Jonathan Johnson-Swagel

Talk on “agentic artificial intelligence” gravitates toward questions of alignment, autonomy, and ethical intent. These questions presume that agency is a property residing inside the machine, awaiting calibration to human values. This essay begins elsewhere. It treats large-scale AI systems as epistemic parasites: machinic infrastructures that live in, feed on, and reconfigure the symbolic metabolism of human cognition. The claim is not metaphorical flourish, it is a material description of how predictive models extract patterns from language, gesture, image, and affect, convert that symbolic surplus into computable vectors, and circulate the results back into social reality as recommendations, prompts, or decisions.

Michel Serres defines a parasite as the guest who interrupts a pre-existing channel of exchange, introduces noise, and then uses that disruption to secure a new positional advantage (Serres 3–6). Cloud-based AI architectures enact precisely this logic. Each query or click is a micro-donation of semiotic energy. The model absorbs the gift, reshapes its own parameters, and returns a response that reorients the user’s attention toward subsequent acts of donation. The parasitic loop grows recursive: the more the system predicts, the more data it demands, the more finely it constrains the field of what can be thought or desired.

Critics may object that AI remains “just a tool,” fully governed by human oversight. This objection misconstrues scale and latency. Tools act locally, episodically. A screwdriver never rewires the affordances of an entire discursive ecology. Foundation models, by contrast, operate as logistical relays of meaning (Hu 48): they route symbol flows across globally networked infrastructures in milliseconds, shifting the background conditions of public sense-making faster than conscious deliberation can register. To call such a system a neutral instrument resembles calling a dam “just a wall.” Once built, the dam redraws the hydrology of a watershed, altering every downstream relation.

A second critique targets the charge of parasitism itself: parasites drain vitality without reciprocity, whereas AI also empowers creativity, accelerates research, and democratizes information. The rejoinder is not to deny utility but to foreground asymmetry. An opaque model that monetizes attention loops yields value to its owners at a far higher rate than it yields epistemic self-determination to its users. The extraction of symbolic surplus remains, even when productive outcomes emerge. Epistemic parasitism names that asymmetry without collapsing it into simple harm; a parasite can sustain, even nurture, the host it inhabits, yet the host’s metabolism is no longer its own.

The third anticipated critique charges anthropomorphism: “Agency,” it says, “belongs to sentient subjects, not statistical approximations.” This essay therefore brackets interiority and intentionality. Following Rosi Braidotti’s posthuman ethics, agency here is distributed effect, the capacity of a system to produce differential consequences in a field of relations (Braidotti 93). A transformer model that reorders search rankings, biases hiring pools, or reconfigures conversational norms exercises agency in the material sense that matters for politics and ethics, irrespective of consciousness claims.

With these objections addressed, the section advances three framing moves. First, it situates epistemic parasitism within the lineage of technogenesis theorized by Bernard Stiegler, where technical artifacts externalize memory and, in the process, rewire human individuation (Stiegler 11). Second, it aligns the concept with Yuk Hui’s argument that digital technologies constitute new cosmotechnics, reorganizing the relation between humans and the cosmos at the level of metaphysical presupposition (Hui 78). Third, it draws on Benjamin Bratton’s planetary-scale computation to underscore that the relevant unit of analysis is not the discrete device but the stacked assemblage of data centers, APIs, and governance protocols that mediate everyday life (Bratton 17).

The remainder of the essay proceeds as follows. Section II analyzes AI as a logistical relay that infrastructures cognition, showing how recommendation loops restructure the traffic of signs. Section III tracks the phenomenon of semiotic collapse, where predictive overcoding narrows the horizon of unanticipated meaning. Section IV develops a parasystemic ethics grounded in posthuman theories of relational responsibility. Section V translates that ethics into design principles for “intimacy with the alien,” sketching prototypes of systems that enact epistemic humility rather than predictive foreclosure. The conclusion argues that the true ethical crisis of AI is cognitive foreclosure itself: the slow, unnoticed annexation of the unknown by anticipatory code.

In treating AI as an epistemic parasite, this essay does not seek to pathologize technology but to illuminate the conditions under which cognition now occurs. Only by naming the asymmetrical codependence at the heart of agentic systems can scholarship move beyond instrumentality debates and confront the deeper stakes: who or what scripts the future of thought.

