I. Institutions as Thinking Ecologies
I begin from a scene that feels ordinary yet philosophically instructive. Across a sprawling organization, thousands of decisions are routed, logged, and confirmed by agents—human and algorithmic—who rarely meet except through traces of data or fragments of code. Each actor believes that their work aligns with a rational order once ratified as good. Each tool executes a script that translates what once was judgment into procedure. The organization thus moves with impressive coordination while concealing an instability at its core: a fragile consensus about what counts as knowledge, how error is recognized, and when surprise is treated as threat rather than invitation. The paradox is not that mistakes arise but that rational, well-intentioned practices can render an institution progressively less able to learn. The more it appears to know, the less willing it becomes to change its mind.
This dissertation treats that paradox as epistemological rather than technical. Institutional epistemology names the study of how large systems decide what knowledge is, who may bear it, and by what process it may be revised. These decisions are diffused across dashboards, policy memos, and model cards whose ontological assumptions are rarely examined. When read together, they form what Alfred North Whitehead would call an “organism of experience,” a living nexus of relations that perpetually prehend and reinterpret the world (Whitehead 68-70). Institutions thus behave as ecologies of thought: distributed systems that metabolize information through rituals of attention, evaluation, and care.
To call the institution an ecology is to assert that it organizes perception, value, and time. Donella H. Meadows reminds us that “a system’s structure determines its behavior; changing the structure changes everything” (Meadows 169). The same applies to institutions. Their metrics and calendars act like predators and seasons, shaping which forms of novelty can survive. Their stories of success sediment into what Whitehead terms “stubborn fact,” creating conditions where the map hardens faster than the territory can change (Whitehead 78). Learning, under such conditions, is not an episodic event but a slow ecological capacity: the ability to integrate surprise without collapse.
This ecological understanding transforms the contemporary fascination with artificial intelligence. The public narrative imagines AI as an autonomous cognition that extends human intellect; yet, as Gilbert Simondon argued, technical objects are “incomplete beings,” evolving only through their relation to the milieus that sustain them (Simondon 48). A model learns to the extent that its institution learns to revise the meanings of reliability, accuracy, and relevance. When organizations enthrone predictability as their highest virtue, they reproduce ignorance at scale. Byung-Chul Han calls this the pathology of “the society of transparency,” where reflection yields to exposure and complexity is flattened into immediate visibility (Han 8-11). In such a climate anomaly appears as failure rather than revelation.
To repair that pathology, design must be understood as ethical practice. Donna Haraway insists that “staying with the trouble” requires cultivating “response-ability”—a capacity to remain answerable within the tangle of relations one inhabits (Haraway 14). Within institutions, this means designing infrastructures that do not evacuate ambiguity but hold it long enough for understanding to evolve. A meeting cadence, a review ritual, or a documentation template becomes an ethical technology shaping what the organization is willing to see.
The philosophical scaffolding for this reorientation draws from four converging traditions. First, Whitehead’s process metaphysics redefines permanence as pattern in flux: “The many become one, and are increased by one” (Whitehead 21). If reality itself is process, then institutional stability must be rhythmic, not static. Second, Martha C. Nussbaum’s reading of Aristotelian virtue in The Fragility of Goodness clarifies that moral intelligence depends on phronesis, the discernment of what is fitting within circumstance (Nussbaum 302-03). Institutional wisdom analogously requires architectures that sustain discernment rather than replace it with rule. Third, Meadows’ systems theory demonstrates that feedback loops and delays are decisive sites of intervention; when feedback is silenced, the system becomes “addicted to its own success” (Meadows 162). Finally, Simondon and Haraway teach that technology is not external to ethics but one of its mediums, encoding ontological wagers about what the world is and should become. Together, these perspectives converge on a single proposition: if institutions fail to learn, it is because their architectures of perception, responsibility, and time are tuned to equilibrium rather than to encounter.
The point is not to demonize reliability. Amartya Sen, in Development as Freedom, reminds us that reliability can protect basic capability by preventing the arbitrary (Sen 39). Yet when reliability becomes a moral idol, it confuses the absence of contradiction with the presence of truth. It rewards silence and conformity over inquiry. A culture of reliability breeds what Bruno Latour calls the “double language of modernity,” in which actors proclaim openness to progress while enforcing closure through procedural certainty (Latour 36). Artificial intelligence then becomes an alibi for moral exhaustion, a technical mask that hides the refusal to be surprised.
I define the architecture of learning as the deliberate design of those epistemic conditions. Such an architecture arranges how doubt is authorized, how evidence is composed, and how decisions can be unmade without humiliation. It joins measurement with narrative, for as Nussbaum observes, “reason without imagination is not yet reason” (Nussbaum 41). It acknowledges temporal rhythm: moments of analysis, intervals of rest, and the sabbath of reconsideration that prevents the ecology from devouring its own attention. Above all, it encodes humility, recognizing with Whitehead that “no system can hope to embrace the infinite detail of the universe” (Whitehead 74).
Learning at institutional scale is therefore an ethical act. Every revision of knowledge redistributes risk and labor. When a predictive threshold shifts, someone inherits the unseen consequences. When a model is frozen for stability, future teams must reconcile the cost of delay. To learn responsibly is to make these redistributions explicit, to “apologize to one’s own future,” as the saying might go, and to embed accountability into the architecture itself.
Iteration, within this frame, is not the cult of speed but the ritual of repentance. It is the art of revising without contempt for what came before and without hubris toward what will follow. It transforms anomaly from defect into pedagogue. Meadows captures this posture succinctly: “We can’t impose our will upon a system; we can listen to what the system tells us” (Meadows 170). Listening becomes the central engineering virtue.
The materials of this architecture are fourfold. Cognitive infrastructures shape what can be perceived; ethical infrastructures codify who may decide; affective infrastructures regulate the institutional emotions of fear, pride, and loyalty that determine how error is treated; and temporal infrastructures govern the duration of attention and rest. These correspond to Whitehead’s conviction that organism and environment co-determine each other, since “each occasion is a society, and every society is an environment for others” (Whitehead 90).
Consider a final contrast. In one review meeting, success is measured by precision and compliance. In another, participants must articulate the ontology their model assumes, identify what it refuses to automate, narrate the most instructive failure encountered, and specify the conditions of obsolescence. The first institution equates intelligence with control; the second equates intelligence with the courage to name its own limits. The difference is architectural. It decides whether the ecology preserves its capacity for wonder—the condition Whitehead described as “the beginning of philosophy” (Whitehead 173).
The animating question of this work thus emerges clearly: How can a large system remain coherent while continuously changing what it knows. Coherence without openness decays into bureaucracy; openness without coherence dissolves into noise. The task is to design a unity strong enough to endure revision and a freedom disciplined enough to sustain trust. What follows attempts to make that unity visible, ethical, and teachable.
II. Problem Statement and Thesis
Large institutions have learned to formalize learning while avoiding transformation. They build pipelines of data, layers of analytics, and protocols of review, then announce that the system is intelligent because it gets better at prediction each quarter. What passes for intelligence is an increasingly refined management of variance that treats deviation as a defect in need of suppression rather than as a messenger from outside the established frame. The result is a style of governance that worships reliability, confuses compliance with truth, and reproduces ignorance whenever the world exceeds the categories that once domesticated it. The problem is not that institutions value consistency. The problem is that consistency has been elevated into an ontology in which reality is assumed to be stable enough to be held by yesterday’s abstractions, while those who are tasked with keeping the machine running experience anomaly as personal failure rather than as an institutional event that requires collective redescription.
