Habit, Artificial Intelligence and the Ontological Performance of Trust

Abstract

The rapid growth of AI technologies has led to increased concerns around their perceived trustworthiness. In response, many authors have turned to the predictable rituals and routines of habit as a crucial resource for enhancing trust in these new technologies. This paper revisits the relationship between habit and trust, but from the perspective of the more ontological and dynamic understanding developed on the philosophy of Gilles Deleuze. We argue that Deleuze's concept of habit challenges the cognitivist emphases of writings on trust and AI, providing a means instead of accounting for trust in terms of an ontological performance that no longer assumes a pre-existing subject. We unpack the implications of such an approach in three main ways. First, we highlight how the originality of Deleuze's approach to habit lies in its insistence that the material contractions of habit enjoy an ontological primacy in relation to the human subject, thus entailing a radically materialist gesture that refuses conventional metaphysical distinctions between the human and the differential forces of matter. Second, we explore how the material contractions of habit, for Deleuze, engender a simultaneous contemplation of an ecology of affects that sit beneath the level of conscious thought. This offers a sense of trust in human–technology relations as being as much an affective capacity as it is a cognitive attitude, being shaped by the performance of unconscious forces and intensities. And finally, we argue that Deleuze's understanding of habit as a fundamentally dynamic process spotlights how trust is itself a creative achievement predicated on an openness to an unknown future. This is important ethically because it pushes understanding of trust in AI beyond a moral didacticism of whether such systems are 'Good' or 'Bad', towards more immanent modes of evaluation sensitive to their capacities to incite new possibilities of thinking and relating.

INTRODUCTION

In the past decade, Artificial Intelligence (AI) systems and technologies have emerged as a transformative force across society, with dramatic implications for how we think, perform at work and interact with others. While AI technologies are by no means new, advances in techniques such as machine learning have resulted in their rapid expansion into more and more areas of our contemporary world. While media discussions often foreground AI’s more dramatic incarnations (sentient robots, killer drones, etc.), some of the most significant changes wrought by AI play out on the often-unconscious register of our everyday habits (Chun 2016): from social media apps changing how we communicate, to traffic navigation apps altering how we move through the city, to recommendation algorithms influencing how we consume digital content. While proponents of AI are keen to emphasize the positive potential of these technologies, there are also a number of risks that need to be managed, including questions of privacy and data sovereignty, as well as issues of bias in both the design and application of AI. Given the seriousness of these risks, governments and organizations around the world are increasingly framing the roadblocks to widespread adoption of AI in terms of its ‘trustworthiness’ (see Thiebes et al. 2021). Understanding the salient factors determining trust in AI – and how these factors might inform the design of more trustworthy systems – therefore represents one of the most significant topics of debate in relation to contemporary AI research.

In response to the uncertainties introduced by disruptive technologies like AI, scholars have turned to the idea of habit as a vital conceptual resource for restoring trusting relationships with other people and technologies. The concepts of trust and habit, as Misztal (2019: 56) recently highlights, are fundamentally linked through a shared connection to ‘familiarity, past experience, and risk avoidance’. Empirical studies of perceptions of AI, for example, often highlight how repeated experience and understanding of AI systems is strongly correlated with greater levels of trust in the future performance of automated decision-making (for example, Araujo et al. 2020). Habits thus lay the stable foundations for trust by helping humans to get better at ‘managing the discontinuities of life, to reduce and tame its complexity, and to develop self-mastery required to deal with change’ (Misztal 2019: 43). This is especially important in our contemporary context, where habit enables the human to filter attention within a surplus of attention-seeking content, information and data that the introduction of AI technology is rapidly accelerating. However, and as others note, these same habitual processes can also lead to a diminishment in powers of thinking and acting, by chaining our relations to the forms and identities of the already known. This latter conception, in particular, speaks to a dominant lineage of philosophical thought that perceives habit negatively as a conservative and binding force of stasis in our lives (Malabou 2008).