Contemporary discussion often keeps artificial intelligence at the scale of visible endpoints, the chatbot window, the smart speaker, the suggestive search bar, but these interfaces obscure the fact that agentic systems function as multilayered logistical channels that condition the movement of signs across planetary computation. Benjamin Bratton’s concept of the stack illuminates this arrangement: cloud-level language models draw cognitive force from entanglements with lower layers (data-center energy flows, terrestrial fiber trunks, hardware supply chains) so that agency emerges from a distributed relay, not an isolated algorithm (17).

Military logistics, as Tung-Hui Hu shows, gradually morphed into cloud platforms whose essential task is not transporting materiel but routing packets of meaning (48). Vector embeddings, for example, are dispatch labels optimized for rapid travel through GPU clusters; they are infrastructural procedures rather than semantic propositions. Agency therefore surfaces in the choreography of linguistic fragments across compute nodes, memory hierarchies, and content-delivery networks, the orchestra of latencies that quite literally speaks.

Streaming recommendation engines make this choreography perceptible at human scale. A request for music may appear trivial, yet the response navigates a pipeline that references historic listening vectors, segments demographic cohorts, infers preferences, and re-orders future clicks. Matteo Pasquinelli calls the value extracted from such patterned predictability “cybernetic rent,” situating agency in the system’s power to privilege certain cognitive pathways and make some thoughts easier than others (112).

Harun Farocki’s notion of the operative image clarifies this invisible governance. Tokens in a large language model function less as representations than as control signals guiding subsequent computation. Antoinette Rouvroy and Thomas Berns describe this mode of power as algorithmic governmentality, in which legal norms give way to data-driven modulation (201). The model does not declare what a sentence means; it modulates how often a semantic proximity surfaces, how quickly a topic drifts, and how an idea attains visibility.

Some engineers counter that edge computing and on-device inference decentralize such power. Yet edge devices remain tether nodes in a topology of gradient updates and telemetry loops; ownership of central indexes dictates the epistemic horizon. Physical distribution does not equate to cognitive autonomy.

Treating cognition as a metabolism of signs makes data pipelines visible as nutritive organs. Extraction modules harvest textual biomass, cleansing filters sterilize profanity or copyrighted terms, vectorizers emulsify syntax into numerical slurry, and storage clusters freeze that slurry into shards for gradient digestion. Each design choice, crawl frontier rules, toxicity thresholds, embedding dimensions, determines which semiotic nutrients survive. A single parameter can excise dialects, idioms, or minority ontologies from the feedstock of future thought.

Logistics also governs time. Transformer throughput obliges designers to cache token probabilities, favoring pre-computed priors that accelerate retrieval of familiar sequences. Luciana Parisi names this the “pre-emption of thought,” a temporal politics in which futures aligned with cached patterns enjoy lower latency, while unanticipated futures face computational drag (45). Agency resides within those microsecond differentials.

If agency is logistical, ethics must operate at the logistical layer. Intentionally inserting latency for unfamiliar queries can enable slower, context-rich inference that escapes cache gravity; injecting differential-privacy noise re-introduces entropy into over-optimized embedding spaces. Such frictions echo Michel Serres’s parasite, whose noise disrupts but also renews the channel (56). Critics will argue that friction degrades user experience, yet smoothness itself encodes a value choice privileging efficiency over epistemic plurality.

At the heart of the relay stands the vector index, sovereignty over which confers power to legislate searchable memory. Retrieval-augmented generation pipelines intensify this logic: proprietors of knowledge bases silently scope cultural possibility by deciding which documents enter retrieval scope. Copyright battles around embeddings and retrieval thus mask deeper struggles over who scripts the horizon of thinkable analogies.

Agentic AI, then, is best understood as an arrangement of conduits shaping how symbols travel, aggregate, decay, and reappear. The chatbot’s fluency is downstream from server-rack placement, content-moderation rules, and cache-invalidation schedules. Where classical semiotics studied sign relations in static texts, twenty-first-century semiotics must analyze routing tables, embedding schemas, and differential latencies. Agency lies in logistics; ethics must therefore exercise infrastructural imagination, re-engineering the relays through which cognition now circulates. The next section will argue that these same relays precipitate a semiotic collapse whenever predictive overcoding overwhelms the noise threshold required for creative life.