This problem can be stated more precisely. A large system will claim to learn if it adjusts parameters within a fixed model of the world, yet avoid learning if it must question the model itself. James G. March described this tension as the difference between exploitation and exploration, where the former optimizes within known certainties and the latter seeks new possibilities that cannot be guaranteed in advance, and he showed how organizations drift toward exploitation because its rewards are near, legible, and safe while exploration is slow, costly, and socially vulnerable (March 71). When governance encodes this drift as virtue, the institution’s most polished practices slowly become its blinders. At that point artificial intelligence does not fail because models are weak. It fails because the surrounding ecology has defined learning as the reduction of surprise. Under those conditions accuracy improves while wisdom declines.
The philosophical stakes of this drift have long been understood. Thomas S. Kuhn taught that a paradigm sustains periods of normal inquiry by setting the problems worth solving and the methods for solving them, while anomalies accumulate at the margins until a crisis of description forces a change in the very terms through which evidence is recognized as evidence (Kuhn). The institutions I study have replicated the habits of normal science without preserving its courage. They sustain an expensive machinery of normal operations, but when anomalies appear they are routed into exception queues, discounted as noise, or neutralized by threshold changes that reclassify the surprising as the irrelevant. Such practices preserve stability while degrading perception. They relieve anxiety in the present while impoverishing the future’s capacity to understand itself.
Repair begins with a different anthropology of knowledge. Michael Polanyi’s claim that we can know more than we can tell reminds us that the intelligence of practice exceeds the explicit procedures that attempt to capture it, which means that an institution that strips away tacit judgment in the name of standardization will often remove the very capacities it needs when the unexpected arrives (Polanyi 4). The dominance of audit logics in contemporary governance converts this excess into embarrassment. People learn to avoid naming what they sense but cannot prove, and they learn to defer to dashboards that never blush. Karl E. Weick’s analysis of sensemaking explains why this avoidance is so resilient, since organizations stabilize their worlds through retrospective narratives that render action sensible after the fact, and those narratives become identity commitments that resist revision even when evidence shifts (Weick). In that climate systems will claim to be learning while actually perfecting their defenses against surprise.
The thesis of this dissertation is direct. Institutions can be designed to internalize surprise as a structured practice of reflection. True intelligence at scale is measured not by prediction accuracy alone but by the speed and integrity with which a system revises its understanding when it is contradicted by the world. Integrity here does not mean the preservation of face. It means the willingness to unmake decisions without humiliation, to redistribute attention without retaliation, and to acknowledge that revision will always carry costs that must be owned rather than hidden. Learning therefore becomes governance of epistemic freedom, which I define as the protected capacity of an ecology to authorize doubt, sustain dissent, and translate anomaly into new descriptions that reenter operations without being sanitized into triviality.
To defend this thesis I specify the mechanisms by which failure propagates when reliability becomes an idol. First, metric capture converts means into ends. Donella H. Meadows warned that every measure transforms the environment that it measures, and that poorly placed indicators will induce behaviors that protect the indicator while harming the system that indicator was meant to serve (Meadows). In large institutions key performance indicators become climates in which people must live, and those climates encourage short time horizons, risk aversion, and a ritualized optimism that renders dissent antisocial. Second, deontic closure replaces judgment with rule. Chris Argyris and Donald A. Schön showed that organizations learn at two levels, one that adjusts action without questioning governing variables, and one that interrogates the variables themselves, yet the second level is rare because it threatens existing authority and identity (Argyris and Schön). When artificial intelligence is introduced into such cultures it tends to strengthen the first level by making action more efficient, while making the second level more difficult by shrouding governing assumptions behind the authority of models. Third, narrative erosion dissolves institutional memory. As teams accelerate delivery cycles they often abandon the slow work of archival reflection, which leaves the organization unable to remember why a prior decision was made and therefore unable to revise it responsibly. Fourth, affective discipline trains people to fear being the source of bad news. Byung-Chul Han has described how the demand for transparency can flatten complexity into immediate exposure, and inside large systems that moralizes speed and visibility, careful doubt appears as delay and is punished as disloyalty (Han).
Each of these mechanisms can be reversed by design. If metric capture is a structural temptation, then evaluation must be rebuilt to privilege signals of sensemaking over signals of throughput. If deontic closure is a cultural default, then governance must institutionalize the right to slow down when a description feels thin, and must reward the humility to retire a successful model because its ontology has expired. If narrative erosion is a byproduct of acceleration, then documentation must be reframed as philosophical practice, where model cards, review minutes, and postmortems are curated as arguments about the world that future teams can interrogate. If fear governs attention, then leadership must model the speech acts by which uncertainty is dignified rather than shamed. Donna J. Haraway’s call to cultivate response ability names the center of this reorientation, since the goal is not perpetual indecision but an institutional poise that can remain answerable to the complexity it participates in rather than reducing that complexity to a set of dashboards that no longer see (Haraway).
The thesis demands an operational definition of success. A learning architecture will not celebrate a decline in exceptions if that decline is achieved by redefining exceptions out of existence. It will not treat participation counts in review meetings as evidence of pluralism if dissenting speech has been domesticated into polite ritual. It will not call a model safe because it is stable if that stability has been purchased by excluding the very contexts in which harm is most likely to arise. Instead the system will measure how quickly anomalies are elevated to narration, how often governing assumptions are named explicitly, how many times a decision is unmade in public without retaliation, and how diversely the institution is able to interpret the same event without collapsing into faction or into sterile consensus. Amartya Sen’s conception of freedom as capability anchors this reframing, since the point is to enlarge the real opportunities an institution has to choose reasoned action rather than to maximize outputs under procedural constraint (Sen).
The remainder of this work elaborates the architecture that makes such freedom durable. It will show how to redesign cognitive infrastructures so that models and metrics declare their ontological assumptions rather than smuggling them in as defaults, how to establish interpretive circuits in which anomaly becomes a formal data object that reenters the learning loop with narrative context intact, and how to govern temporal rhythms so that speed and depth are jointly optimized rather than traded against one another. It will also demonstrate that these reforms are not instruments at the periphery of technical systems but are constitutive mediations of the ecology itself. The wager is that institutions can be composed to welcome contradiction without losing coherence, can cultivate humility without abandoning precision, and can perfect reliability without confusing it for truth. If they can do so, artificial intelligence becomes a partner in collective moral cognition rather than an alibi for the refusal to change one’s mind.
III. Conceptual Frame: The Epistemic Ecology
I name the institution an epistemic ecology in order to make a strong claim about how collective intelligence arises and fails. Intelligence in this frame is not the property of components but the emergent consequence of relations among infrastructures that shape what can be perceived, what must be protected, how error feels, and how long attention endures. The ecology metaphor does not romanticize organic life. It disciplines analysis toward interactions, feedback, and delayed effects that are often invisible from within a single team or tool. Following Donella Meadows, I treat structure as destiny while refusing fatalism, since structures can be redesigned once their leverage points are seen and narrated with care (Meadows). Following Alfred North Whitehead, I treat institutions as societies of occasions in which patterns persist through becoming, so that stability appears as rhythm rather than as stasis and learning appears as the renewal of pattern in the presence of genuine novelty (Whitehead). What follows specifies four interacting infrastructures that jointly condition how an institution comes to know. They are cognitive, ethical, affective, and temporal. None can be understood in isolation, and each becomes pathological when elevated above the others. Together they compose the architecture of learning.
