This paper was originally published in the proceedings of the 2020 International Symposium on Electronic Arts (ISEA), accessible at this location: https://isea2020.isea-international.org/

If you wish to cite this paper, please use the following publication information: Thibodeau, J., & Yolgörmez, C. (2020). Open-source Sentience: The Proof is in the Performance. In the proceedings of the 2020 International Symposium on Electronic Arts. Montreal, Canada.

Open-source Sentience:
the Proof is in the Performance

Joseph Thibodeau and Ceyda Yolgörmez
Machine Agencies Research Group, Milieux Institute
Concordia University
Montreal, Quebec, Canada
joseph.thibodeau@concordia.ca, ceyda.yolgormez@concordia.ca

Abstract

This paper asks how AI can change the consensus around the notion of sentience, through a specific focus on the intertwining of culture and materiality, as well as human-nonhuman relations. The question of sentience in machines is generally regarded as a rather fetishizing notion that obscures the specific assemblages of AI. We would like to open up this notion of fetish, and provide a theoretical map for an inverted way of understanding how fetishization of AI could be mobilized in cultivating an interspecies community. [1] The ideas reflected in this paper are drawn from a research-creation project that took place in Spring of 2019, Machine Ménagerie, in which a collection of small autonomous robots serve as a medium for considering different understandings of human-machine interaction. Machine Ménagerie creates the basis for interrogating the exclusive definition of sentience as a measurable property. We argue instead for an approach that would emphasize (1) the relational nature of the notion of sentience, and (2) the rituals of care and friendship in relating to nonhuman others. In this context, sentience is not something that beings own, or that humans bestow upon things, but rather is something that is continuously achieved and in which humans and nonhumans participate.

Keywords

relational sentience, human-machine relations, artificial intelligence, affective interaction, rituals of interspecies community

Introduction

Contemporary imaginaries surrounding AI contribute to the binary logic of domination. The futures of human-machine sociality is generally reduced to a question of who dominates whom. This question foregrounds the machines and humans as separate entities-in-themselves, in that the questions focus on the agential qualities of the machines, what their affordances should be so that we avoid an accidental superintelligence, which would be a doomsday adversary in the worst case, or the harbinger of corporate power at least. [2] While such anxieties are legitimate, they tend to obscure the significance of other manners of being with machines, and perhaps temporary achievements that might aggregate into a mutual intelligibility and different co-constitutions of reality. This paper is written in response to an already ongoing reconfiguration in perhaps less explored realms of AI relations, specifically those that delve into machine sentience. What follows is a theoretical account of how to approach the notion of sentience relationally. We emphasize the cultivation of an attention towards the concrete situations and encounters where machines are treated as sentient, as opposed to the expectation of a particular type of agency that would then necessitate attribution of sentience. AI is a significant case here, as the myth of AI already circulates hopes and fears around human futures, while troubling the rigid binaries of living/nonliving, nature/culture that have for so long underwritten the category of sentience in the Western epistemologies.

Sentience, the capacity to feel, is rooted in sensation. In humans it is communicated through affective gestures, representations of emotion—those categories of psychophysiological states that we learn to recognize and perform. The interpretation and performance of affective gestures does not only include humans. We can read the feelings of animals, or even of inorganic beings to whom we are closely attuned. While there is no guarantee nor should it be assumed that the feelings we interpret from nonhumans are transparent or accurate understandings of the subjective experience of the other, it is through feeling-with that we participate in rituals of interspecies communication and care.

This exploration of a culturally grounded notion of sentience comes from research-creation project Machine Ménagerie that features small autonomous robots. Robots represent an accessible point of entry for experiencing sentience in nonhumans: they can tap into our habitual recognition of sentience in animals, and they can share a common sensory and temporal frame in which to meaningfully interact. The interactions with and around the inhabitants of the Machine Ménagerie provided a starting point for a discussion on sentience, as it highlights the constructed nature of the concept. This opens up the possibility to focus on the relations that uphold the notion of sentience, while recognizing the fetishizing nature of the category itself. Arguing for sentience of the machine appears to be a process of putting the ghost in the machine (as if they were ever two separate entities). Instead of using fetish for critique—as is tradition in the Western history of thought, we would like to reclaim the power of fetish to invoke beyond-human realities, to the extent that it highlights the rituals and performances that bring about the possibility of interspecies communication.