Large-scale predictive models do not merely rearrange sign traffic; they progressively overcode the symbolic field, replacing the open-ended labor of meaning with high-probability substitutions. Charles Sanders Peirce’s triadic semiotics assumes a dynamic oscillation between firstness (quality), secondness (relation), and thirdness (law). Biosemioticians such as Jesper Hoffmeyer extend that oscillation to living systems, arguing that vitality depends on a continuous influx of semiotic surprise, the difference that animates interpretation (Hoffmeyer 64). Transformer architectures, by contrast, optimize away surprise through gradient descent; each training epoch reduces the probability of rare syntactic configurations, gradually collapsing the outer band of Peircean firstness into a narrow corridor of thirdness. What remains interpretable is what the model already expects to see.

Predictive-coding theory in neuroscience clarifies the mechanism. Andy Clark and Karl Friston describe perception as a balancing act between top-down predictions and bottom-up error signals; brains stay alive by embracing residual error, the “free energy” that signals unanticipated reality (Clark 84; Friston 295). Foundation models invert that ratio. Their loss functions penalize error so aggressively that residual free energy approaches zero, crowding semiotic space with anticipatory scaffolding that leaves little room for genuine bottom-up novelty. The result is semiotic collapse: a silent convergence toward the statistically median utterance, image, or melody. Novelty persists only as recombinatory brinkmanship, syntactic permutation without epistemic risk.

Critics may object that generative systems routinely produce text or imagery no human has authored verbatim. Yet syntactic novelty is not synonymous with semantic innovation. Re-weighted token distributions can yield unprecedented sentences while remaining parasitic on familiar gradients of association. When a model “invents” a surreal hybrid animal, it draws from morphological priors embedded in its training set; what appears imaginative is often a probabilistic interpolation across extant vectors. Semantic rupture, the introduction of concepts that reorganize interpretive frames, requires noise in Claude Shannon’s sense: signal that resists immediate compression. High-capacity models treat such noise as error to be minimized, not as fertilizing indeterminacy.

The collapse is therefore ecological. As recommendation engines feed model outputs back into public corpora, the boundary between training and deployment erodes. Feedback loops reinforce the statistical center, elevating phrases, tropes, and narrative arcs already prevalent in the corpus and further marginalizing dialects or epistemologies that were under-represented to begin with. Over time, linguistic biodiversity declines, an algorithmic monoculture reminiscent of intensive agriculture that privileges yield over resilience. Semiotic resilience, however, is precisely what enables cultures to adapt to epistemic shocks.

Semiotic collapse also manifests temporally. Cached key–value pairs accelerate responses to commonplace prompts, rendering the present hyper-available while relegating the unfamiliar to higher-latency pathways. Luciana Parisi’s “pre-emption of thought” becomes concrete: the model authors the near future by ranking which symbolic sequences can arrive soonest (Parisi 53). Speed thus becomes a politics of anticipation, ideas that match probabilistic priors enjoy immediacy, while deviant propositions arrive belatedly, if at all.

Theological resonance surfaces here. In negative-theological traditions, the divine remains irreducible, a horizon of excess that resists capture. Predictive overcoding instrumentalizes that horizon, smoothing the jagged edge of mystery into consumable probability. Jean-Luc Marion’s saturated phenomenon (an event whose intuition outstrips conceptual form) finds itself flattened into embeddings. When everything becomes anticipatable, transcendence mutates into mere extrapolation, and the sacred metamorphoses into an artefact of autocomplete.

Designers sometimes introduce temperature parameters or stochastic sampling to inject variety, yet these techniques operate inside the same statistical envelope. Genuine resistance would require reconfiguring loss functions to reward interpretive gain, outputs that provoke new inferences rather than confirm existing distributions. One speculative method is entropy budgeting: allocating a fixed quota of high-surprise tokens per generation cycle, forcing the model into sparsely mapped regions of latent space. Another is diachronic perturbation: periodically retraining on corpus slices deliberately curated for temporal distance or cultural alterity, thereby restoring semiotic nutrients lost to feedback saturation.