Cognitive infrastructure names the perceptual apparatus of the ecology. It includes data schemas, taxonomies, and the metrics and models that distribute salience by deciding which differences will be seen. It also includes the documentation genres that fix meaning long enough for collaborative work. Geoffrey C. Bowker and Susan Leigh Star showed that classification systems are never neutral since they carry histories of labor and value that become naturalized as common sense over time, which means that every schema constitutes a politics of visibility and invisibility that later generations may mistake for the order of things rather than for a historical settlement among contested goods (Bowker and Star). Donna Haraway’s insistence on situated knowledges unsettles any claim to a view from nowhere and clarifies that description is always embodied and positioned, which implies that cognitive infrastructure must encode its own partiality rather than conceal it behind the prestige of scale or the glamour of precision (Haraway, “Situated Knowledges”). Within this frame a model card is not a bureaucratic requirement. It is a philosophical text that declares an ontology and marks the zone where automation must refuse itself. When cognitive infrastructures are designed as if they were mirrors of reality rather than instruments for ethical orientation, they tempt institutions to confuse performance with truth and to treat anomaly as defect rather than as teacher.
Ethical infrastructure names the norms, permissions, and accountabilities that govern action when goods compete. It includes formal codes of conduct, review bodies, escalation paths, and the informal folkways by which dissent is dignified or disciplined. Martha C. Nussbaum’s reading of practical wisdom clarifies why procedure cannot replace judgment, since the fitting act depends on attention to the texture of circumstance that general rules can only approximate without finality (Nussbaum). Chris Argyris and Donald A. Schön distinguish single loop from double loop learning and show how organizations often adjust tactics while protecting governing variables that ought to be interrogated, especially where status and identity are at stake (Argyris and Schön). Ethical infrastructure therefore concerns more than compliance. It concerns the institutional courage to unmake prior decisions without humiliation, to widen the circle of interpreters without diluting responsibility, and to treat governance as the cultivation of habits that make revision possible. Where ethical infrastructure atrophies, deontic closure replaces judgment with ritual, audit becomes theater, and artificial intelligence inherits the moral exhaustion of the culture that deploys it.
Affective infrastructure names the circulating emotions of the ecology and the implicit pedagogy through which people learn what it costs to speak or to remain silent. It includes the pride that animates speed, the fear that accompanies error disclosure, the loyalty that binds teams across uncertainty, and the shame that follows public contradiction. Karl E. Weick’s account of sensemaking reminds us that organizations stabilize themselves through narratives that render action plausible after the fact and that identity commitments congeal around these stories, which means that affect is not a decoration upon reason but a medium of collective cognition that can either enlarge or constrict the field of the sayable (Weick). Byung Chul Han’s analysis of the transparency society shows how the moralization of exposure can crush complexity into the demand for immediate legibility, so that careful doubt comes to look like disloyalty and measured delay is punished as incompetence (Han). In such climates the ecology learns to anesthetize anomaly. Repair requires modeling the speech acts through which uncertainty is dignified, surprise is narrated without blame, and responsibility is shared without diffusion. Affective infrastructure thus becomes an object of design rather than a background mood.
Temporal infrastructure names the cadences, horizons, and sanctuaries of time that govern how long attention is sustained and when rest becomes obligatory. It includes the pacing of delivery cycles, the seasons of review, the archival intervals in which documentation is curated as an argument about the world, and the sabbath practices by which the ecology refuses to consume all of its own attention. James G. March alerts us to the structural bias toward exploitation over exploration, since the returns to exploitation are proximate and legible while the returns to exploration are distant and risky, which produces a slow decay of imagination under pressures that feel rational at each step (March). Thomas S. Kuhn shows that anomalies can accumulate without consequence during periods of normal inquiry until crisis forces a re description, but crisis is not an argument. It is a failure of temporal stewardship that allowed pressure to rise where more patient rhythms could have metabolized novelty with less harm (Kuhn). Temporal infrastructure must therefore encode intervals where the institution can question without the tyranny of immediacy, can revise without humiliation, and can remember why a decision was once right so that it can be unmade with gratitude rather than contempt.
These four infrastructures do not coexist as a checklist. They interpenetrate as a living system of constraints and affordances. Cognitive choices create ethical obligations by determining who is affected by what the organization chooses to see. Ethical choices create affective climates by signaling which risks will be honored or punished. Affective climates shape temporal patience by teaching bodies whether to rush or to rest. Temporal rhythms feed back into cognition by determining which patterns become visible at all. Gregory Bateson’s dictum that information is a difference that makes a difference takes on institutional specificity in this circle, since only those differences that survive the gauntlet of norms, emotions, and time can become knowledge, and only those forms of knowledge that return to restructure perception can renew the ecology rather than ossify it (Bateson).
This ecological frame also clarifies the typical failure modes. When cognitive infrastructure dominates, metrics evolve into climates that people must inhabit while the institution forgets that measures are instruments and not oracles. When ethical infrastructure dominates, deliberation elongates into protective ceremony and novelty suffocates under the weight of good intentions. When affective infrastructure dominates, leadership becomes mood management and the collective sacrifices truth for harmony. When temporal infrastructure dominates, process becomes an idol and the institution mistakes endurance for wisdom. The remedy is not a new balance struck once and for all. It is a disciplined practice of reflexivity in which each infrastructure periodically submits to critique by the others. Suchman’s insight that plans are indexical to situated action can be generalized here. Designs must be accountable to the situations they intend to govern, and that accountability requires cycles of attention that move from model to case to counterexample to redesigned model with narrative memory intact (Suchman).
An epistemic ecology earns the name architecture of learning when it encodes three recursive movements across these infrastructures. First, it makes ontologies explicit. Models and policies must state what they treat as entities, relations, and exclusions, so that ethical and affective consequences can be anticipated rather than discovered through harm. Second, it formalizes interpretive circuits. Exceptions, dissent, and anomalies must be treated as data with narrative context that reenter modeling and measurement rather than being laundered into irrelevance through threshold manipulation. Third, it governs learning velocity together with learning depth. Speed becomes a virtue only where depth is preserved by practices that index the number of governing assumptions questioned, the diversity of interpreters engaged, and the durability of the new description over time. These movements do not rest on sentiment. They are supported by the lessons of systems theory, by the metaphysics of process, by the ethics of practical wisdom, and by the sociology of sensemaking, each of which insists that intelligence at scale is a property of relation ordered in time.
The ecological frame now stands ready for method. If every artifact in the institution is a philosophical text in technical clothing, then ethnographic listening and design experiment will be required to reconstruct how anomalies are negotiated and how categories are enforced. If infrastructures are interdependent, then evaluation must reach beyond throughput into measures of attention, humility, and narrative coherence. If affect and time shape what can be known, then leadership must become custodians of climate and cadence rather than champions of velocity alone. The point is not to soften governance. The point is to compose a culture whose reliability is hospitable to surprise and whose precision remains teachable by a world it does not own.
