Examples from real-world use
The data-driven animal technology focusing on measuring and suggesting interventions that are on the market are primarily focused on domesticated animals, that is, animals with whom we as humans have a mutual relationship affecting caregiving and reproduction. This includes distinct categories like companion animals, such as dogs (canis familiaris) and cats (felis catus), and farm animals such as cows (bos taurus). The focus of such technology on such domesticated animals likely reflects our closer relationship to these species and the need for support in our interspecies caregiving.
These examples are based on extensive research and interactions with vendors of animal-centered technology, market analysis reports, and insights from recent research which has analyzed over 8000 Amazon reviews of commercially available animal-centered technology (activity trackers, location trackers, etc.) , and studies investigating users of commercially available animal wearables  and their perceived impact .
Companion animal wearables are a quickly growing sector in the companion animal industry and cover a variety of, often data-driven, technology. While such technology is primarily visible in the context of pets, they may be suitable for a given species regardless of it’s exact role. That is, technology designed for dogs may be equally useful and suited for pet dogs, working dogs (e.g., detection dog, search and rescue dog), or service dogs (e.g., blind guide dogs or emotional assistance dog). Pet wearables in particular have been noted as one of the top industries for aspiring entrepreneurs to enter, given a large and growing customer base, relatively low investment for entry, and fairly low competition . It should be no surprise that there is a proliferation of different data-intensive animal-centered technology being released and promised. Indeed, a recent article in the New York Times  focused on AI-driven technology for pets discussed the diversity of technology under development and on the market for pets.
Not all of such technology will promote interaction between animals and humans and give rise to an IIS. For example, in 2016, Wagg Pet Foods produced a prototype of a television remote control  optimized for dog physiology (color schemes fitting to dog vision, buttons suited to dog physiology). This is an example of a technology developed for use by one species, but not capturing or sharing data with the dog’s human owner to inform their understanding of, say, the dog’s likes or dislikes for particular TV channels based on their interactions with the remote. It thus remains an isolated technology, rather than giving rise to an IIS.
More potential for an IIS to emerge comes from the growing number of interactive speakers and cameras developed to increase interaction between owners and pets when pets are left alone at home. An example of such devices is Petcube , marketed as an interactive assistant for ‘pet parents’. Research has shown that the core functionality of such devices makes sense, as pets are capable of interacting with their owners through such technology, such as e.g., dogs using Skype to communicate with their owners . These devices, while not seemingly giving rise to an IIS yet, make an important first step by enabling interspecies interactions.
When such technology goes further, and captures and processes data to inform humans how to structure those interactions, and in doing so inform interspecies interactions, they become an IIS. A technology closely related to the interactive assistant shows just such an example. Smart food dispensers are based on such interactive assistants, representing more complex technology where a pet owner can see the amount of food currently in the bowl through a weight sensor, or receive a ‘communication’ from their dog, and in return instruct the technology to dispense food. Such technology may even aid in veterinary care by providing veterinarians with more objective diet information than owner reporting. This shows that such information flow may be both indirect and direct, either when a dog indirectly triggers the system to send a signal to its owner by emptying a bowl of its food, or by doing so directly by barking into the speaker, whereupon the owner may be stimulated to release food. If such interaction is intentional, interspecies communication is indeed enabled by the technology.
Consider one of the more prevalent types of technology for companion animals, pet wearables. The market is filled with devices to monitor location of pets, track their activity and fitness, or even provide detailed insights into their health. These wearables, similar to human wearables, typically exist of a piece of sensor-laden hardware, worn by the pet, and relevant controlling technology, usually in the form of an app for the owner’s smartphone. In the context of pet wearables, location trackers are typically based on GPS or RF-based solutions, while activity trackers are typically based on accelerometers, using Bluetooth or WiFi connectivity to share data with controlling devices . A popular activity tracker such as FitBark , thus consists of a device worn by the dog, measuring its activity akin to a regular human wearable, whose data is processed into human readable form and accompanied by suggestions for interactions (e.g., walk the dog more). Information thus flows from raw captured accelerometer data, to processed human-readable descriptive data of daily activity, to normative instructions informing concrete interactions.