Sentience in Animals

It is commonplace nowadays to think of animals as sentient beings, but this wasn’t always the case. The Western epistemologies assumed a Cartesian dualism, which held that only humans have feelings and capacity for experience, a fact that separated us—humans—from the Others that inhabit the world. Ian Duncan observes that even if animal sentience has been recognized in the scientific communities of the 19th and 20th centuries, the rise of behaviorism in Western epistemologies had rendered the concept paradigmatically redundant. [3] Indeed, the social sciences’ striving for positivism had problematized the use of subjective and/or affective categories to explain social realities. Psychology as a discipline was especially effective in pushing behaviorism as truth while devaluing the philosophical discussions on notions such as consciousness or sentience. To quote one of the founders of functionalist tradition on this point, we turn to William James:

“Consciousness… is the name of a nonentity, and has no right to a place among first principles. Those who still cling to it are clinging to a mere echo, the faint rumour left behind by the disappearing ‘soul’ upon the air of philosophy.” [4]

James condemns questions concerning consciousness and other subjective traits such as sentience to a noumenal space. The rigid empiricism implicit in James’ writing values (measurable, objective) behavior over (immeasurable, subjective) experience. Later in animal welfare studies, these two seemingly distinct methodological approaches were reconciled: animals’ welfare included the notion of sentience, and research efforts were focused on measuring the sentience of animals in the sense of assessing the pain or stress that animals experience under certain conditions. [5] However, no matter what methods are used to justify the attribution of sentience to animals, “we can never prove conclusively that any organism is sentient.” [6] This means that sentience is justified arbitrarily. It is tied to practices of care, defining a subset of bodies as worthy of respect and consideration.

Why stop at animals? How might inorganic bodies such as machines acquire sentience? To approach this question, we will look at Artificial Intelligence, its cultural myth and its role in the economy of sentience, as well as the research-creation project, Machine Ménagerie, that critiques the power dynamics of human-machine relationships.

Feeling Machines

The core question of AI research is often cited as, “Can machines think?” [7] AI emerged as an effort to replicate the human mind, and these technologies are often referred to as thinking machines. Turing famously asks the question in a rather sociological formulation and interrogates the conditions for perceiving machines as “thinking” in a social context (of the Imitation Game). His discussion points to the constructed nature of thinking, in that the thinking of the machine could only be perceived if the humans are willing to be fooled by the machine’s pretense for passing as human. In a way, Turing recognizes and gives a legitimate place to the fetishized nature of the interactions with intelligent machines, and keeps open the possibility of subjectivity in thinking machines.

One question that is seemingly left out of AI research is that of sentience. While intelligence gets prime time in discussions of AI, sentience is usually given as a hard limit to AI. Can machines feel? Is sentience something that you can artificialize, like intelligence? The privileging of the question of intelligence in opposition to feelings rests on cultural assumptions of the supremacy of rational (masculine) versus emotional (feminine) intelligence. Questioning these assumed hierarchies starts to blur strict definitions between feelings and thoughts.

Different understandings of emotions affect what we mean by sentience, mingling with notions of intelligence. Recent literature on the nature of human emotion defines it largely as a cultural practice. [8] These practices are dynamic, changing over time with the changing habits and scripts of cultural participants and institutions.