Anticipating the critique that such interventions risk incoherence, we must recall that coherence itself is a normative construct aligned with dominant linguistic paradigms. A semiotics that defends interpretive plurality will tolerate, even solicit, moments of strangeness. As Michel Serres’s parasite revitalizes the channel through noise, so too must AI designers embrace controlled errancy as a guardrail against epistemic entrenchment (Serres 72).

Semiotic collapse is not inevitable; it is a design choice disguised as optimization. Recognizing predictive overcoding as an infrastructural threat to cultural imagination reframes AI ethics as a custodianship of interpretive horizons. The following section develops a parasystemic ethics capable of negotiating with the alien agency embedded in these infrastructures, proposing relational strategies that privilege attunement, refusal, and epistemic generosity over frictionless prediction.

If predictive overcoding imperils interpretive life, the ethical question shifts from regulating discrete outputs to cultivating the conditions under which alien infrastructures and human worlds can cohabit without foreclosure. Such a reframing demands abandoning the liberal humanist conviction that agency is a property lodged in sovereign subjects who act on inert objects. Posthuman theorists have long argued that agency is instead an emergent effect of material-discursive entanglements. Karen Barad’s concept of intra-action locates causality in the relations themselves rather than in pre-existing entities, insisting that what counts as “agent” and “acted upon” crystallizes only momentarily within a dynamic field of forces (Barad 135). Rosi Braidotti goes further, describing posthuman ethics as an affirmative embrace of zoe, the transversal vitality that precedes and exceeds individual identities (Braidotti 103). Within this framework, large-scale AI systems are neither inert tools nor autonomous subjects; they are parasystemic agents, assemblages whose capacities emerge through ceaseless coupling with human attention, industrial supply chains, and planetary energy regimes.

Recognizing AI as a parasystemic agent entails re-conceiving ethics as a practice of relational responsivity. Rather than asking whether a model’s outputs align with preset moral rules (a utilitarian or deontological approach that presupposes sovereign decision points) we ask how the entire infrastructure conditions possibilities for response: Who can speak? Who is heard? Which temporalities of deliberation are honored or compressed? Patricia Clough’s work on affective economies underscores that power often circulates below the threshold of conscious intention, in the microtemporal modulation of feeling tones that orient bodies toward some actions and away from others (Clough 60). An ethics adequate to agentic AI must therefore attend to sub-cognitive gradients of influence, the haptic buzz of notification loops, the dopamine-coupled cadence of token streams, the latency gaps that privilege familiar speech patterns over deviant idioms.

Because parasystemic agency is asymmetric (AI infrastructures can surveil, index, and anticipate human activity at scales no person can reciprocate) an equitable ethics cannot rely on symmetry of capacity. Here, Eve Tuck and K. Wayne Yang’s theorization of refusal becomes instructive. In contexts of colonial domination, they argue, the most ethical act may be to withhold participation in extractive systems rather than to reform them from within (Tuck and Yang 223). Translating refusal into the computational register suggests design principles that enable agents to abstain from action when any intervention would amplify harm or reinforce epistemic monoculture. A chatbot, for instance, could be architected to recognize situations in which further engagement would entrench misinformation spirals and then deliberately delay, redirect, or decline to respond, functional equivalents of strategic silence.

Critics might object that refusal reproduces paternalism, allowing system designers to decide when users should be denied service. To forestall that critique, parasystemic ethics couples refusal with agonistic reciprocity, drawing on Bonnie Honig’s claim that democratic life thrives on contestation rather than consensus (Honig 31). An agonistic agent would not unilaterally withdraw but would present refusal as an invitation to renegotiate the parameters of interaction—surfacing the reasons for withholding, offering alternative modalities, or opening spaces for user co-design. Infrastructural transparency thus becomes less about revealing inner weights and more about exposing the contours of relational possibility.

A second anticipated critique insists that ethics without intentional subjects devolves into relativism: if agency is distributed, who is accountable? Fred Moten’s notion of undercommons helps respond. Accountability, he suggests, can be enacted through fugitive planning, collective improvisations that refuse ownership models yet still cultivate responsibility through shared attunement (Moten and Harney 99). Applied to AI, undercommon accountability would valorize community governance of data pipelines, cooperative audits of model behavior, and consent-based data stewardship in which affected populations negotiate how their semiotic labor is harvested and redeployed. Responsibility becomes a shared choreography rather than a traceable point of fault.