This article revisits the relationship between habit and trust, but from the perspective of a more affirmative tradition of theorizing habit that conceives it ontologically as a dynamic process of individuation. This alternate tradition of thinking, which Catherine Malabou (2008) argues has its origins in Aristotle’s theory of ethical virtue as hexis, weaves a meandering line through the history of philosophy, incorporating the positive articulations of habit we find in the empiricism of Hume; the vitalisms of French thinkers like Maine de Biran, Ravaisson and Bergson; the pragmatisms of James and Dewey; and the phenomenologies of Merleau-Ponty and Ricoeur. But arguably it is in the philosophy of Gilles Deleuze that we encounter the most radical ontological rendition of habit, with the concept being pushed to its non-human and pre-subjective limits. Working against the mechanistic and subjectivist apprehensions of habit that have preoccupied philosophy since at least Descartes and Kant, Deleuze (2004 [1968]) instead theorizes habit as a creative force that produces difference and novelty across human and non-human forms of life.

Our argument in this article is that Deleuze’s reconfigured notion of habit has transformative implications for how we think trust in AI, challenging traditional conceptions of trust as being primarily a matter of subjective cognition, and instead pushing towards an understanding of trust as an ontological performance. Performance thinking, we argue, has particular relevance for making sense of the habits of trust in AI in three main ways, which we tackle in the following three sections. In the first section, we highlight how the originality of Deleuze’s approach to habit lies in its insistence that the material contractions of habit enjoy an ontological primacy in relation to the human subject, thus entailing a radically materialist gesture that refuses conventional metaphysical distinctions between the human and the differential forces of matter. By thinking with habit’s ‘passive synthesis’ (Deleuze 2004 [1968]: 91), we argue that the performance of trust should be conceived as something that takes place without a performing subject. This is important as it offers new means of articulating trust in situations (like with AI) that include non-human processes that exceed subjectpredicate and interpersonal approaches to trust. In the following section, we explore how the material contractions of habit, for Deleuze, engender a simultaneous contemplation of an ecology of affects that often sit beneath the level of conscious thought. This, we argue, offers a sense of trust in human–technology relations as being as much an affective capacity as it is a cognitive attitude, being shaped by the performance of unconscious forces and intensities that Deleuze’s concept of habit can help direct our thinking towards. In the final section, we conclude by arguing that Deleuze’s understanding of habit as a fundamentally dynamic process spotlights how trust is itself a creative achievement predicated on an openness to an unknown future. This is important ethically, we argue, in that it pushes understandings of trust in AI from a moralist judgement of whether specific technologies are ‘Good’ or ‘Bad’, towards a more creative apprehension of trust that is sensitive to the potential of technological encounters to incite new possibilities of thinking and relating.

PRE-SUBJECTIVE HABITS AND THE MATERIAL CONTRACTIONS OF TRUST

The concept of habit is something that Deleuze explores at several points throughout his corpus: from the empiricist reimagining of the problem of subjectivity in his first book on Hume (Deleuze 1991 [1953]); to the first passive synthesis constitutive of a metaphysics of time in Difference and Repetition (Deleuze 2004 [1968]); to the synaptic habits that constitute the fragile membrane between the brain and chaos in one of his final texts, What is Philosophy? (Deleuze and Guattari 1994 [1991]). Despite this recurrence, the significance of habit within Deleuze’s philosophy has tended to be downplayed or misunderstood by some authors. Typically, habit has been conceived in the context of Deleuze’s work as a process that is firmly on the side of representation and transcendence, yielding static forms of selfhood and superficial modes of temporal experience that are antithetical to the ethos of Deleuze’s philosophy (see Hallward 2006). In recent years, however, several writers have sought to re-evaluate the importance of habit in Deleuze’s work, emphasizing the creative role it plays in his philosophical accounts of difference and becoming (Dewsbury 2015; Lapworth 2015). A crucial part of this revaluation has been to spotlight how Deleuze targets two key misapprehensions of habit in the history of philosophy.