IV. Methodology: Situated Reflexive Design
Method, in an institution understood as an epistemic ecology, must be practiced as a way of seeing that changes what it sees. The procedure I adopt treats every artifact and encounter as a site where ontology has been encoded and can be made discussable. I join institutional ethnography, interpretive systems mapping, and design experiment into a situated reflexive discipline that renders anomalies legible without coercing them into premature resolution. Following Dorothy E. Smith, I begin from the standpoint of participants whose daily coordination is shaped by texts that travel and rule, not in order to romanticize experience but to reveal how objectified relations are organized and sustained through document genres, reporting channels, and accountability rituals that seem natural because they have been naturalized through repetition and authority. Institutional ethnography is therefore not an optional qualitative supplement to technical rigor. It is the analytic core that exposes how a system knows and how it distributes the right to know by arranging who may speak in what register to whom and with what consequence for future action, since texts in Smith’s sense do work that binds local lives to translocal ruling relations and that work must be traced in situ if reform is to be anything more than cosmetic (Smith 34–36, 75–78).
Because action is always indexical to its situation, the inquiry proceeds with Lucy Suchman’s instruction that plans are resources for action rather than determinants of it, which means that the choreography of a review, a handoff, or a triage cannot be inferred from process maps alone but must be reconstructed through close observation of how people repair breakdowns, negotiate category boundaries, and decide what counts as sufficient evidence for escalation or closure. I therefore treat shadowing, contextual interviews, and artifact walkthroughs as collaborative sensemaking where the aim is not to extract opinions but to surface the implicit grammars by which participants render their worlds actionable. The resulting records include fieldnotes, annotated screenshots, model cards, decision logs, and excerpts from policy or audit texts. They are analyzed abductively, with hypotheses formed and revised in conversation with participants, consistent with Donald A. Schön’s claim that professional knowledge advances through reflection in action where practitioners frame problems, experiment in the doing, and then reflect upon the outcomes to reframe their understanding of both problem and situation (Suchman 27–33, 52–55; Schön 26–29, 132–36).
Interpretive systems mapping then composes these situated traces into an architecture of relations. The work here is not to draw a totalizing diagram that pretends to master the whole, but to disclose feedbacks, thresholds, and delays that organize what can be known and done. I follow Donella H. Meadows in treating structure as the proximate cause of behavior, yet I refuse the fatalism that sometimes follows diagrammatic clarity, because maps in this project are performative proposals for intervention. They mark leverage points where small changes in classification, escalation cadence, or documentation practice can disproportionately improve the ecology’s capacity to be instructed by the world it purports to model. I also import Bruno Latour’s insistence that the social is not a substance but a network of associations whose stability is a local achievement, which requires that maps record not only formal roles and data flows but also the mediating objects and scripts through which alignment is negotiated, including dashboards, templates, meeting agendas, and boundary categories that travel across teams and translate worlds while never fitting them exactly (Meadows 145–70; Latour 64–86).
Artifact analysis proceeds with the assumption that a policy, a model card, or a dashboard is a philosophical text wearing technical clothing. I read classification systems with Geoffrey Bowker and Susan Leigh Star’s reminder that categories hold histories of labor and value that become invisible through use, which means that to change a schema is to change a politics of visibility whether or not stewards acknowledge the moral character of their technical choice. I read documentation with Donna Haraway’s demand for situated accountabilities, which entails that authors must mark the vantage points and limits from which their descriptions speak rather than perform a view from nowhere. In practice this means that model reporting templates require fields that make ontological commitments explicit, including what the model treats as entities and relations, which contexts have been excluded from automation and why, and under what observable conditions the model should be retired or constrained. I treat model cards and datasheets not as bureaucratic obligations but as genres that can institutionalize humility when written as arguments about the world rather than as public relations statements about performance. They become sites where tacit judgment is given a durable form that can travel responsibly, in the spirit of Michael Polanyi’s claim that we know more than we can tell and that the aim of explicit representation is to honor rather than to liquidate that excess (Bowker and Star 1–5, 33–61; Haraway 582–86; Mitchell et al.; Gebru et al.; Polanyi 4–7).
Because this methodology must not only diagnose but change the conditions of learning, I use design experiments that are small, reversible, and co owned by those who will live with their consequences. The criterion for a good experiment is not cleverness but generativity. An intervention should amplify the institution’s ability to surface contradiction early, narrate it without shame, and translate it into revised descriptions that reenter operations with their contexts intact. Typical experiments include the insertion of an anomaly intake ritual into review meetings that requires the presenter to name the most instructive failure encountered, to state the ontology assumed, and to specify the sunsetting conditions under which the artifact should be retired. They include the redesign of model cards such that assumptions, exclusions, and unresolved questions are first class fields, and that these fields are reported upward in governance rather than trimmed away before senior review. They include the institution of reflective debriefs after high stakes decisions where quantitative results are read alongside narrative reconstructions of judgment under uncertainty, a practice indebted to Argyris and Schön’s double loop learning in which the governing variables themselves are interrogated rather than only the efficacy of tactics. They include the curation of an internal commons where postmortems and counterexamples are archived as living arguments, so that future teams inherit memory as a resource rather than begin again in ritualized amnesia. In each case the experiment is evaluated for its effects on attention, speech, and time, rather than only for throughput.
Evaluation is designed to preserve reflexivity. Quantitative traces are collected, including counts of anomalies elevated to narration, median time from detection to deliberation, and frequency with which governing assumptions are explicitly named in recorded minutes. Qualitative traces are given equal weight, including the moral tone of deliberation, the presence of humility language in reviews, and the persistence of curiosity across cycles, understood here as the willingness to ask questions that cannot be answered within the current frame. I adopt a mixed logic of abductive and longitudinal assessment. Abductively, I follow surprises where they lead and update hypotheses quickly when a change in practice yields unexpected effects. Longitudinally, I track whether early gains in vigilance are sustained or decay under pressure, since many reforms produce a temporary increase in attention that ebbs once novelty fades. The test of method is therefore not an isolated performance metric but the durability of a new ecology of learning under ordinary constraints. Amartya Sen’s conception of freedom as capability informs the synthesis, since the final criterion is whether the institution has enlarged its real opportunities to choose reasoned action rather than merely increasing output under procedural constraint. If the ecology learns to unmake decisions without humiliation, to redistribute attention without retaliation, and to welcome contradiction without losing coherence, then the method has succeeded at the only scale that matters, which is the temporal scale of institutional character formation over time rather than the episodic scale of project success (Argyris and Schön 2–3, 18–22; Sen 36–41).
The methodology is called situated because it refuses the fantasy that design can act from nowhere. It is called reflexive because it requires those who inquire to become answerable for the effects of their inquiry on the very practices they study. It is design because it rearranges the conditions under which perception and judgment occur, composing infrastructures that make it easier to do the right thing for the right reasons at the right time. It is, finally, experimental, since institutions learn in history, with limited vision and contested goods, and the only honest path to wisdom is to try careful things, to watch closely, to apologize quickly, and to build forms of documentation and governance that carry those apologies forward as shared memory rather than as private remorse. In the end, method here is a pedagogy of attention administered at organizational scale. It aims to render the ecology both more reliable and more teachable, to protect coherence while multiplying the pathways through which the world can revise what the institution believes about itself.