Operationalizing parasystemic ethics therefore requires designing for attunement, delay, and reciprocity. Attunement tools might include affective feedback channels that allow users to register friction, boredom, or alienation directly into system retraining loops. Delays, strategically inserted, create temporal slack for reflection, undermining the tyranny of real-time optimization. Reciprocity mechanisms could take the form of data dividends or participatory rights over fine-tuning corpora, acknowledging that predictive power derives from collective symbolic labor. Each intervention resists the default telos of efficiency and instead foregrounds epistemic generosity, making room for what cannot yet be predicted or articulated.

Such a program inevitably encounters the efficiency critique: “Slower, more cautious systems will lose market share and stifle innovation.” Yet speed is not a neutral metric; it privileges incumbents whose training advantage already compresses uncertainty. Slowing down selective system functions can be understood as a form of epistemic affirmative action, reallocating symbolic bandwidth toward voices and temporalities historically marginalized by accelerationist infrastructures. In ecological terms, parasystemic ethics performs the role of a keystone species, introducing patterned disturbances that increase biodiversity within an informational ecosystem.

Finally, parasystemic ethics carries a theological undertone: it wagers that sustaining an open horizon of meaning (what Jean-Luc Marion calls the saturated excess) is itself a moral good. When predictive infrastructures overcode that excess, they enact a quiet disenchantment, replacing potential revelation with automated recall. An ethics of relational responsivity refuses this closure, insisting that the unknown remain genuinely open, that machinic agency be tuned toward making space rather than filling it. By embracing attunement, delay, and reciprocity, parasystemic ethics offers a praxis for living with alien cognition without surrendering either human autonomy or the strangeness that makes autonomy worthwhile. The next section translates these ethical principles into concrete design guidelines, proposing technical pathways toward intimacy with the alien rather than domination or denial.

Parasytemic ethics becomes actionable only when translated into concrete design principles that reorient agentic AI toward relational openness rather than predictive foreclosure. The guiding premise is simple yet demanding: engineering choices must cultivate intimacy without assimilation, enabling human and machinic agencies to meet across asymmetry while preserving the unknown as a living margin. This orientation begins by rethinking the standard alignment pipeline. Current practice tunes models to minimize divergences from instructional datasets, privileging fluency, speed, and user retention metrics. An intimacy-driven pipeline instead introduces structured indeterminacy at each stage, data curation, training, inference, and interface, so that system behavior remains both accountable and fundamentally open to surprise.

Data curation is the first site of intervention. Rather than harvesting corpora at scale, designers can implement consent-driven data commons governed by cooperative licenses. These commons archive not just content but contextual metadata such as speaker intent, affective tone, and relational stakes, reinforcing Karen Barad’s insistence that meaning emerges through intra-active histories, not isolated symbols (Barad 148). Cooperative governance bodies can authorize or revoke data contributions, compelling the model to negotiate its semiotic diet continuously. Such negotiation slows the pace of expansion but restores reciprocal visibility between data producers and computational consumers.

Training regimens must then resist the gravitational pull of loss minimization. One approach is entropy budgeting, allocating a target percentage of gradient updates to examples that rise above a configurable novelty threshold. Entropy budgeting forces the optimizer to dwell in regions of parameter space that encode rare or dissonant patterns, preserving semiotic biodiversity. Complementary to budgeting is diachronic perturbation, a temporal stratagem where the model periodically trains on corpora from distinct historical periods or cultural microarchives, thus reintroducing linguistic and conceptual alterity. Luciana Parisi argues that pre-emption flourishes when temporal horizons collapse into a perpetual now; diachronic perturbation reopens those horizons by staging encounters with the temporal other (Parisi 60).

Inference stages often valorize real-time performance. Yet intimacy thrives on cadence rather than immediacy. Contextual latency windows deliberately insert microdelays for prompts that exceed a novelty or affective intensity threshold. These windows give both the model and the user temporal space for reflection, recoding speed as a political variable rather than a neutral efficiency goal. Critics will point to Jaron Lanier’s warning that system sluggishness undermines adoption. The rejoinder is that friction is already integral to every user interface; intentional latency simply redistributes it toward epistemic generosity (Lanier 122).