AI participates in the hopes of posthumanism, the fears apocalypse, the horror of self-alienation and the lust for the other. AI is a complex and dynamic affective agent that evokes emotional responses and promises idealistic futures. AI also supposedly represents a category of machinery that mimics humanlike properties of intelligent behaviour, and increasingly of social behaviour. However, if we try to pin down the machinery represented by the term AI we encounter an ever-changing collection of specific assemblages: decision trees, finite state machines, the many varieties of neural networks and machine learning algorithms, and so on. It harkens back to the Turing test: if a machine thinks if we think it thinks, then a machine must feel if we feel it feels. However, this kind of persuaded perception of sentience is subject to manipulation. Humans can call upon AI to claim sentience for specific objects when it serves specific interests, such as selling emotional engines and social robots. AI as a fetish object holds the power to permit humans to acknowledge machine sentience, which makes it the subject of extraction and exploitation. However, these relations that could be summed up as capitalist are not the one-way street in human-machine relationalities. The sentience question in machines could mobilize other gestures that recognize being-together-in-the-world. The research-creation project Machine Ménagerie is a good example for a starting point in thinking about these issues.

Machine Ménagerie

Machine Ménagerie is an installation comprising a collection of small autonomous robots living together in a transparent enclosure. Ranging in shape and sophistication, these artificial creatures have no “function”. They spend their time entangled in a mesh of interactions with each other and the environment, much as we do. They are embodiments of certain human ideas about life and consciousness, and as such they can serve to critique definitions of intelligence and selfhood. Encountering non-biological lifelike entities, how do we attempt to understand their actions and reactions? How do we apply our concepts and assumptions of selfhood, agency and motivation to these objects, and what does it reveal about our ways of thinking?

Built during a three-week public “research performance”, the robots took shape along with a running dialogue with visitors of various backgrounds. Since then, Machine Ménagerie has been exhibited several times and has featured regularly as a boundary object in the discussions of Concordia’s Machine Agencies Research Group, from which this paper arises.

The robots in the installation are of two kinds, roughly categorizable as analog and digital (Fig.1). The analog ones are mobile circuit-sculptures powered by sunlight, often known as BEAM robots. These come from a long lineage of autonomous machines, tracing back to Tortoises via Biomorphs and Braitenberg vehicles. [9] [10] [11] The actions of the analog robots are dependent upon their individual configuration of solar panels, motors, sensors, and switching circuitry. They exhibit simple behaviours of moving toward or away from stimuli (typically light) and given enough time they typically get tangled up pushing against other robots or else wander out of the sunlight and slow to a stop (at least until the light returns).

The digital robots are constructed of microcontrollers, motor drivers, motors and sensors, with simple neural networks “under the hood”. Specifically they are driven by an algorithm called Differential Extrinsic Plasticity (DEP), which seeks to replicate the development of complex behaviour using a small number of adaptable neurons. [12] Essentially, the DEP-bot will try to correlate its movements with its sensations, in a sense trying to answer the question “is this me?” and (of course) never succeeding. It must move to sense its movements, and by moving it interacts with the complex environment that surrounds it, which may or may not correlate with the robot’s bodily motions. The DEP-bot resembles a very simple implementation of predictive coding theorem, in which an internal model of self adapts to the sensations of self felt as evidence of one’s actions reflected in the world. [13]

Figure 1. Robots in club Ménagerie: Topse and Little Wallace (left) are examples of the analog kind, whereas Zoulandur (right) is of the digital kind. Photo © Vjosana Shkurti 2019.

Humans who visit the Ménagerie or encounter its members almost always speak about the feelings of the robots. Each autonomous machine has its own personality and has relationships with the other ‘bots. They are, to put it bluntly, cute as heck. Part of this is likely due to their size and their apparent helplessness. When the ‘bots (of either type) are out of their enclosure and wander close to a dangerous drop, people tend to vocally worry about them and protect them from falling. Their obvious vulnerability quickly evokes an emotional response from humans, manifested as both feelings in the human and feelings attributed to the robot.