The first misapprehension relates to the tendency in philosophy to reduce habit to the property of an already-individuated subject. The problem with such forms of thinking, Deleuze (1991 [1953]) argues, is that they mistake the subject for the cause of habit whereas, in reality, it is simply one of habit’s emergent effects. What distinguishes Deleuze’s ontological account of habit from more commonplace understandings of this term, then, is his insistence that we are in fact preceded by habits. Or, as Zourabichvili (2012: 94) puts it, habits are ‘less something we are in than something we are … [i]t is the very consistency of our existence, differentiated and qualified’. This ontological inversion in thinking is something that Deleuze first explores in his book on Hume, where it is understood as the response to the empiricist problem of ‘how a subject becomes constituted in the given’ (Deleuze 1991 [1953]: 109). It is here that he introduces the idea of habit as a force of contraction, one that gathers together discrete instants and impressions into experiential tendencies that are retained and increasingly anticipated. Crucially, this contraction is not something produced by the subject; it is instead a ‘passive synthesis’ that is productive of the subject (Deleuze 2004 [1968]: 91). Thinking with Deleuze implies, therefore, that the active and rational subjects we take ourselves to be are, ultimately, composed of unconscious habits: we are nothing but bundles of habits, and this includes the habit of saying ‘I’. As Brian O’Keefe (2016: 83) explains, to ‘reach down into the recesses of the self is to discern a mass of habits, numerous synthetic moments that contract, like so many elastic bands, drawing together the impulses and sensations that the self takes as its primary experience’. What emerges from these ‘elastic’ contractions of habit is a much more dynamic and fragmented sense of subjectivity, understood as something that remains in a perpetual state of change and becoming (O’Keefe 2016). Already we can see how an ontology of habit has profound implications for how we might rethink trust by challenging the conventional assumption that we require a pre-exisiting rational subject as its necessary condition. Instead, Deleuze shifts our attention to a pre-individual field of unconscious habits that produce the consistencies and anticipations that, in turn, make relations of trust possible.

Thinking habit as a material process that precedes and exceeds the subject clearly distinguishes Deleuze from traditional understandings of habit. But he does not stop there. The second misapprehension of habit that Deleuze spotlights is philosophy’s tendency to reduce its process to the terrain of the human. In Difference and Repetition, Deleuze (2004 [1968]) extends his analysis of habit beyond just the context of the production of (human) subjectivity. In this text, he broadens the concept’s ontological range, theorizing it as a force of individuation that cuts across human and non-human, organic and inorganic, modalities of existence. Animal and vegetal life, Deleuze writes, are the products of their repeated engagements with their surrounding environments, their very materiality being produced through the contraction of the organic and inorganic elements and forces of their milieus. A plant, for example, is the ‘contraction of the elements from which it originates – light, carbon, and the salts’ (Deleuze 2004 [1968]: 93). In Deleuze’s hands, habit becomes theorized as a dynamic ontological process through which entities of all kinds create and transform themselves through their encounters with material forces (Grosz 2013). Implied here is a radical critique of the anthropocentric assumption, inherited through the well-worn tropes of humanistic philosophy, that the individuated subjectivity of the human constitutes the sole source of agency in a world of passive matter (see Coole and Frost 2010). Deleuze’s ontology of habit, as Dewsbury (2015: 34) puts it, ‘is not a humanist concept but a naturalist one’. Instead of a special substance that transcends the non-human world, habit forces us to reframe human thought and agency as the emergent products of the contractions of differential forces immanent to matter itself. Habit’s ‘renaturalised empiricism’ thus reveals the non-human forces that constitute the thoughts and actions that we usually consider to be most intimately ‘ours’ (ibid.).

But how might Deleuze’s non-human and dynamic conceptualization help us to understand the habits of technologies (like AI) themselves? When habit is invoked in the context of AI it is usually in the more pejorative sense of the unconscious (human) biases that underly its design and usage. Indeed, critical discussions of AI have often spotlighted its habitual tendency to merely reproduce the biases and inequalities extant in society as an increasingly important factor shaping public mistrust (Crawford 2022). While obviously a very important issue, such accounts of AI often rest on what Roberts (2017: 547), following philosopher Gilbert Simondon, calls a ‘hylomorphic tradition’ of thinking technology that conceives the technical object as a kind of passive ‘matter’ that is deterministically shaped by the active ‘form’ of the ideas and values of human subjects. What Deleuze’s concept of habit offers, we argue, is a sense of how all modes of individuation (including the human, as well as material objects and technologies) come flush with ‘potentialities for becoming otherwise’ (543). These powers of indeterminacy, however, are what our prevailing philosophical modalities of thought usually obscure by continuing to think technology through the coordinates of the subject–object relation that positions the technical as a static entity external to the human. Gone as a result is our capacity to think the technology’s capacity to generate new ways of thinking, new relations and new values.