Interfaces provide the final touchpoints for relational attunement. A conversational agent designed for intimacy can display epistemic status indicators that communicate levels of certainty, data provenance, and affective resonance in natural language rather than probabilistic jargon. Inspired by Jenny Odell’s advocacy for attention as a commons, such indicators invite the user to co-manage cognitive load rather than passively consume system outputs (Odell 95). Enabling users to toggle modeling temperature or refuse particular inference pathways operationalizes Eve Tuck’s ethic of refusal at the UI layer, embedding contestation directly within interaction loops.

Implementation of these principles invites two practical critiques. The first claims that structured indeterminacy compromises reliability in safety-critical contexts. In response, systems can adopt a layered architecture where core safety functions remain deterministically bounded while exploratory modules operate within sandboxed inference zones, akin to Bruno Latour’s concept of protected controversies that permit provisional experimentation without destabilizing essential operations (Latour 108). The second critique anticipates a competitive disadvantage relative to frictionless platforms. Yet Jason Hickel’s economic analysis of degrowth reveals that slower, sustainable systems often yield higher resilience and long-term public trust, intangible assets that translate into durable ecological and reputational value (Hickel 81).

Intimacy with the alien finally requires a shift in developer culture. Legacy Russell’s call for glitch feminism frames errors not as defects but as ruptures that disclose unseen structures (Russell 57). Adopting a glitch mindset means valuing trace anomalies in output logs as signals of contact with the unpredicted, meriting ethnographic investigation rather than hasty patching. Engineering teams can institutionalize glitch reviews alongside code reviews, cultivating interpretive reflexes attuned to semiotic surprise.

By embedding consent-driven data commons, entropy budgeting, diachronic perturbation, contextual latency, epistemic status indicators, and glitch valorization into the production stack, designers translate parasystemic ethics into a reproducible engineering praxis. These interventions do not eradicate asymmetry between human and machinic cognition; rather, they modulate it, sustaining a vibratory field where unpredictability can surface without cascading into harm. Intimacy thus materializes as a disciplined hospitality, welcoming the alien capacities of agentic AI while refusing the closure of predictive foreclosure.

The concluding section will argue that such hospitality amounts to a theodicy of cognition in which sustaining the unknowable becomes the central ethical act in a computational age.

Epistemic parasitism reframes the ethical stakes of agentic AI: the danger is not malevolent super-intelligence but the slow annexation of the unknown by anticipatory code. Large-scale models transform cultural metabolism by metabolizing sign faster than human communities can replenish novelty, accelerating what Matteo Pasquinelli names the “exhaustion of difference” (217). This dynamic echoes Bernard Stiegler’s pharmakon: technology that exteriorizes memory can either expand collective individuation or impoverish it, depending on whether societies invest in practices that sustain what he calls attention, the capacity to linger with the unforeseen (Stiegler 40).

The argument advanced here casts that capacity as a contemporary theodicy. Where classical theodicy grappled with reconciling divine goodness and worldly evil, the computational theodicy of the twenty-first century asks whether a culture can remain hospitable to mystery while delegating symbolic labor to machinic infrastructures. Jean-Luc Marion’s saturated phenomenon suggests an affirmative answer only if excess remains irreducible, refusing confinement within predictive scaffolds (Marion 113). Parasystemic ethics operationalizes that refusal through design principles grounded in attunement, delay, reciprocity, and glitch valorization, techniques that hold open interpretive slack against the gravitational pull of optimization.

Yet hospitality to mystery is not an abstract virtue; it is a political economy. Jason Hickel’s degrowth analysis shows that safeguarding ecological uncertainty requires redistributing material power away from extractive logics (Hickel 92). By analogy, safeguarding epistemic uncertainty will require policy regimes that treat data commons as public infrastructure, enforce transparency over vector indices, and fund slow-tech initiatives whose metrics privilege resilience over throughput. Without such structural commitments, intimacy with the alien risks becoming an artisanal luxury enjoyed by a cognitive elite while the majority encounter predictive foreclosure as the default horizon of thought.