It is this exact information flow that gives rise to the IIS. This therefore goes beyond simple technology enabled interactions, as the IIS provides an information loop in which a data-driven system tells the human owner how to intervene in different aspects of their dog’s life, or, more accurately: suggests interventions to different processes affecting their caregiving to the dog. For example, increasing or decreasing activity, increasing or decreasing food and calories based on that activity, and so on. An IIS has emerged, as visualized by the simplified data flow in Fig. 1 consisting of actors of different species, technology capturing data of one species, and processing it for consumption and acting upon by another species.
Even before technology for companion animals became widespread, farm animals had been subject to increasing use of technology to optimize different processes.
Similar technology is available for farm animals, often at larger scales. Rather than individual food dispensers, auto-feeding solutions for livestock consist of sensor-driven systems which estimate the amount of food needed for farm animals based on their physiology and environmental conditions, such as e.g., determining feed during developmental stage when growing chicks for poultry meat. Such technology, however, do not give rise to an IIS as they are one-sided data-driven systems which automate the decision-making, taking the human out of the loop to decrease workload.
Yet, different systems built to optimize scaling and reduction of workload place the human central in the loop. For example, a sensor-based system for the monitoring of health and welfare data of dairy cows. Monitoring technology worn by each individual cow contains sensors which capture activity data and vital signs, sending this towards a central IS where it is processed and visualized for a farm operator to keep track of the physical and mental state of each cow. Based on data analysis, the software can inform the farm operator of cows which are showing indicators of factors that may impact the quality of their produced milk (e.g., stress levels, lameness, overheating), and, just as with companion animal systems we discussed before, inform them of concrete interactions that are required to correct this. In this context, this may be both one-on-one interactions between a human and an animal, such as stimulating a cow to walk around, or provide them with additional cognitive enrichment, while it may also be indirect technology mediated interactions, such as turning on air-conditioning to reduce overheating.
Here, just as with the case of pet wearables, an IIS emerges where a combination of technology measuring data of one species are processed to inform a human actor how to best intervene in the support of a particular process, as visualized in Fig. 2.
It would be remiss, however, to move on without explicitly considering that an IIS in the context of farm animals, even if ostensibly meant to inform about animal welfare (albeit with a commercial ulterior motive), does eventually lead to fatal outcomes for many of the animals within the IIS. Consider, for example, the use of data-driven technology to reduce stress for pigs (Sus scrofa domesticus) in pre-slaughter phases in order to improve meat quality . An IIS may similarly emerge that processes data (e.g., cortisol level monitoring) to assess and inform whether certain actions should be taken to improve welfare on the short term (e.g., provide cognitive enrichment, avoid particular handling), which is ultimately tasked with the process of improving meat quality on the long-term (cf. the context of using enrichment devices with large pigs for such purposes ). Thus, while I phrased that the IIS in this farm context would inform a human actor “how to best intervene”, it must not be taken for granted that this is in the best interest of animal actors per se—especially when these interventions serve a business process that is not at all concerned with animal best interests (i.e., optimizing meat quality and conditions for slaughter).
Wildlife has been subject to human technologies for a significant amount of time. For instance, the use of wildlife crossings, whether tunnels or bridges, that allow wildlife to safely cross human made barriers like highways. Or their inverse, wildlife grids, that discourage animals from crossing into particular areas. These are concrete, physical examples of technologies with an underlying similar purpose—to manage how humans co-exist with wildlife, and to reduce our negative impact on their very existence.
In other words, these are technologies to manage how we co-exist with other species—an increasingly important topic as human civilization encroaches on animal populations and affects the way they survive . Wildlife monitoring, in particular may be of most interest as a source of technologies that give rise to an IIS, as these technologies and the data they capture are essential to “inform conservation and management decisions to maintain diverse, balanced and sustainable ecosystems.” . Indeed, wildlife monitoring technologies are essential to understand the many ways in which human activity and civilization impact on wildlife . This seemingly less anthropocentric purpose than the companion and farm animal examples I have discussed so far is achieved through a variety of technologies—from simple manual counting to cameras, to even unmanned aerial vehicles . The data they capture is already raising discussions on what such data contains beyond its primary purpose, and what (malicious) behavior it may unintentionally inform—see e.g., debates on the potential of digital poaching .