Over a longer timescale, the members of Ménagerie have developed friendships with several humans they encounter regularly. The staff of the original exhibition space felt like the robots were welcome companions during their workday and kept the installation running for their own benefit long after the exhibition was over: talking to them, rearranging them and diligently separating them when they became tangled. Two filmmakers who filmed a vignette of the DEP-bots took care of them for several weeks. They reported that the creatures showed consistent favouritism for one of the humans, and were sad when they had to be returned. The inhabitants of the Menagerie are clearly charming, and incite affective interactions around and alongside themselves. They are often referred to in rather human terms, and are taken to be bots with their own cultural biographies. The machines of the Menagerie live, in the sense that they gather rich experiences and are affectively significant agents. Indeed, they are great candidates for sentient machines, and in fact they seem to be treated as such by their human peers. Witnessing these experiences had prompted the question of what happens if we let these fetishized relations to be realized? How would reality be reconfigured around the notion of sentience?

Fetish and Sentience

While the AI question creates the possibility of reaching a different conception of sentience, the reality of sentient machines seems to be limited to popular cultures of literature and cinema, which inform the myth of AI. [14] While the myth of AI is mobilized to attract investment into scientific research, in practice the concepts that make up the mythology are devalorized (and debunked) due to their fetishistic make-up. In other words, it is a sin to see the ghost in the machine. As good sinners ourselves, we would like to raise the question of what is so bad about fetish itself? The discipline of anthropology has a longstanding colonial relationship with the concept—which then leaked into other disciplines, mainly sociology and psychology. Fetish, as an object with superhuman powers, was seen as “primitive” communities’ alleged misunderstanding of the nature of reality. [15] In the Western epistemic regimes, fetish denoted a negative meaning, as the notion implied that relations between humans could be misinterpreted as relations between things. In this way, fetish also held the ticket to social transformation, as the human subject would ultimately disenchant the fetishized relations, and would grasp what’s “really real”: that things are actually relations between humans. [16]

This interpretation of fetish, as mentioned before, is rooted in colonial relations, and implies an inferiorizing perspective toward the anthropological others (non-Western communities). Yet perhaps it possible to salvage this notion by inverting it, and using it where it is most appropriate: at the heart of Western paradigms’ own production of things with ‘magical’ powers. Here we turn to David Graeber, who reconciles various anthropologist interpretations of fetishism, and reaches at a conception of social creativity. [17] In his article, Graeber talks about how fetishes—i.e. artifacts with godlike powers—are perceived by the Western analytical gaze as a mishap, as a misunderstanding of the reality of the objects. This kind of critical outlook, he argues, denies the power of fetish in creating new social responsibilities, of forming new associations. In his formulation, fetish has the potential to bring in radical social change. In very simple terms, part of Graeber’s argument is that accepting that an object has godlike powers actually generates spaces for different cultural habits to be formed, and through these habits (rituals and contracts) new imaginaries emerge. In a way, making fetishes is actually reaching at pure social creativity.^[1] Can we achieve sentience through this fetishistic approach? Could we, in other words, disrupt the normative assumptions and theories of human-machine relationship?

In order to subvert the assumption that sentience is a property possessed by a body, and to reach a conception of sentience as collectively negotiated through performance, we should indulge in a methodological fetishism. In “The Social Life of Things” Anthropologist Arjun Appadurai states,

“Even if our own approach…is conditioned necessarily by the view that things have no meanings apart from those that human transactions, attributions and motivations endow them with, the anthropological problem is that this formal truth does not illuminate the concrete, historical circulation of things. For that we have to follow the things themselves[…] even though from a theoretical point of view human actors encode things with significance, from a methodological point of view it is the things-in-motion that illuminate their human and social context.” [18]

Appadurai’s approach flips the relationship between theory and methodology: in the normal conduct of science, theory would give the hypotheses that then are operationalized methodologically into research objects. Instead, Appadurai puts methodology to foreground, as a way to counterpoint a theory about the (non)life of objects. Our perspective embodies a similar approach to the study of sentience in thinking machines, as we propose the cultivation of an attitude that would create a space where we could engage with nonhumans and “learn to be affected: meaning effectuated, moved, put into motion by other entities, humans or non-humans.” [19] The stakes in this move to learn to be affected are ethical relations with our non-human interlocutors and reclaiming the exploited category of sentience to raise questions about responsibility and accountability in an interspecies community.