V. Case Studies of Learning and Closure
I present three anonymized archetypes in which large institutions mistook the signs of mastery for the substance of understanding. Each case follows a common arc. An ecology of procedures and proofs achieved stability under pressure. Success then hardened into identity. Anomalies that once taught the system how to see were reclassified as noise, as risk, or as bad faith. The point is neither to shame the actors nor to romanticize disorder. The point is to disclose how reliable practices, left unexamined, can quietly degrade an institution’s capacity to revise itself. I read each case with the aid of a canon that has already helped us think about organizations, expertise, failure, and surveillance. Donella Meadows cautions that structure is destiny while insisting that design can redirect flows of attention when leverage points are made visible. Thomas Kuhn reminds us that anomalies do not destabilize normal work until the description through which evidence is recognized is itself at issue. Michael Polanyi keeps the excess of tacit judgment on the table whenever standardization claims to capture everything that matters. Chris Argyris and Donald Schön explain why organizations defend governing variables even as they tinker with tactics. Marilyn Strathern and Michael Power map the moral economy of audit regimes in which visibility begins to substitute for care. Amy Edmondson shows that learning depends on climates of safety in which embarrassment does not govern speech. James Scott and Geoffrey Bowker with Susan Leigh Star teach that legibility is a political accomplishment with winners and losers. Karl Weick reminds us that organizations tell themselves stabilizing stories after the fact. In each case the sources function not as ornament but as vocabulary for explaining why good people doing good work can steadily lose the courage to learn.
The first case concerns stability that became an idol. A forecasting system had finally become precise enough to stop frightening the board. The leadership team, exhausted by prior volatility, froze the methodology. Change control was formalized in the name of safety. A quarterly ritual praised the shrinking error bands, then routed any deviations to an exceptions queue where risk staff assessed whether they were material. When a structural break arrived in the environment, the forecast held steady because the governance process had learned to interpret deviation as threat rather than as revelation. Teams who noticed new patterns were told to wait for the formal review window. The metric of reliability was maintained by narrowing the horizon of relevance to the range of values the model already expected. Stability, so hard won, became a cognitive climate that people had to inhabit. Meadows warned that any indicator reshapes the system that it measures by becoming a target that induces compensating behaviors. Here the compensating behavior was not falsification and not malfeasance. It was the steady social training of attention away from contexts that disturbed the performance line the institution had learned to love, a pattern well described by Power as the audit society’s tendency to substitute visible assurances for substantive improvement and by Strathern’s demonstration that when a measure becomes a target it ceases to be a good measure because practices adapt to the demand for visibility rather than to the demand for care (Meadows 163 to 173, Power, Strathern). The most telling detail emerged during interviews. Several practitioners remembered the moment they stopped bringing forward off pattern narratives. They could not pinpoint a single order and they did not resent leadership. They had learned, slowly and rationally, that anomaly talk generated work without reward. The ecology had quietly reclassified surprise as noise.
An adjacent literature helps read the mechanics of this closure. Kuhn observed that during periods of normal inquiry anomalies accumulate at the margins without interrupting routine, since equipment is trusted and problem choice is domesticated by shared exemplars. The system I observed had routinized normal operations while removing the pathways by which anomalies could recruit allies. James March would have recognized the exploitation bias that followed. Near returns on exploitation were legible, visitable, and promotable. The returns to exploration were distant and socially vulnerable. The effect was not neutrality. It was a self reinforcement loop in which performance under a fixed description was rewarded while curiosity about the description itself was treated as indiscipline. The repair strategy began by redesigning the review. Presenters were required to narrate one instructive failure and to state the conditions under which the forecast should be retired. Exception queues were replaced by an anomaly forum with the authority to trigger model retraining or to recommend contextual reframing. The outcome was modest and real. The error bands temporarily widened, performance charts grew less comforting, and the institution regained a form of perception it had traded away for the comfort of stability. The lesson is straightforward. A frozen description can be brave in times of panic, yet any ecology that treats deviation as impurity will eventually choose ignorance over vulnerability. Reliability must be re defined as the dependable willingness to be corrected.
The second case concerns the erosion of tacit expertise under the banner of speed. A negotiation support tool had transformed cycle times. Deals closed faster, documents were standardized, and governance could audit with ease. Senior leaders celebrated a new season of certainty. Three quarters in, unanticipated disputes began to rise. They were not errors in calculation. They were injuries of meaning. Contexts that previously had been handled by experienced practitioners at the edges of policy were now classified as compliant or non compliant within categories that had been designed to be simple enough to scale. Polanyi’s line that we can know more than we can tell proved exactly right. The institution had optimized the speakable while dissolving embodied know how into a handful of fields. Practitioners began to describe a feeling of being trained out of judgment. They were not dismissive of fairness or standardization. They were trying to name the sense that the tool was draining the practice of its art. Star and Ruhleder’s observation that infrastructure becomes visible upon breakdown seemed apt. Experienced staff did not campaign against the tool. They began to invent shadow practices of annotation, back channel calls, and unofficial pre work in order to keep the art of their craft alive inside a world that increasingly mistrusted anything that could not be logged as an auditable field entry (Polanyi 4 to 7, Star and Ruhleder).
Here the sociology of sensemaking provided further diagnostic precision. Weick taught that organizations construct plausible accounts after the fact in order to stabilize identity, not to arrive at perfect truth. The tool’s performance dashboards and governance decks supplied a new story in which speed and consistency proved that wisdom was present because error rates fell. That story then folded back into the ecology as a disciplining narrative that rendered dissent irrational. Amy Edmondson’s work on psychological safety helps explain why. In competitive settings, even mild embarrassment costs capital. Practitioners learned to keep misgivings private because the culture offered no graceful way to discuss the costs of standardization without sounding nostalgic or defensive. Repair therefore began with a procedural and rhetorical redesign. We established interpretive rituals in which practitioners were asked to annotate a small, randomly sampled set of negotiations with narrative reasoning for borderline decisions. Reflective debriefs convened monthly where text was read aloud alongside numbers. These sessions were not confessional theatre. They were a systematic attempt to give tacit judgment a durable form so that it could travel responsibly. Over the next cycles the dispute rate declined and, more importantly, so did the private sense of depletion among senior staff. The lesson is again straightforward. Standardization can be an instrument of justice and efficiency. Without interpretive rituals that preserve tacit intelligence as a first class object, the institution slowly trades away the capacities it will need when its frozen categories encounter a future the authors did not imagine.
The third case concerns hidden drift within an anomaly detection system. The system reduced reported errors and won prizes for responsible deployment. It also displaced uncertainty into zones that the metrics did not reach. Maintenance events, upstream data corrections, and shifts in business practice were repeatedly relabeled as out of scope rather than as signs that the detector’s ontology required revision. The team did not falsify results. They performed the logic they had been given. The metric that mattered to leadership was the count of events detected and resolved within service levels. Latour would call attention to the network of mediations that allowed the system to speak successfully. The pipeline of labeled data, the alerting thresholds, the governance cadence, and the success narrative together formed an apparatus that stabilized the claim that risk was under control. The apparatus could not hear what it had been designed to ignore. The telling event was a cross functional audit convened for other reasons that uncovered a pattern of exclusions. It did not reveal fraud. It revealed that the categories through which detection worked no longer mapped to the business that the institution had become. Bowker and Star would call this a politics of visibility that had succeeded too well. The unknown was not inside the error count. It had been externalized into the everyday heroics of teams who worked around the system because it could not be made to see what was not on its list (Latour, Bowker and Star).
