we highlight the importance of world design for robots; of designing key aspects of the places they inhabit alongside companion animals and humans, including providing safe spaces and observation spaces for all parties, enabling robots to flexibly interact with passive objects, and finding a decoration scheme appropriate to all. Second, we reveal how designing a robot to autonomously engage animals requires consideration of interspecies interactions, from spectating, to overseeing, to cleaning up the resulting mess. Our aim is to sensitise designers to the complexities of robot encounters with animals—intended or otherwise.

Related work falls into two categories. First, with the immediate goal of Cat Royale in mind, we review previous attempts to design digital technologies to play with cats. Second, we consider a body of work that speaks to the wider idea of designing robot worlds and digital ecologies that has emerged from our reflections.

With Cat Royale, we have followed the approach of Performance-led Research in the Wild [5] as a distinctive way to engage HCI research with the performing arts to the benefit of both. Falling under the broad umbrella of Research Through Design, this is design-led, unfolding through the practical creation of public performances and installations and subsequent reflection on these to generalise design knowledge. It also sits within a broader history of HCI engaging the arts, both to address their distinctive challenges with regard to engagement, presence and other matters [17], but also to surface fresh perspectives for the field, such as: provoking interpretation through ambiguity [49]; deliberately designing uncomfortable interactions [6]; promoting environmental sustainability through activist art [15] and various provocations arising from artistic speculative design [2].

Papers grounded in Performance-led Research in the Wild have appeared in HCI for more than 20 years with the approach being explicitly named and articulated in 2013 [5]. Its distinctiveness arises from the particular relationship it establishes between art practice and HCI research. Typically, the process begins with a team of professional artists and computer scientists, including HCI researchers, building a relationship. At some point the artists propose a creative vision and/or opportunity to create a new artwork. While this may be in part grounded in the technical capabilities of the researchers, and perhaps even broadly inspired by their research (especially if the collaboration is partly funded by a wider research grant), the vision is essentially an artistic one, and often takes the technology in an unexpected direction, sometimes offering a critical perspective on it. The first role of the researchers is to help the artists realise their vision, collaborating with them to iteratively develop software and hardware, and enable early public performances. In return, the researchers get to study the artists’ rationale and process, and the audience’s experience. Reflection across these often yields unusual perspectives on interaction.

In this case, the artists—who have been collaborating with some of the authors on multiple projects for decades—were invited to make some form of artwork that would explore the question of trust in autonomous systems. The artists have over 30 years of experience using new technologies to create a variety of artworks, including performances, games, films, apps, and artistic installations. Following exploratory workshops at which they experienced various robots, their artistic response was Cat Royale, which the researchers then helped implement over a period of 18 months. The following account draws on design documents, meeting notes, observations, and autoethnographic field notes and video recordings (especially from the ‘robot operator’) to reflect on the process, up to and including what happened when the cats played with the robot. Observations and autoethnographic field notes used to give the account of Cat Royale were made by various members of the team including the artists, the robot operator, and members of the research team who observed the activities taking place. This material does not cover public reaction to watching the final edited film, which is still being deployed in museums and galleries and will be reported in subsequent papers. Instead, the contribution of this paper is to report on the lessons that emerged through designing a unique and challenging artefact.

Our project was advised by, and received ethical approval from, three ethical review boards at the lead researchers’ University. The Animal Welfare and Ethical Review Body (AWERB) ensured compliance with the regulations and law governing research involving animals in our country; the Committee for Animals and Research Ethics (CARE) in the School of Veterinary Science shaped and approved our cat welfare protocol. The Computer Science research ethics Committee (CSREC) approved the human-facing aspects of the project including consent, anonymity and data collection and management [8].

To investigate multispecies interaction between human and non-human actors with a (semi)autonomous system, the artist-led Cat Royale project team, comprising artists and a multidisciplinary set of researchers, asked itself the following question:

This section outlines the project and its non-human and human components and actors. The design process and the accompanying challenges of Cat Royale, which this paper focuses on, were developed over the course of eighteen months.

The Enclosure. The project’s centrepiece was a bespoke enclosure, built for and inhabited by three cats. The three cats, a parent and two offspring, spent six hours a day (two periods of three hours) over twelve consecutive days within the enclosure, a total of 72 hours. As we describe below, the enclosure was designed, in the artists’ words, to be a ’utopia’ for the cats, a luxurious space that could cater to their various needs, including play. Its striking visual design was also intended to convey the idea of luxury to the audience when filmed via the eight cameras embedded inside.

The Robot. Placed in the centre of the enclosure was the robotic arm, whose sole purpose was to play with Ghostbuster, Clover, and Pumpkin, the three inhabitants of the enclosure. Provided with access to a series of custom end-effectors and a wide range of different toys and attachments, the robot was able to offer a variety of games and treats. At regular intervals, the robot’s underlying system proposed a robot action aimed at maximising a specific cat’s assessed happiness (see Section 4.5). If approved by the lead artist of the installation, the robot would execute a playful activity targeted at increasing the cat’s happiness, while being monitored by the robot operator to ensure safe robot movement in close proximity to the cats. To further ensure animal welfare, each cat’s individual stress score [28] was ranked at 15-minute intervals by the team’s specialist in feline behaviour. The robot performed more than 500 activities over the course of the 12 days, ranging from providing small cat treats², to simple toys like a cardbox or a string, all the way to more elaborate toys such as an orange bird toy or a battery powered wiggling fish toy. Following each activity, each cat’s engagement—as a proxy for positive attitude towards the robot’s provided entertainment—was ranked using the Participation in Play scale [18]. The task executed (i.e., ‘present a prey game with a toy bird’), the targeted cat, and the impact on happiness where subsequently fed back into the robot’s machine learning system (i.e., the decision engine) informing future proposed activities.The Installation. All cat and robot activities within the enclosure were constantly filmed using eight cameras (see Figure 5 1), the output of which were mixed in real-time by an experienced television vision mixer to capture the best possible view of the action and to add descriptive captions to it. The footage was subsequently edited into an eight-hour long film that was exhibited to public audiences at the Curiocity Brisbane World Science Festival³ (see Figure 2 left), and at the National Science Gallery in London⁴ (see Figure 2 right) for six months. Short highlight videos (2 – 5 minutes) were also edited and released online during filming.