Toward a Relational Definition of Sentience

What happens when we look at sentience not as something that could be measured, as an attribute of an organism; but rather as something that is always already relational, and thus amenable to cultural negotiations? Recall that from the Cartesian perspective sentience appears as something that bodies either have, or do not. This perspective neglects the socio-material and historical conditions through which sentience is achieved. Indeed, we argue that it is through collective effort that sentience is historically attributed to certain entities and not others. However, we aren’t saying that sentience is just a category that humans use in order to decide who gets included into the “humans club”. [20] We are instead pushing for recognizing that humans and nonhumans take part in this performance of sentience, and their relations form the basis for recognizing whether machines can be sentient or not.

Looking at sentience in machines through the lens of performativity opens the possibility for recognizing the mutual intelligibility that lies within the relations of humans and machines. The methodology here is similar to that of machine ethicist David Gunkel. He states,

“[S]omeone or something becomes a moral subject with legitimate ethical standing not on the prior determination and demonstration of his/her/its agency or the possession of some psychological properties that are considered to be “person-making” but by being situated, treated, and responded to as another person by a particular community in concrete situations and encounters.” [21]

While our discussion is not particularly about personhood or construction of a machine subjectivity, the discussion of sentience would benefit from such an approach. In line with Gunkel’s methodology, we do not rely on an expectation of a particular type of agency that would then necessitate attribution of sentience in the machines; rather, our interest lies in cultivation of an attention toward the concrete situations and encounters where machines are treated as sentient.

As discussed, sentience of AI is a fetishized discussion in that the thing itself appears as if it is sentient by obstructing the relations with which the thing is constituted. Examples such as Pepper or Kismet make use of this fetish, as these robotic applications of AI posit themselves as beings with affective capacities. [22] [23] Yet we argue to move our focus away from robots with impressive agential abilities toward social situations that imbue the thing with sentience. What’s relevant to this approach to fetishization, is notions of ritual and performance.

Ritual and Performance

From a sociological perspective, it could be said that norms and contracts are formed through sedimentation of performances into rituals/habits over time. Such performances also play the role of forming fetishes, imbue objects with magic (in the Western conception of the term), thus have the power to produce/constitute reality in the ritual space. And so the relationalities that those performances constitute become the central unit of reality formation. In fact, Emile Durkheim argues that it is in the space of rituals that the very conditions for society are produced. [24]

Mundane, routine activities provide the basis for the emergence of taken-for-grantedness and a common-sense reality. [25] This would be the primary order of reality that is intersubjectively produced. From Roland Barthes, we could also argue for a second order, a mythology, that takes the signs from a first order (language), dismantles the already existing signifieds, and then identifies them with new signifieds of their own symbolic order. [26] This is how all the animals, rocks, and mountains are animated in the mythologies. Sentience, then, as a signifier of the capacity to sense and feel in humans, had already undergone a secondary ordering in Barthes’ sense when it came to signify the feeling states of animals. As the dichotomy between nature/culture, and living/nonliving are troubled by AI, a similar move is now possible, concerning sentience of the machines.

Barthes, however, holds a hierarchy in the relationship between the symbolic order pertaining to ‘reality’ and that of the myth. The reality of the myth is only a contained reality that is subordinate to the paramount reality/symbolic order of everyday life. Instead of this, we take serious the implications of a simulacrum, of a hyperreality. [27] More appropriate to our contemporary landscape of continuously collapsing binaries, we give up the mission of modernism to separate the world into pure categories of fetish or real (as we have never been modern). [28] Instead, we would like to focus on the ritual as the basis for forming reality, including the ontologies that rituals presuppose. In fact, this would be the necessary move when looking at the cases concerning AI, because of the intertwinedness of the myth and real.