It seems evident, therefore, that wildlife technologies similarly give rise to interspecies information systems (IIS). There is an important distinction that becomes apparent, however, going even beyond the lesser anthropocentric nature of such an IIS. Data collected by these devices, often in major project and group efforts do not tend to inform actions with similar immediacy as in the context of companion and farm animals. Rather, information collected by wildlife technologies (e.g., herd observations, impact of human activity on quality of life) informs separate decision-making processes that set policy towards wildlife, often on governmental level. There is an additional layer of complexity with a decision-making structure that essentially introduces a gulf of execution between the initial observations, interpretations, and the actual actions taken towards wildlife–making these additional layers vital to the actual decision-making that informs the actual actions that then affect wildlife populations .
In terms of the IIS that emerges, as Fig. 3 visualizes, the flow of information thus does not go simply from observation to action, but necessarily includes an additional step of interpretation, only then resulting in a mandate for action towards animals.
Given the focus of wildlife monitoring being on the preservation of entire populations of animals, a wildlife IIS may also elicit debate on what it means to “best intervene” when animals are in need. While in the case of farm animals human interests likely trump animal interests on the long-term, with wildlife it may be more like that interests of populations of animals likely trump individual animal interests on the long-term. Consider, for example, a wildlife IIS that informs human actors of the need to cull a rapidly growing population of animals to ensure they do not threaten their own survival by entirely depleting food sources. This shows a further nuance in the notion of interventions and impact, in that it may be targeted either at individuals or entire populations of some species.
Through these examples, I have shown that the essence of an IIS is to inform humans of action to take towards animals, and established some important commonalities that IIS share regardless of functionality (e.g., activity tracking, health monitoring) or involved species (e.g., companion animals, farm animals, or wildlife). Specifically, that:
Commonalities of an IIS
an IIS enables a flow of information across species, typically informing human stakeholders of physiological or behavioral states of another species;
this information informs, often intentionally, informed interventions from one species to another (whether to individual or groups), typically to affect their physiological or behavioral state (whether positively or negatively); and
those interventions impact a species (whether individuals or groups) to aid in an external process (whether informal or well defined).
It is therefore important for the informed design and use of IIS that they understand in detail where and how information is created, and how it flows between components of the system.
Below I discuss how these commonalities may be informed by, and grounded in, relevant theoretical frameworks, in turn informing a model of a general IIS.
Interspecies information flow
The flow of information between species in an IIS is typically meant to enable interactions, or communication between distinct species. From the microbial level to interaction of different mammalian species, interspecies communication has been studied extensively. From differences species of old world monkeys (cercopithecus) having mutually intelligible warning calls , play between chimpanzees (pan troglodytes) and bonobos (pan paniscus) , to the well studied interspecies communication between dogs and humans, showing the formation of such communication even from a young age . Research has suggested that we should not restrict ourselves solely to reciprocating communication, as unilateral ‘interactions’ play a major role in maintaining interspecies communities . Kostan proposed a theory of interspecies communication  based on assessment and management of information which provides insight into how the direction of information flow may enable interventions in increasing levels of reciprocity. It classifies interspecies communication into:
Unidirectional assessment (one species acting upon another species’ intra-species communication)
Bidirectional assessment (both species acting upon each other’s intra-species communication)
Asymmetric communication (one species informing another species)
Symmetric communication (two species information one another)
Assessment, whether uni- or bi-directional, are not relevant to understanding the information flow in an IIS as these constitute one-sided ‘consumption’ of information, where no interaction between species is enabled. For example, in the context of dogs, an example of unidirectional assessment could be a person hearing several dogs barking loudly in a street, and inferring that it must be a sign of danger, hence deciding to avoid walking down that street. As the enabling of communication is key for IIS to emerge, the primary distinction to make is thus whether such communication is asymmetric or symmetric. From the examples I have discussed above in Sect. 3.1, this shows for example:
Directionality in an IIS
Asymmetric communication: When an IIS enables an actor of one species to intervene in the behavior of another species.
For example, the monitoring of livestock, where a human operator monitors data of a herd of cows and intervenes where appropriately, while cows are unaware of the monitoring. Similarly, pet wearables present an asymmetric information flow, where a dog is monitored and software suggests how the owner may interact with them or intervene in their behavior, while dogs are unaware or engaging similarly in the IIS.
Symmetric communication: When an IIS enables actors of multiple species to intervene in each other’s behavior.