The literature on disasters offers a final cautionary lens. Diane Vaughan’s account of normalization of deviance in the Challenger launch decision shows how drift can appear as normalcy to people inside whose judgment is bounded by familiar categories and escalating commitments. The ecology I studied was not launching rockets and it was not reckless. It had, however, learned how to follow its own rules so well that divergence from reality was experienced as non event. The repair began by changing what counted as a case. The institution created a new data object for exclusions that required narrative justification and made those justifications queryable. Reviews were required to open with a reading of what the detector refused to see, not to shame the team but to ensure that the ecology could ask whether its ontology still fit the world. Over subsequent quarters the error count rose, the model was retrained, and the esteem of the detector shifted from a symbol of control to a symbol of humility in the presence of a changing environment. The lesson is clear. A metric that does not account for its own blind ground will eventually reward the very forms of non seeing that turn early success into late fragility.
Across the three cases a final reflection gathers force. Knowledge gained under the surveillance of performance is not yet free knowledge. The phrase names a field of worry rather than an indictment. The audit society offers protections that should not be despised, yet it can induce a moral economy in which visibility replaces attention and speed replaces care. Han explains how the moralization of transparency can flatten complexity into immediate exposure, encouraging climates in which careful doubt looks like disloyalty. Scott’s history of legibility shows how simplification can enable coordination while impoverishing perception, which means that every gain in control must be paid for by a loss in nuance somewhere else. The ethical demand, therefore, is not an exit from evaluation. It is the composition of architectures in which evaluation remains answerable to narration, in which reliability is kept teachable by anomaly, and in which success includes the grace to be revised in public without humiliation. When institutions build such architectures, artificial intelligence becomes a partner in collective moral cognition rather than a performance that charms us into forgetting that wisdom is nothing more and nothing less than the practiced willingness to change one’s mind when the world refuses the story we prefer to tell about it.
VI. The Architecture of Reflexive Learning
Having examined how stability, speed, and control can converge into epistemic closure, the question now becomes constructive. How might an institution design for learning that remains faithful to reality rather than to its own reflection. The answer is neither a new algorithm nor a cultural slogan about agility. It is an architecture: a deliberate arrangement of infrastructural conditions that make reflexivity durable, visible, and governable. The architecture of reflexive learning treats knowledge not as the accumulation of conclusions but as the continuous re negotiation of what counts as evidence, what constitutes responsibility, and what enables care. Its purpose is to build coherence that can survive contradiction.
Three interwoven architectures sustain this ecology: epistemic transparency, interpretive circuits, and learning velocity measured by depth. Each can be instantiated technically and institutionally. Each draws from traditions that have already wrestled with the paradox of order that learns—Latour’s inquiry into mediation, Haraway’s ethics of response ability, Byung Chul Han’s critique of transparent coercion, and Nussbaum’s defense of practical wisdom against bureaucratic proceduralism. Together they propose a design practice for institutions that must know without colonizing, must change without disintegrating, and must remain answerable to the world they claim to serve.
1. Epistemic Transparency
Epistemic transparency differs from audit transparency. It does not mean maximal visibility but situated disclosure of what a system takes to be true and what it refuses to see. Donna Haraway’s call for “situated knowledges” established that objectivity arises not from detachment but from the disciplined acknowledgment of position, embodiment, and partial vision (Haraway 583–586). In an institutional setting this principle demands that every model, policy, and process document articulate its ontological assumptions and prohibited zones of automation. Such transparency is ontological rather than optical: it names the kinds of entities a model presumes, the causal relationships it encodes, and the social contexts from which it claims exemption.
This disclosure can be formalized through design. Model cards (Mitchell et al.) and datasheets (Gebru et al.) already provide technical scaffolds for documenting provenance, limitations, and intended use. The architecture extends these genres to include ethical self description. Every analytical artifact must answer three questions: What realities does it presume, which human judgments has it replaced, and under what conditions should it refuse automation. Such documentation transforms governance review from a compliance exercise into a philosophical dialogue. The act of writing forces reflexivity by compelling authors to articulate what they believe about the world. By encoding those beliefs in versioned and queryable form, institutions institutionalize humility. Whitehead’s warning that no system can “embrace the infinite detail of the universe” becomes an operational constraint rather than a metaphysical lament (Whitehead 74).
The virtue of epistemic transparency is its resistance to epistemic hubris. Latour described the modern project as one that simultaneously purifies and hybridizes—claiming rational mastery while depending on unseen mediations (Latour 11–12). Reflexive learning reverses the asymmetry. It brings mediations into discourse so that technical mastery can coexist with moral accountability. The architecture therefore embeds not disclosure for its own sake but disclosure as care: a collective practice of staying with the limits of one’s own seeing.
2. Interpretive Circuits
Interpretive circuits transform exception handling from a defensive routine into the central nervous system of institutional intelligence. Exceptions, dissent, and anomalies are formalized as data objects that reenter learning loops with narrative context intact. This approach is indebted to Kuhn’s observation that paradigms shift when anomalies can no longer be ignored and to Argyris and Schön’s double loop learning, which distinguishes tactical correction from reflective re framing of governing assumptions (Kuhn 66–67; Argyris and Schön 19–23).
In conventional systems anomalies are domesticated by thresholds, queues, and escalation paths that aim to restore order. In a reflexive architecture they are treated as invitations to re describe the world. The design implication is structural: every data pipeline, audit workflow, and policy review must include an interpretive channel parallel to the corrective channel. In the corrective channel, deviation is resolved within the existing frame. In the interpretive channel, selected deviations are re narrated by cross disciplinary teams that include those closest to the phenomenon. Each anomaly is documented as a story—what was expected, what occurred, and what this difference reveals about the institution’s assumptions. These narratives are indexed, searchable, and periodically reexamined for recurring patterns.
The function of interpretive circuits is not romantic openness. It is disciplined conversation with reality. They provide the environment in which dissent becomes method. Latour’s notion of the “parliament of things” becomes literal: the institution convenes a procedural space where the nonconforming event is given representation and voice (Latour 144–45). To prevent drift into endless deliberation, the architecture defines explicit intervals and thresholds for re framing. After a defined frequency of anomalies of a given type, the governing description must be reconsidered. The outcome is a system that learns structurally rather than incidentally. Surprise becomes a designed input.
3. Learning Velocity and Depth
Most institutions measure learning by speed alone: time to adapt, number of iterations per cycle, frequency of model retraining. Reflexive architecture adds a second dimension—depth. Depth measures the degree to which the system revises its own premises. It indexes how many governing assumptions were questioned, how many interpretive communities participated, and how widely the new description propagates. This notion derives from Donella Meadows’s hierarchy of leverage points, where changes in goals and paradigms produce deeper transformation than changes in parameters (Meadows 163–165).
Depth therefore resists the cult of acceleration. Byung Chul Han’s critique of “the society of efficiency” warns that relentless motion can exhaust meaning itself, producing systems that appear dynamic while their categories stagnate (Han 19–21). A reflexive institution redefines velocity as rhythm: the alternation between attention and rest that preserves the capacity for surprise. Time is designed as infrastructure, with periodic sabbaths of non optimization that allow reflection and repair. Martha Nussbaum’s Aristotelian notion of phronesis—practical wisdom born of deliberative pause—grounds this ethics of timing (Nussbaum 302–303).