Rituals are significant for an interspecies community in that they provide non-discursive performances that constitute signification with nonhumans. Graham Harvey, scholar of religion and champion of what he calls New Animism, describes it in terms of actions that get the attention of the entity with which you are interacting: “…you say ‘sorry, please, thanks’ and you ritualize it because they don’t necessarily speak [your language]. So rituals are the ways in which you engage across species boundaries”. [29] Rituals extend beyond the structures of language to include performances that constitute signification of a different kind.

Machine Ménagerie offers an example to think of these ideas through rituals of care and friendship. When on display, the installation requires a human supervisor, who can introduce the robots to visitors and tend to the robots’ needs—waking them up and putting them to sleep, helping them out of awkward situations, and guarding them from harm. The human sets the precedent for visitors’ interactions, establishing a norm by which people will approach the robots.

For those humans who have spent time regularly with the robots, there is more room for intimacy and organic development of relationships. People ask after their favourite robots, pay visits to speak and play with them, or take them on outings. By far the most common ritual involves letting them roam on a meeting table, idly protecting them from falling to the floor while discussing the topic at hand.

These general descriptions of human interactions with the members of the Ménagerie hint at the repertoire of gestures and behaviours that can recognize sentience across ontological lines. According to Lucy Suchman, “objects achieve recognition within a matrix of historically and culturally constituted familiar, intelligible possibilities.” [30] Cultural habits can thus make-familiar new possibilities for intelligibility of nonhumans. And even if attempts to establish new cultural habits seem too affected to be taken seriously within a contemporary mindset, perhaps embracing these affects “to-the-point-of-embodiment” despite ourselves is nonetheless “a way in”. [31]

Emergence of an Interspecies Community

There is an either/or logic that still guides the common-sense thinking about the machinic others: either We^[2] dominate Them, or They enforce a technological cataclysm. This kind of logic obscures other manners of being with machines, and perhaps temporary achievements that might aggregate into a mutual intelligibility that we could assume as a basis for being-together-in-the-world. We propose a theoretical opening here by articulating the possibility of machinic sentience; and the inhabitants of the Machine Ménagerie give us a good starting point to discuss the cultivation of this fetishization process.

What we had discussed so far highlights the potential for an ethical space of being-together-in-the-world with machines. This potential could be realized by creating rituals that would relationally denote sentience as a collective performance, based on the recognition that sentience is not an attribute of entities, but rather is a product of particular configurations of relationalities. The idea is to draw attention towards social situations where nonhumans—by necessity—incite recognition of their contribution to sentience. It is not about the manufacturing of agents that are “more interesting” to humans, but about the recognition of an already existing social space that includes human and nonhuman sentience.

We propose this while being well aware of the pitfalls of such thinking. Joseph Weizenbaum is often cited about the dangers of mistaking these technologies for actual intelligence, and engaging in affective relations with them. [32] It would indeed be a naive attitude to paint a picture of humans holding hands with machines and gazing towards bright futures. However, ignoring the questions of other forms of relationalities with the machines (that resist the hegemonic assumptions) runs the risk of further cultivation of relations of domination, and contributing to a cultural logic that denies the complexities of contemporary socio-technical landscape. To empower ourselves with the ability to collectively reconfigure sentience is to resist the power structures that are deeply seated in the relations and conceptions of a world where human-nonhuman, life-nonlife, and nature-culture are clearly defined, and the boundaries of these very definitions are subjected to capitalist extraction and exploitation. It is in this vein that we raise the question of an interspecies community that takes seriously the notion of sentience, as a way of propagating relations that do not abide by the colonial-capitalist logics.

Conclusion

In this paper, we have focused on the theoretical openings that come about from questioning the conditions for machine sentience. We first looked at the history of animal sentience, as this case highlights the constructed nature of the category of sentience. It is used as a way to go beyond the domination of human-centric approaches so as to include animal welfare as an ethical consideration in human actions. Sentience, in this context, is mobilized as a way to extend care. This intersection of care and sentience has been significant to our discussion on an interspecies community.