The examples of interactive assistants connected to feeding systems enable symmetric communication. A dog may share information and request action of their owner (e.g., barking to request food), while the human owner similarly may engage in interactions and request action of the dog through audio/video link.
It is thus important for the informed design and use of an IIS to account for the directionality that its information flow enables.
The information flow within an IIS is meant to inform interventions from one species to another. It is thus important to understand how the key components like actors and technology within the IIS relate to each other to enable such interventions. As a start for theoretical grounding of how different components of an IIS are needed to enable interventions, consider the SHELL conceptual model that describes the interactions between the four main components of a socio-technical system: Software, Hardware, Environment, and Liveware . Each of these components interacts in a given system, where here, in particular, the interactions of actors to other components are key to understanding how an interspecies intervention is enabled. As the examples in Sect. 3.1 already revealed, depending on the exact technology, actors of a given species may only interact with some of the hardware, and these interactions may be passive or active. This means that we need to explicitly distinguish the interactions that actors have with hardware, software, and each other, showing that interspecies interventions are effectively enabled as three successive interactions:
Key interactions in an IIS
Actor–hardware interactions can be active or passive, For example, with wearable technology, animal actors typically have a passive relation to the hardware, simply being made to wear it. Other technology, such as smart food dispensers show passive interactions between animal actors and the hardware, triggering a signal for more food simply by emptying the bowl. Human actors, however, will typically interact with both the hardware worn or used by animal actors in order to ensure its suitability and appropriate fit (e.g., ensuring the animal actor is not bothered by a wearable), as well as separate hardware used to control and monitor these devices.
Actor–software interactions, are the critical aspect enabling an interspecies intervention, as human actors consume information and suggestions how to interact with, or intervene in another animal actor’s behavior.
Actor–Actor interactions, finally, are both the information-driven interventions that a (typically) human actor takes towards animal actors in the IIS to aid in an external process such as caregiving or quality management, and the human-human actor interactions that may first occur as a prelude to informing those cross-species interactions, as e.g., in the context of wildlife management decision-making.
This emphasizes the importance for designers of having a detailed view on what technology actors of different species interact with, and explicitly distinguishing between human and animal actors in terms of their interactions to other components of the IIS.
Many of the interventions that actors make across species boundaries informed by an IIS will lead to concrete impacts on an animal’s physical wellbeing, both on the short and long term. Over twenty-five years ago, Hirschheim et al.  already noted that IS design is “not merely a technical intervention but involves social and ethical dilemmas that affect the human, social and organizational domains.” Interspecies interventions enabled by an IIS showcase this complexity: both human and animal actors, as well as the wider societal and organizational environments in which both species co-exist are affected by the interventions that the IIS suggests–and as Sect. 3.1.2 has shown, not always for the better. Moreover, the very design of the IIS may even further strengthen assumptions that are (unintentionally) harmful to animals . As interspecies relations are highly complex , it is therefore important to systematically treat the (potential) impact that actors have on actors of other species.
Understanding interspecies interventions enabled by an IIS as symbiotic relationships, allows us to distinguish between different levels of harm and benefit to the involved actors of different species , including relationships that are:
Amensalistic (harming one species, while not affecting the other)
Parasitic (benefiting one species, while harming the other)
Commensalistic (benefiting one species, while not harming the other)
Mutualistic (benefiting both species)
The impact of interventions enabled by an IIS through the subsequent interactions described before can thus be described in increasing levels of desirable symbiosis:
Impact levels in an IIS
Amensalistic impact, to a certain extent, may primarily arise unintentionally during the design and use of an IIS if technology is designed without due regard for all involved actors. This may be linked to hardware, such as for example wildlife technology leading to unintentional death of its subjects . Perhaps more critically for an IIS, software-linked harms may arise if interventions suggested by the IIS need are not accurate and appropriate. For example, a dog owner may unintentionally cause musco-skeletal injury in a dog through overtraining as a result of erroneous advice generated by an activity tracker.
Parasitic impact is unlikely to arise as an intended consequence of an individual interspecies intervention. However, interventions are in the aid of external processes–not all of which will serve the best interest of animals in the long term. We might thus critically assess whether the long-term benefits of processes supported or enabled by these interventions harm a species in the IIS, while benefiting another by e.g., trivializing their caring needs while giving a human owner a false sense of security in their caregiving capability, or more commonly, preparing farm animals for slaughter.