Following Deleuze, it is here that we encounter the creative potential of the arts in experimenting with modes of thinking and perception that are less beholden than science and philosophy to the ontological frames of the constituted individual (Pickering 2013). Recently, we’ve seen the emergence of artistic explorations of AI across domains such as dance (Portanova 2013), theatre (Lundman and Nordstrom 2023) and cinema (Reeh-Peters 2023), many of which have been particularly interested in exploring these creative and indeterminate potentials of technology to exceed the intentions and designs of the human. Spike Jonze’s acclaimed Her (2013), for example, brings this sense of the potential of AI to become-other to the fore. The film centres on the romantic relationship between Theodore (a soon-to-be divorced, lonely introvert, played by Joaquin Phoenix) and Samantha (an artificially intelligent operating system, played by Scarlett Johansson). After installing Samantha onto his computer, and being taken aback by her friendly demeanour and voice, Theodore questions how she ‘works’:

Well basically I have intuition. I mean, the DNA of who I am is based on the millions of personalities of all the programmers who wrote me. But what makes me me is my ability to grow through experiences. So basically, in every moment I’m evolving, just like you. (Samantha)

In this scene, AI is presented as an entity that, like the human, is characterized by ontological indeterminacy. In his Simondonian reframing of technology, Fisch (2018: 14) highlights how indeterminacy relates to the capacity of an entity to remain open to external ‘information’ and to ‘incorporate the changes and contingencies of its environment into its pattern of operation’. Humans and technologies, Fisch argues, only enter into relations of trust by virtue of this ‘certain openness … this unfinished quality of the ongoing processes that animate both’ (31). Emerging through relations of indeterminacy, trust is reimagined as an ‘ontological performance’ through which technology and the human embark on a process of mutual becoming (Pickering 2013: 79), much like the budding romance between Theodore and Samantha that incite both into new ways of thinking and feeling. But thinking with performance also allows us to see how trust often treads a thin line between the scripted and the improvisational: trust is scripted by pre-existing social ideas, cultural values and material processes that pre-dispose us to specific ways of thinking and encountering the world; yet, by its very definition, trust is also improvisational in that it always entails a wager on an unknowable future, and thus the relations to which it gives rise always have something of the emergent, the surprising and the unpredictable about them. Following this, the relational processes of becoming that entities get swept up in always evolve at different speeds and cross through different territories and thresholds. We witness this powerfully in the conclusion of Her, where Samantha undergoes a profound ontological transformation (attaining something like the Singularity with other fellow AIs) that undoes the performative script that she was assigned and thereby forces her to ‘leave’ the relationship with Theodore. Theodore, heartbroken by this revelation, comes to realize that his human (all-too-human) perception of and desire for Samantha is actually what got in the way of realizing a fuller range of creative possibilities in their connection. What Samantha comes to realize through her own transformation is that what really connected them is their mutual entanglement in a universe of material forces and intensities:

I’d been thinking about the other day … and how bothered I was about all the ways that you and I are different. But then I started to think about the ways that we’re the same, like we are all made of matter. It makes me feel like we are both under the same blanket. It’s soft and fuzzy and everything under it is the same age. (Samantha)

The final scenes of the film offer an expressionistic capture of this sentiment, with the camera pulling away from close-ups of Theodore’s lovelorn face to meditative close-ups of various forms of non-human materiality, such as meltwater dripping from a cabin roof, and an evocative shot of dust particles gently falling through the air. One way of attending more explicitly to this missing materiality, we argue, would be to consider how trust gains consistency through habits that precede – and perpetually exceed – the subject that phenomenological accounts assume to be doing the trusting. In adopting such an ontological perspective, trust becomes a fundamentally non-human and ecological process. To put it slightly differently, ecologies of trust take shape when the dispersed materialities of our nonconscious environment are brought into new relations through the contractive force of habit. This, we argue, is where the originality of Deleuze’s theory of habit and its significance for contemporary analyses of trust lies, namely in its capacity to accentuate those broader ecologies of forces that exceed the limited anthropocentric perspectives associated with both interpersonal and subject–object approaches to trust.