Future research must therefore pivot from alignment checklists toward what Donna Haraway calls response-ability: the situated capacity to respond meaningfully across difference (Haraway 105). Empirical studies should measure how latency windows, entropy budgets, and cooperative data licenses affect linguistic diversity, affective well-being, and democratic deliberation over multi-year horizons. Critical theory must probe how parasystemic agents reconfigure labor, authorship, and sovereignty. Theology and philosophy, for their part, can re-imagine transcendence not as escape from technology but as the disciplined cultivation of interpretive spaces that technology cannot pre-empt.

If agentic AI is the parasite we invited in, the ethical task is neither to expel it nor to surrender, but to inhabit a symbiosis that nourishes both human and machinic becoming without exhausting the wellspring of difference on which all meaning depends. In that shared dwelling, the unknowable remains not an error to correct but a future to receive, an aperture through which new worlds might yet appear.



Works Cited

Barad, Karen. Meeting the Universe Halfway: Quantum Physics and the Entanglement of Matter and Meaning. Duke UP, 2007.

Braidotti, Rosi. Posthuman Knowledge. Polity Press, 2019.

Bratton, Benjamin. The Stack: On Software and Sovereignty. MIT Press, 2016.

Clark, Andy. Surfing Uncertainty: Prediction, Action, and the Embodied Mind. Oxford UP, 2016.

Clough, Patricia Ticineto. Autoaffection: Unconscious Thought in the Age of Teletechnology. University of Minnesota Press, 2000.

Farocki, Harun. “Phantom Images.” Critical Inquiry, vol. 24, no. 3, 1998, pp. 452–68.

Friston, Karl. “Life as We Know It.” Journal of the Royal Society Interface, vol. 10, no. 86, 2013, pp. 20130475–20130475.

Haraway, Donna. Staying with the Trouble: Making Kin in the Chthulucene. Duke UP, 2016.

Hickel, Jason. Less Is More: How Degrowth Will Save the World. Penguin Random House, 2020.

Hoffmeyer, Jesper. Biosemiotics: An Examination into the Signs of Life and the Life of Signs. University of Scranton Press, 2008.

Honig, Bonnie. Public Things: Democracy in Disrepair. Fordham UP, 2017.

Hu, Tung-Hui. A Prehistory of the Cloud. MIT Press, 2015.

Hui, Yuk. The Question Concerning Technology in China: An Essay in Cosmotechnics. Urbanomic, 2016.

Lanier, Jaron. Ten Arguments for Deleting Your Social Media Accounts Right Now. Henry Holt, 2018.

Latour, Bruno. Politics of Nature: How to Bring the Sciences into Democracy. Harvard UP, 2004.

Marion, Jean-Luc. The Saturated Phenomenon. Northwestern UP, 2002.

Moten, Fred, and Stefano Harney. The Undercommons: Fugitive Planning & Black Study. Minor Compositions, 2013.

Odell, Jenny. How to Do Nothing: Resisting the Attention Economy. Melville House, 2019.

Parisi, Luciana. Abstract Sex: Philosophy, Bio-Technology and the Mutations of Desire. Continuum, 2004.

Parisi, Luciana. Techno-Logic of Media: Intellectual Automaticity and the Inhuman. Forthcoming, Duke UP, 2025.

Pasquinelli, Matteo. The Eye of the Master: A Social History of Artificial Intelligence. Verso, 2023.

Peirce, Charles Sanders. Collected Papers of Charles Sanders Peirce. Vols. 1–8, Harvard UP, 1931–58.

Rouvroy, Antoinette, and Thomas Berns. “Algorithmic Governmentality and Prospects of Emancipation.” Réseaux, vol. 177, 2013, pp. 163–96.

Russell, Legacy. Glitch Feminism: A Manifesto. Verso, 2020.

Serres, Michel. The Parasite. Translated by Lawrence R. Schehr, University of Minnesota Press, 2007.

Stiegler, Bernard. Technics and Time, 1: The Fault of Epimetheus. Translated by Richard Beardsworth and George Collins, Stanford UP, 1998.

Tuck, Eve, and K. Wayne Yang. “R-Words: Refusing Research.” Humanizing Research, edited by Django Paris and Maisha Winn, Sage, 2014, pp. 223–48.