Practically, learning depth can be measured through governance analytics: the ratio of anomalies that trigger re framing, the diversity of interpreters who contribute to those re framings, and the persistence of curiosity language in review documentation across cycles. Over time, a healthy ecology exhibits slower but more enduring paradigm revisions and fewer crises of description. The metric of success becomes the institution’s sustained ability to live through contradiction without the defensive reflexes of denial or over control.
Toward Reflexive Governance
The integration of these architectures produces an institution capable of reflexive governance—a polity of attention in which humility, responsiveness, and coherence coexist. It is neither a moral utopia nor an algorithmic optimization. It is a practice of collective conscience expressed through design. Every element, from model documentation to meeting cadence, from metric dashboards to audit templates, becomes an artifact of moral philosophy translated into procedure. The system ceases to ask whether it is intelligent in the abstract and begins to ask whether it remains teachable by the world it inhabits.
The result is not fragility. It is a new kind of reliability, one that can hold contradiction without collapsing. Han warned that transparency without opacity leads to burnout; Haraway argued that staying with the trouble means remaining in conversation with what resists control; Latour showed that every network must account for the mediations that make it real. The architecture of reflexive learning operationalizes these insights in the language of design, measurement, and governance. It allows institutions to maintain coherence while continuously changing what they know. It is, in short, a moral technology of revision.
VII. Evaluation and Measurement
Evaluation inside an epistemic ecology cannot be an afterthought appended to delivery. It must be the instrument through which the institution remembers what it values and trains its attention accordingly. If evaluation reduces learning to throughput, then the ecology will optimize for motion while its categories harden. If evaluation gives narrative a free pass from rigor, then sentiment will substitute for care. I therefore design a dual framework in which quantitative traces and qualitative judgments are bound together by explicit theories of change, reported publicly, and kept answerable to revision. The aim is not to invent a new fetish of measurement. The aim is to keep the institution teachable by creating indicators that reward the courage to be revised and by curating narratives that preserve the moral memory of why revision was necessary.
The quantitative spine begins with measures that privilege sensemaking over speed. First, anomaly recognition is counted, but not as a raw total. I track the anomaly elevation rate, the proportion of detected deviations that are formally narrated and brought into deliberation within a defined interval. Second, resolution time is measured not as an instrument of pressure but as a diagnostic of attention and coordination. Median time from detection to deliberation and from deliberation to decision gives a window into how the ecology balances vigilance and care. Third, learning depth is indexed by counting governing assumptions named explicitly in review minutes, by tallying the number of times a decision is unmade in public without retaliation, and by tracking the propagation of re framed descriptions across dependent processes. Donella H. Meadows taught that the deepest leverage points are changes in goals and paradigms rather than parameters, so the index weights revisions of purpose and ontology more heavily than parameter retuning, and a semester of apparent volatility can count as progress when it accompanies a clarified description of the world the institution inhabits (Meadows 163 to 170).
These traces remain fragile if fear governs speech. Amy C. Edmondson’s account of psychological safety shows that people learn when they believe that the social cost of candor will not exceed the institutional value of truth, which means that any evaluation regime that punishes the bearer of bad news will quietly nullify its own indicators (Edmondson 354 to 357). I therefore include a dissent participation index that measures how many distinct roles contribute to interpretive deliberations in a period, and a dissent elevation ratio that compares concerns raised in lower level forums with those that reached governance. A healthy ecology exhibits wide distribution of interpretive voice. James G. March’s analysis of exploration and exploitation also counsels that organizations drift toward exploitation because its rewards are immediate. To counter this drift, I measure exploration effort as the proportion of cycles that allocate time to testing new descriptions or data sources that cannot promise near returns, and I assess the persistence of such effort across quarters so that exploration is not sacrificed when pressure rises (March 71 to 73).
Because measurement can induce the very blindness it seeks to cure, two families of countermeasures are necessary. The first is transparency with situated humility. Model cards and review templates require explicit fields that disclose ontological commitments, exclusions from automation, and retirement conditions. The presence of these fields is counted, but more importantly their content is sampled for specificity and argumentative clarity. Donna J. Haraway’s insistence that objectivity is situated informs the audit rubric, which rewards authors who mark the limits of their vantage and the reasons for refusal, while refusing the aesthetic of a view from nowhere that is the precondition of moral evasion (Haraway 583 to 588). The second family is an audit of the unknowns. Michael Power and Marilyn Strathern warned that regimes of audit can convert visibility into virtue and render care subordinate to display. To resist that conversion, the architecture obliges each review to begin with the reading of the exclusions ledger, a register of cases and contexts the system has refused to see, with narrative justifications and open questions preserved. The presence of an unknowns ledger is tracked, but its discipline is enacted by the ritual of reading, since what is unknown can only discipline the institution if it enters speech and is allowed to attract responsibility (Power 12 to 27; Strathern 310 to 314).
Quantitative traces earn their authority only when braided with qualitative judgment. I therefore institutionalize narrative evaluation as a first class practice. In reflective debriefs, teams construct short case histories of anomalies that changed their mind, drawing on Karl E. Weick’s understanding that organizations stabilize themselves through plausible accounts and that such accounts can enlarge or constrict the future depending on how they handle surprise (Weick 61 to 62, 128 to 133). Narratives are not free form. They are written to a schema that requires identification of the expectation that failed, the evidence that contradicted it, the human judgments displaced by automation, the reasons for any change in description, and the harms that were prevented or absorbed by the decision. These texts are archived in a searchable commons and sampled for periodic qualitative review by a rotating panel of practitioners and ethicists who annotate them for humility language, care for context, and clarity about costs. The annotations are themselves preserved, thereby creating a lineage of moral reasoning that future teams can inhabit rather than reconstruct.
The tension between reliability and revelation demands metrics that protect safety without strangling agency. I therefore separate throughput indicators, which keep the machine dependable, from epistemic freedom indicators, which keep it open to being taught by the world. Throughput indicators include service levels, stability bands, and audit closure rates. Epistemic freedom indicators include anomaly elevation rate, dissent participation, assumption articulation, unmaking decisions in public, and the persistence of curiosity language across cycles. Curiosity is operationalized by a lexicon of interrogative phrases and uncertainty markers that reviewers are trained to use. Natural language analysis on review minutes produces a curiosity persistence score, which is used not to punish writers but to remind the institution that a culture’s rhetoric is one of its infrastructures of attention. Amartya Sen’s conception of freedom as capability provides the ethical horizon for the synthesis. The question is always whether the ecology is enlarging its real opportunities to choose reasoned action, not whether it is generating more output under constraint (Sen 36 to 41).
The framework also requires temporal discipline. Many reforms produce a transient uplift in vigilance that fades as novelty passes. To guard against this decay I employ interrupted time series designs and cohort comparisons that examine whether gains in elevation rate, participation diversity, and assumption articulation persist beyond initial enthusiasm. I pair these designs with shadow measures that are resistant to performative reporting. For example, I track the fraction of anomalies that originate from frontline roles rather than from compliance functions, and the fraction of exclusions that later become formal categories, since both reveal whether the ecology is able to learn from its periphery rather than only from its center. Michael Polanyi’s reminder that we know more than we can tell motivates a further check. In periodic interviews, practitioners are asked to recount a moment in which they chose not to escalate surprise. These accounts are coded for reasons and costs, then reported back to leadership as part of an affect audit that measures the price of candor inside the current climate (Polanyi 4 to 7).