The remainder of this section, elaborates on core elements of the design of the cat and robot world that is Cat Royale.

We now turn our attention to the events that transpired during the deployment of Cat Royale. While our aim in this paper is not to evaluate the success of Cat Royale in terms of effectively playing with the cats, we do note that, from the perspective of the artists and wider team, the experience was judged to be successful. The successful completion of Cat Royale, while not formally evaluated in this paper, is multilayered. Firstly, the cats safely inhabited the enclosure for twelve days without needing to be withdrawn due to stress or injury. The cats waited in front of the door to the enclosure, seemingly awaiting to gain access to it. Secondly, within the enclosure, they voluntarily chose to engage with the robot’s games, typically descending from their perches to play for several minutes at a time. Thirdly, the team were informally (i.e., not backed up by a scientific analysis), of the view that, on the whole, the play was pleasurable and not frustrating for the cats. This sentiment was shared by the animal welfare officer based on the cats’ body language, as well as the cat owner. Lastly, Cat Royale seemed enjoyable for humans to watch. With countless visitors at the installation in Brisbane, tens of thousands of views on YouTube, thousands of likes on social media, and extensive engagement in comment threads across various platforms, the project effectively captured public attention. However, there were revealing incidents along the way. In the following, we draw on autoethnographic field notes and video recordings from the team, especially from the two robot operators and the lead artist, to paint a picture of how Cat Royale unfolded. We present a range of vignettes from typical operation to more challenging examples—requiring careful management—all of which presents the orchestration of Cat Royale.

Reflecting on the artists’ final design and the experience of the cats, we suggest that Cat Royale provides an existence proof that it is indeed possible to design a general purpose robot to play a variety of games that repeatedly engage cats in what appear to be naturally playful behaviours (e.g., that at first sight appear to engage their natural hunting behaviours). The cats did not appear to get bored, but rather to become confident, even assertive, in engaging the robot in such play. It may even be possible for a future robot to learn the preferences of individual cats and to adapt its play to them.

However, as we saw in the many details of the artists’ design, this was far from being a simple proposition. Delivering Cat Royale required the artists to innovate solutions to a wide variety of challenges, including: choosing an appropriate robot; designing its movements; introducing systems to make these safe; developing a decision engine to recommend games; choosing suitable (passive) toys and adapting them so the robot could deploy them; building a magnetic rack storage system for the toys; creating a surrounding enclosure with spaces and facilities to make the cats comfortable and meet their other needs; appropriately decorating this for the cats, computer vision system and human audience; creating an external control room from which the play could be monitored and broadcast; and putting in place a wide variety of supporting humans roles. Furthermore, to make Cat Royale a success they had to ensure that all of these elements worked together as a coherent whole.

What is notable about this list of activities is just how much of it extends beyond designing the robot itself or its direct interactions with the cats (though these certainly appear). Rather, the artists designed an entire ‘robot world’. This reflects the idea of understanding and designing interaction in terms of entire ecologies, not just devices, as we reviewed earlier. Cat Royale provides an extreme example of designing an entire information ecology of people, technologies, practices and values in a local environment, now extended to also include animals. However, it is also a highly unusual example; in this art project, the artists had more or less free licence to be creative with the design of their robot world. They also had the luxury of starting with a blank page. Cat Royale seems unlikely to directly transfer to people’s homes due to its size, cost, need for a humans in supporting roles, and the fact that most homes are already replete with other ‘stuff’. So what broader lessons can we generalise from this example. What does Cat Royale tell us about designing more everyday robot worlds for cats or other animals?

We now look under the surface of its design to distil underlying principles, grouping them under two broad themes that capture the essence of how the artists approached their robot world: designing the robot’s physical world and designing multispecies interactions also involving humans.

In this paper, we have presented the artist-led project Cat Royale, an exploratory investigation at the intersection between art and human/animal-robot interaction. We presented a detailed account on the design of the project, emphasising both the robot design and the design of the wider context in which it was situated, i.e., its ’robot world’. We also revealed that the importance of humans-in-the-loop cannot be understated. We described the unravelling of the events that transpired as part of the twelve day long installation, aiming to offer the reader insight into both the routine and the more challenging animal-robot interactions of Cat Royale. We then present two key implications which emerged through the Cat Royale project, which could inform future research and development of domestic robots. Firstly, we stress the importance of considering the environment beyond the robot design to include the design of the wider context, i.e., the physical world the robot and other actors are expected to share. Secondly, we highlight the need for designing multispecies interactions to include human involvement during robot supported enrichment for companion animals, as necessary functions implicated in animal-robot interaction (e.g., holding responsibility for animal welfare or recovering from breakdowns) cannot be outsourced to autonomous systems.

This work was supported by the Engineering and Physical Sciences Research Council [grant number EP/V00784X/1] UKRI Trustworthy Autonomous Systems Hub.