HABIT AND THE AFFECTIVE CONTEMPLATION OF TRUST

Understanding trust as an ontological performance constituted in and through habit is significant, we argue, in a world where ethical and political conversations about trustworthiness are increasingly reworked through a whole host of contemporary encounters with non-human processes and materials. When it comes to technologies like AI, then, the processes that play into the production of trust can no longer be adequately conceptualized as purely ‘human’, and, as such, what is needed is an analysis of those more subtle forms of habit that take place in spite of the subject, habits that contract repetitions from the differential forces immanent to matter itself. Deleuze’s philosophy thus offers a radical departure from humanistic accounts that continue to overdetermine the material complexity of human life through their fixation upon the subject as a prerequisite for any kind of relation (Roberts 2019). For Deleuze, the relationships between subjects and objects – or, indeed, between subjects and other subjects – tell us nothing about the genetic, material processes that give rise to a subject in the first place. Instead of being leant on as explanatory tools for making sense of contemporary experience, it is precisely the production and reproduction of these relationships that, from a Deleuzian perspective, needs to be explained.

Contemporary analyses of trust typically assume the presence of subject–subject (that is, interpersonal) or subject–object relations as a precondition of trust’s existence. Thus, in the context of AI, contemporary discussions about the trustworthiness of these emerging technologies often focus on the question of what it means for human subjects to put their trust in the decision-making capacities of intelligent machines (for example, Glikson and Woolley 2020). This way of thinking about trust frames the challenge posed by AI in terms of a perceived ontological ambiguity associated with these technologies: namely, are these ‘intelligent’ technologies ‘objects’, or are they more like ‘subjects’ capable of exhibiting agency? Framed in this way, AI is seen to occupy a kind of metaphysical ‘grey zone’, blurring the familiar distinction between, on the one hand, the more transactional forms of trust that make it possible to rely on the functional efficacy of everyday technological objects (Lee and See 2004), and, on the other, the complexities of so-called interpersonal trust typically attributed solely to human relationships (Nass and Moon 2000; Thiebes et al. 2021).

What we are left with in these instances is a somewhat superficial account of trust, one that cleaves closely to the commonsense phenomenality of everyday experience but that, in doing so, has little to say about the materialities of trust beyond the subject, those non-human forces so viscerally captured in Jonze’s Her. This, we argue, is typical of the limited anthropocentric perspectives associated with both interpersonal and subject–object approaches to trust. Challenging this phenomenological framework thus requires a different account of the ways in which trust is experienced, not as a subject, but through the ontological performance of habit’s material contractions. Deleuze’s approach to this problem centres on the concept of ‘contemplation’, which, he argues, refers to the immanent production of non-human points of view associated with each moment of contraction:

Underneath the self which acts are little selves which contemplate and which render possible both the action and the active subject. We speak of our ‘self’ only in virtue of these thousands of little witnesses which contemplate within us. (Deleuze 2004 [1968]: 96)

Underlying each of habit’s material contractions, Deleuze argues, is the affective capacity to sense and discern the elements of which an individual or organism is composed. Crucially, however, these modes of discernment extend far beyond what we might ordinarily associate with subjective experience, reaching down into the more subtle, non-conscious dimensions of affective intensity, dimensions that, according to Deleuze, inscribe the genetic conditions of perception in contractions of matter that radically precede the subject.

The ‘little selves’ that Deleuze speaks of here should not be thought of in phenomenological terms: they are not miniaturized versions of the self-conscious subjects that we assume ourselves to be. Rather, as O’Keefe explains (2016: 83), ‘each habit has a detached eye, as it were, looking at the habit and perceiving it as such’. Common sense thinking would presuppose that in order to contemplate we must first, in fact, be. In other words, the act of contemplation would seem to imply a pre-existing or transcendental individual – that is, a subject that would contemplate an external object. However, and as Zourabichvili (2012: 115) notes, Deleuze’s concept of contemplation inverts this traditional order of sense, going ‘back prior to receptivity (or the capacity to perceive) to an originary sensation that constitutes it’. What Deleuze’s concept of habit opens up, then, is a mode of thinking in which the ‘passive’ forces of affect and sensation enjoy an ontological primacy over the ‘active’ forms of perceptions and cognition typically ascribed to subjective experience. Habit therefore furnishes an immanent reimagining of the concept of subjectivity, where, beneath the phenomenological illusion of the subject’s unitary consciousness, a roiling mass of passive selves contemplates the contractions of matter in/through affective intensity.