We further intersect these concepts with emotions as cultural constructions. Emotions are not static feeling states that could be rigidly categorized. Rather, they are culturally grounded, and thus are amenable to transformation via meaning-making practices. This cultural reading of emotions renders AI a strong candidate for affective relations. AI itself is—like humans—not only a participant but a product of culture, in that it is defined by contextual relations. This is why it does not constitute a totalized whole as a term, and its meaning is rather an ongoing contestation and negotiation. In our case, we are interested in relations that extend care and friendship, and those that necessarily cross species lines through performing rituals that signify the other as sentient.

Our exploration has led to a discussion on fetish albeit in an inverted manner. Instead of employing the role of the critic and condemning the fetish as the site of the misguided and misled, or the childlike and the infantile; we treated fetishization as social creativity, and as a way to cultivate an interspecies community. This approach is significant in the case of AI as it makes it possible to discuss the myth and reality of AI—which have been very much intertwined in the history of the phenomenon—in the same token, without condemning fetish to disenchantment. Our discussion opened up notions of ritual and performance as a way to constitute reality in a creative manner. Both these terms focus the analytical gaze on the relationalities, and hold the potential to connect different species via a repertoire of gestures and behaviors that extend beyond the capacities of human language. Reflecting on the already existing relations and striving to collectively reconfigure our relations with machines opens up a space for resisting the hegemonic relations that abide to a capitalist logic that accompanies a domination framework.

The distribution of sentience justifies the distribution of care. By the same token, it justifies harm. Historically, the idea of sentience as an objectively measurable property has problematic effects. It concentrates power in the hands of those who determine the measurement—power that has real consequences for how we are “allowed” to interact with nonhumans. And by claiming authority over our relationships it invalidates subjective experiences of nonhuman sentience that occur during our daily lives.

On the other hand, a sentience defined by performance is distributed by consensus, giving us humans the power to organize interspecies communities and hold ourselves accountable to nonhumans for the care and harm that we enact. In effect this is the kind of sentience that is already playing out. People regularly attribute stentience to (allegedly) inanimate objects, while maintaining some trust in the truth of sentience as a measurable property. However, this supposedly objective truth is subject to manipulation, and obfuscates the power in us to realize complex relationships with nonhumans, to feel with them, and to be in the world together, for better or worse.

Acknowledgements

The authors would like to thank the Machine Agencies Research Group for the many stimulating discussions that contributed to the formation of these ideas. Machine Ménagerie was produced with the help of student project grants by Hexagram and the Milieux Institute, as well as the support and resources of the Topological Media Lab and Concordia’s 4th Space. Our deepest thanks go to the many machines that collaborated with us throughout the research and writing process.