Commensalistic impacts are seen in e.g., the examples of farm animal technology used for short-term beneficial interventions. The interventions that the IIS in Fig. 2 enables give direct relief to the cows by e.g., reducing their heat stress, or providing cognitive enrichment, but do not directly provide benefit to the human actor in the IIS. Rather, as a reverse of parasitic impacts, here benefit may more likely arise on the long-term as a result of the external processes that these interventions support by e.g., increasing the quality of the produced milk, which in turn brings commercial benefits.
Mutualistic impacts of an IIS are when an intervention benefits both species of actors. Pet wearables, activity trackers in particular, provide an example of such benefit. A typical concrete intervention that a dog activity tracker may suggest is to simply take the dog for a walk. As research into the motivations and actual use of dog activity trackers has shown, the use of these trackers leads to improved activity and potential health benefits not only for the dog, but to increased motivation for fitness of the human owner as well .
The critical reader may recall that Sect. 3.1.3 discussed another complication beyond that of balancing short and long-term impacts. Indeed, it may be the case that interspecies interventions harm some individuals of a species, or even prove to be fatal. For example, killing (‘culling’) a set number of animals in a population of wildlife species, benefiting their population as a whole by ensuring the population does not deplete its food sources to the point of no recovery. Here, a seeming amensalistic interspecies intervention (killing the set of number of wildlife animals) leads to an eventual commensalistic outcome (benefitting the wildlife, while not directly benefiting the human actors involved (although, of course, on a much longer term, these actions can be interpreted as benefiting human actors by ensuring stability of our overall shared ecosystems). Designers of an IIS may thus also need to consider how to balance the short-term and long-term impacts of interspecies interventions their technologies inform on, potentially separating them into separate interventions so as to allow for critical reflection on whether some harm gives way to the greater good.
Does an IIS control or inform?
We have now established certain commonalities of IIS, in that they enable an interspecies information flow with a given directionality, in turn informing and enabling interactions that lead to interspecies interventions with concrete, real-world impacts on one or more species. An important consideration designers might reflect on is whether that means an IIS informs interspecies interventions, or whether it controls them.
It might be tempting to consider the IIS as a system that controls interspecies interventions, akin to a cyber-physical system (CPS) that extends human capabilities in terms of sensing, decision-making and action , while keeping humans in the loop . There are indeed similarities to CPS from what I have established as key requirements for an IIS–both are preoccupied with enhancing capabilities , need to be data-driven, and values-based (cf. ), concerned with continuous detection of challenges , and ensuring safety of more-than-human interactions becomes an important requirement (cf. ). In some specific contexts, parts of an IIS may indeed be preoccupied with (semi-)automated control—in particular in the context where animals are effectively seen as resource, such as e.g., automated milking systems which control milking processes while also informing a farmer of further potential actions to take. Designing such systems indeed requires significant considerations on data capture and management to ensure adequate control (cf. [13, 14]).
Yet, as the considerations and design challenges from other examples analyzed in Sect. 3 show, the key essence of an IIS as a whole is to inform of interspecies interventions to take, while not actually intending to control such actions in (semi-)automated ways. Moreover, as we have seen from just the three examples of different technologies analyzed, the interspecies interventions taken suggested by an IIS pass through multiple layers of additional consideration and complexity before culminating in a realized interspecies intervention. In its simplest form, this might mean whether a dog owner followed up on suggestions provided by the processed data. Indeed, deviation where necessary is important so that, e.g., veterinarians can overrule suggestions from the IIS. In the context of wildlife IIS, Sect. 3.1.3 there is an additional gulf of execution that separates the actual apparatus of the IIS itself and the interventions taken: culling of wildlife is certainly informed by an IIS through its monitoring or recording of animals and suggestion of potential interspecies interventions, but before any such interspecies interventions are taken, they go through external decision-making, policy considerations, likely involving entities and resources entirely separate to the IIS itself.
Thus, designers, in their design thinking, ought to think of an IIS as a system that enables information flow between species to inform of interspecies actions to take–whether improved caregiving, optimization of farming, or tracking of wildlife populations, and focus on that information flow and the interspecies interventions it enables.