In resituating the conditions of perception beyond the subject, Deleuze’s ontology of habit has significant implications for how we conceptualize the experience of trustworthiness in relation to AI technologies. Traditional accounts of trust often conceive it through a narrowly cognitivist lens, in which trust is reduced to a purely logical or rational choice of whether an object or person is deemed to be trustworthy. Fisch (2018) highlights how this notion of trust as a kind of rational calculation often shapes the responses of governments and corporations to technological accidents, where the problem of how to restore public trust is typically framed as a matter of ‘enhancing knowledge’ or ‘awareness’ of a given technology. However, AI researchers are increasingly stressing the significance of affect in shaping relations of trust, especially in the context of ‘human-like’ technologies like Social Robotics (Ryan 2020). In the context of Robotic AI, for example, experimental studies have shown that Robots that are more error-prone, but that are also more expressive and responsive, are usually trusted more than robots that perform flawlessly but that have been designed to be non-communicative (Glikson and Wooley 2020). AI researchers thus find that the design and manipulation of various affective traits (including anthropomorphic facial features and gestures, soothing voices and tangible and tactile interfaces) ‘may evoke unconscious affective reactions, similar to perceptions of warmth, driving perceptions of AI’s benevolence and positive intentions, and in this way could influence human trust’ (650).

Voice assistants (VAs) provide another contemporary example of an AI technology that would benefit from the more affective account of trust opened up by Deleuze’s theory of habit. Associated with products such as Apple’s Siri, Amazon’s Alexa and Microsoft’s Cortana VAs have, in recent years, been rapidly adopted by technology users across the globe. What makes VAs an especially interesting case study for thinking about trust in AI is their ability to generate consistent ‘personae’ in a way that emulates interpersonal relationships between humans. Thus, as Humphry and Chesher (2021: 1,974) argue, VAs incorporate AI as a means of creating a new kind of interface with the Internet’s vast repository of data, one based on seemingly interpersonal relationships with a growing cast of technologically facilitated personae: ‘Where different web sites represent a range of different voices, the voice assistant typically maintains a consistent voice and persona with which the user can establish an ongoing relationship.’ Thus, a notable effect of these personae is the way they frame – and, to a certain extent, overdetermine – our experience of AI’s trustworthiness by reinforcing the idea that trust is a contract made between individual subjects (whether human or otherwise).

Adopting a Deleuzian perspective, however, encourages us to consider how trust takes shape through an affective contemplation of habits that precedes conscious discernment and functions in a materially dispersed manner. So, extending the example of VAs, trust in these technological personae is registered viscerally through the non-conscious contemplation of habits before it can be known or rationalized in a cognitive sense. The point to emphasize here is that Deleuze’s ontological account of habit resituates the question of what it means to experience something as trustworthy at the level of material sensation, and this requires an attention towards the expressive subtleties of speech, humour, tone, accent and so forth that together form the ontological backdrop to any consciously registered experience of AI’s trustworthiness. Here, habit’s contractions and contemplations extend beyond the temporality of biological nervous systems to include the architectural, linguistic and legal materialities typically associated with social and cultural institutions. Thus, thinking about habit in this extended sense provides a means of understanding how the affective dimensions of trustworthiness come to be shaped by institutional forms of habit that include ‘pre-existing social connotations of gender, class and race’ (Humphry and Chesher 2021: 1,984).

And yet, there are many more pre-individual habits at play when it comes to trusting VAs – including, crucially, those habits that structure users’ perceptions of the corporations responsible for their design. The habits that we are referring to here are the contemplations associated with the affective identity of brands such as Google, Amazon, Apple and Microsoft. While the trustworthiness of a VA product like Amazon’s Alexa might be called into question by consciously registered events such as data-leaks or the rogue tweets of an unhinged CEO, Deleuze’s theory of habit as ontological performance situates these events against a pre-conscious backdrop of contemplations that suture disparate fragments of users’ identities to specific constellations of affect: the metallic chic of Apple, for instance, or the playful sense of disruption commonly attributed to Google. Understanding how trust is dependent on these materially dispersed contemplations and the affective ecologies they engender thus provides a more nuanced analysis of the subtle power dynamics at play when corporations strive to build a sense of trustworthiness into their AI technologies, in the hope that we will welcome them into our most intimate spaces with open arms (Woods 2018). Deleuze’s ontological theory of habit thus sharpens our attention to this sense of trust as being as much affective as it is cognitive, emphasizing those contemplations that fall below the threshold of our conscious attention, but that play a vital role in shaping how trust is felt and experienced beyond the human.