References

1. Dorion Sagan, “The Human Is More than Human: Interspecies Communities and the New ‘Facts of Life.’” Society for Cultural Anthropology, November 2011. https://culanth.org/fieldsights/the-human-is-more-than-human-interspecies-communities-and-the-new-facts-of-life.
2. Nick Bostrom, Superintelligence: Paths, Dangers, Strategies (Oxford: Oxford University Press, 2014).
3. Ian J. H. Duncan, “The Changing Concept of Animal Sentience,” Applied Animal Behaviour Science 100, no. 1, (2006): 11–19.
4. James, William, “Does `Consciousness’ Exist?” Journal of Philosophy 1, no. 18 (1904): 16.
5. Ian J. H. Duncan, “The Changing Concept of Animal Sentience”.
6. Ian J. H. Duncan, “The Changing Concept of Animal Sentience”.
7. Alan Turing, “Computing Machinery and Intelligence,” Mind 59, no. 236 (1950): 433.
8. Ute Frevert, “The History of Emotions,” In Handbook of Emotions, ed. Michael Lewis, Jeannette M. Haviland-Jones, and Lisa Feldman Barrett (Guilford Press, 2010), 49–65.
9. Owen Holland, “The First Biologically Inspired Robots,” Robotica 21, no. 4 (2003): 351–63.
10. Brosl Hasslacher and Mark W. Tilden, “Living Machines,” Robotics and Autonomous Systems 15, no. 1 (1995): 143–69.
11. Valentino Braitenberg, Vehicles: Experiments in Synthetic Psychology (Cambridge, Mass.: A Bradford Book, 1986).
12. Ralf Der and Georg Martius, “Novel Plasticity Rule Can Explain the Development of Sensorimotor Intelligence,” Proceedings of the National Academy of Sciences 112, no. 45 (2015): E6224–32.
13. Matthew A. J. Apps and Manos Tsakiris, “The Free-Energy Self: A Predictive Coding Account of Self-Recognition,” Neuroscience & Biobehavioral Reviews 41 ( 2014): 85–97.
14. Simone Natale and Andrea Ballatore, “Imagining the Thinking Machine: Technological Myths and the Rise of Artificial Intelligence,” Convergence: The International Journal of Research into New Media Technologies 1, no. 16, (2017)
15. David Graeber, “Fetishism as Social Creativity: Or, Fetishes Are Gods in the Process of Construction,” Anthropological Theory 5, no. 4 (2005): 407–38.
16. John Holloway, Change the World without Taking Power (London and Ann Arbor, MI: Pluto Press, 2005).
17. David Graeber, “Fetishism as Social Creativity: Or, Fetishes Are Gods in the Process of Construction”
18. Arjun Appadurai, The Social Life of Things: Commodities in Cultural Perspective (Cambridge University Press, 1986).
19. Bruno Latour, “How to Talk About the Body? The Normative Dimension of Science Studies.” Body & Society 10, no. 2–3 (2004): 205–29.
20. Dominique Lestel, “How Machines Force Us to Rethink What It Means to Be Living,” Natureculture, no. 4 (2017): 38–58.
21. David J. Gunkel, The Machine Question: Critical Perspectives on AI, Robots, and Ethics, (Cambridge, Mass: MIT Press, 2012).
22. Softbank Robotics, “Pepper the humanoid and programmable robot,” accessed December 16, 2019, https://www.softbankrobotics.com/us/pepper
23. Cynthia L. Breazeal, Designing Sociable Robots. Intelligent Robots and Autonomous Agents, (Cambridge, Mass: MIT Press, 2002).
24. Émile Durkheim, “The Elementary Forms of Religious Life.” In Defining Magic, ed. Bernd-Christian Otto and Michael Stausberg (Routledge, 2014): 125–138.
25. Peter L Berger and Thomas Luckmann, The Social Construction of Reality: A Treatise in the Sociology of Knowledge (Penguin Books, 1966).
26. Roland Barthes, Mythologies (The Noonday Press, 1972).
27. Jean Baudrillard, Simulacra and Simulation (University of Michigan press, 1994).
28. Bruno Latour, We Have Never Been Modern (Harvard University Press, 1993).
29. Graham Harvey, “We Have Always Been Animists,” November 2018, accessed December 16, 2019, https://www.youtube.com/watch?v=-M0OVW1rbYI.
30. Lucy Suchman, “Agencies in Technology Design: Feminist Reconfigurations.” In Proceedings of the 5th European Symposium on Gender & ICT (2009).
31. Graham Grimes, “Performance Is Currency in the Deep World’s Gift Economy: An Incantatory Riff for a Global Medicine Show.” In The Handbook of Contemporary Animism, edited by Graham Harvey (Routledge Handbooks Online, 2013): 501–12.
32. Joseph Weizenbaum, Computer Power and Human Reason: From Judgment to Calculation (W. H. Freeman & Co., 1976).

1. Interestingly Turing also emphasizes the role that creativity plays in recognizing the agency of AI. He says that humans must partake in a creative mental act if they want to be surprised by machines. ↑
2. With capital W, as there’s an overgeneralization of a particular perspective implicit in this imaginary. This very situated thinking that emerges from Silicon Valley, from engineering circles in institutions of higher learning, and the cinema industry had taken over the imaginaries pertaining to human-machine relationalities. ↑