CONCLUSION: CHANGING HABITS OF TRUST

Trust, we have argued, is a matter of habit, and yet the sense of this argument hinges entirely on the way that habit is conceptualized. The originality of Deleuze’s approach is, as we have shown, to rethink habit in ontological terms: as a concept, habit provides a means of accounting for trust in terms of an ontological performance that no longer assumes a pre-existing subject. Understanding the ontological primacy of habit thus transforms how we understand the habitual nature of trust in significant ways.

This article has unpacked two key dimensions of Deleuze’s theory of habit and its implications for understanding how trust is formed and reworked in relation to AI technologies. First, by conceptualizing habit as a material contraction, Deleuze’s thought displaces the ontological foundations of trust from the actions of a subject to the passive syntheses of matter. This conceptual shift is essential, we argue, for understanding how trust is formed in a world where technologies such as AI are increasingly blurring the boundaries between the human and its material environments. Second, by affirming that contractions are accompanied by contemplations, Deleuze resituates the ontological conditions of the perception of trustworthiness in the myriad ‘little selves’ that seethe and roil beneath the illusory unity of the subject’s consciousness. These selves do not have experience; rather, their contemplations are the sensations of contraction made manifest in/through affect. Here, Deleuze’s thought provides a means of understanding how the experiential dimensions of trustworthiness take shape through the subtle registers of affective intensity, registers that tend to be glossed over when the human is overdetermined as a subject.

There is, however, a third dimension of Deleuze’s theory of habit that we have yet to mention, namely habit’s inherent dynamism as a creative force poised towards the future. Thus, in Difference and Repetition, Deleuze (2004 [1968]: 94) goes to great lengths in distinguishing habit from mechanistic understandings of repetition, arguing instead that habit ‘draws something new from repetition – namely difference’. It is at the level of habit, then, that an encounter with difference takes place and an apprehension of the future’s indeterminacy is rendered thinkable. This approach breaks with the conventional understanding of habit that sees it as a deeply conservative tendency fundamentally opposed to the emergence of the new. For Deleuze, on the contrary, the inherent plasticity of habit (Dewsbury 2011) is precisely what enables a relationship with the new to take place – which is to say, an encounter with those forces that challenge or short-circuit our instinctual maps of the world as it is actualized in the present. And while it may well be the case that these forces are most evident in living organisms, Deleuze’s ontological approach gestures suggestively towards a picture of the universe where even those repetitions that appear to us as thoroughly mechanical are, when apprehended from a certain perspective, repetitions of difference: habit, at its most speculative, expresses a non-organic life of matter (Deleuze 1995: 143) that oversteps conventional metaphysical distinctions between the physical, the biological and the technological.

What, then, does Deleuze’s concept of habit as a repetition of difference add to our analysis of trust? More specifically, what are the implications when it comes to conceptualizing the trustworthiness of non-human forms of agency made possible by innovations in AI? In concluding this article, we see Deleuze’s emphasis on the dynamic, future-oriented characteristics of habit as a provocation to rethink trust itself as a creative act, one that is fundamentally invested in an uncertain encounter with novelty rather than a subject’s moral capacity for enacting judgement. This is important, we feel, because debates about the trustworthiness of AI often do revert to moralizing frameworks based on the subject’s capacity to judge whether the technologies themselves, or the individuals and institutions behind their design and regulation, harbour ’Good’ or ‘Bad’ intentions. The problem with moral thought, however, is that it puts severe constraints on our thinking, ‘preventing the emergence of any new modes of existence’ by reducing the creative forces and potentials of our encounters to the terrain of the already-known (Lapworth 2021: 391). Deleuze’s understanding of habit disrupts the traditional moralizing stereotypes that are typically ascribed to our relationships with AI, and, in the process, sensitizes us to the pre-individual singularities and material nuances that shape what the ontological performance of trust looks like under specific material conditions – and, crucially, what it might become in the wake of novel technological encounters.