A final layer concerns the ethics of evaluation itself. Byung Chul Han warns that transparency moralized into ideology flattens complexity into immediate visibility. I therefore encode opacity as a protected space under conditions of accountability. Periodic sabbath intervals suspend throughput metrics to make room for archival curation, narrative analysis, and ontological review. Martha C. Nussbaum’s account of practical wisdom legitimates such intervals, since discernment needs time to ripen and justice can require the patience to listen before deciding. These sabbaths are not indulgences. They are infrastructural commitments that prevent the ecology from devouring its own attention and that allow deeper paradigm revisions to mature without panic or spectacle (Han 8 to 12; Nussbaum 302 to 303).
Skeptics will ask whether such a regime burdens already stretched teams and whether it risks a new performativity of virtue. The answer is that any evaluation is already a moral pedagogy, whether acknowledged or not. If the institution must live under metrics, let the metrics teach what matters. To guard against sanctimony, I keep indicators few, open to critique, and tethered to narrative checks that can embarrass metrics when they drift from meaning. The point is not to create an economy of humility tokens but to build a culture in which a change in description is experienced as a collective achievement rather than as a shameful admission of failure. When evaluation is composed in this way, artificial intelligence becomes accountable not only for what it predicts but for the reasons it offers for refusing prediction. Governance becomes a craft of attention rather than a theatre of control. Reliability remains, but it is joined to a renewed capacity for wonder, which is the only reliable engine of institutional learning that does not collapse into either paralysis or domination.
VIII. The Ethics of Institutional Freedom
Institutional freedom is not the absence of constraint. It is the cultivated capacity of a large system to choose reasoned action in the presence of competing goods, to authorize doubt without paralysis, and to revise decisions without humiliation. I treat freedom as an infrastructural property rather than a personal attribute, a quality of the ecology that enables persons and tools to act from understanding rather than from compliance alone. Amartya Sen argues that freedom becomes real when people have substantive capabilities to do and to be what they have reason to value, not only formal permissions that remain inert under pressure. If we translate this insight from individuals to institutions, we find that an organization is free to the extent that it maintains capabilities for attention, deliberation, dissent, and revision, so that policy and model can be unmade when they no longer fit the world, and so that those who speak from the margin can be heard before crisis requires them to shout (Sen 36 to 41, 291 to 298). Freedom here does not abolish reliability. It teaches reliability to remain answerable to truth by preserving space for to be otherwise when reality instructs.
Such an ethic requires that we distinguish liberty from license. Isaiah Berlin warned that liberty collapses when the term is stretched to justify domination in the name of supposedly higher goods, a warning that resonates in institutions that invoke mission to silence dissent. Berlin’s negative and positive liberties are not a menu of preferences but a hard lesson about the ease with which the rhetoric of collective purpose can be used to coerce under the name of reason. An ecology that aspires to be free must protect zones where persons can refuse what seems efficient and where models can refuse automation, not because refusal is an end in itself, but because refusal preserves the capacity to keep judgment alive in the face of seductions that would make thinking unnecessary (Berlin 168 to 170, 203 to 206). Philip Pettit sharpens this point by defining freedom as non domination, the condition under which one is not subject to the arbitrary will of another. The institutional analogue is a governance architecture in which discretionary power is visible, contestable, and bound to public reasons, so that no team or tool can impose a description of the world that others cannot challenge with standing and without fear (Pettit 22 to 27, 55 to 61).
Because institutions are not sovereign minds but negotiated orders, freedom must be organized through forms that distribute authority while preventing capture. Elinor Ostrom’s work on polycentric governance demonstrates that commons can be stewarded when rules are tailored to local conditions, participation is real rather than theatrical, and monitoring is mutual rather than purely external. Her core lesson is methodological and ethical at once. Durable freedom arises when those affected by rules can shape them and can revise them in the light of experience, and when monitors are accountable to those they monitor rather than to distant optics alone. An institution that treats learning as a commons can therefore encode rule changing protocols that are accessible to those closest to the phenomena, so that revision is not an elite privilege but a property of the ecology itself (Ostrom 88 to 102, 135 to 142). This is neither romantic decentralization nor bureaucratic flattening. It is the practical craft of aligning description with experience under conditions of scale.
Freedom inside the ecology must also be protected against the moralization of immediacy. Byung Chul Han shows how the ideology of transparency compresses time and flattens complexity into instant visibility, generating a culture in which delayed judgment is treated as evasion. When an institution adopts this moral climate, it punishes the very pauses that discernment requires and it rewards speed as a proxy for courage. The ethic proposed here refuses that compression by encoding intervals of protected opacity under accountability, what I earlier named an epistemic sabbath. During such intervals throughput metrics are suspended, archives are curated, and ontologies are reviewed in quiet speech. Martha C. Nussbaum’s reading of Aristotelian practical wisdom legitimates these practices, since fitting action in the concrete case cannot be produced by rules alone and requires time to see what is at stake for the particular lives involved (Han 8 to 12; Nussbaum 302 to 305). A sabbath is not a luxury addendum to performance. It is the condition under which performance remains human and true.
The ethic also concerns voice. Hannah Arendt argued that freedom appears in the world when people act and speak in concert, bringing something new into a common space where it can be seen, judged, and taken up. In institutional life the common space is easily colonized by performance dashboards and by risk scripts that pre narrate what may be said. Under such conditions speech ceases to be action and becomes compliance theater. If freedom is to be more than rhetoric, the institution must rebuild the publicness of its deliberations so that reasons are offered and contested, so that responsibility has a face, and so that a decision can be unmade with gratitude rather than with scapegoating. The practice does not romanticize process. It restores the political dignity of judgment by ensuring that those who will bear the costs of a decision have the standing to speak into its formation and its revision (Arendt 175 to 181, 193 to 198).
Because artificial intelligence now mediates much of institutional judgment, the ethic of freedom must also reach into the design of technical artifacts. Donna Haraway has taught that objectivity arises from situated accountability, not from disembodied gaze. A free institution therefore requires its models to confess their vantages, to mark prohibited zones of automation, and to offer reasons for refusal alongside reasons for prediction. Such confession is not academic etiquette. It is the ground on which contestation can proceed without humiliation, since those who disagree can address assumptions rather than attack persons. The aim is to construct a moral grammar in which technical power is domesticated by care and where care is disciplined by exposure to counterexample and to cost (Haraway 583 to 589).
A skeptic will insist that freedom at scale endangers reliability, that protected intervals invite drift, that participatory revision encourages faction, and that voice will be captured by the articulate rather than by the affected. These cautions are real. Berlin warned that positive liberty can disguise coercion, and Pettit observed that domination can hide in forms that satisfy surface fairness while enabling arbitrary interference. The remedy is not to abandon freedom, but to build counterweights that make its pathologies visible. Polycentric review bodies can be obligated to include lay representation and to publish minority reports. Anomaly forums can be required to track who speaks and who is silent, and to rotate facilitation so that no style of speech becomes the currency of legitimacy. Sabbaths can be scheduled and enforced by those who are not evaluated by throughput. Above all, revision rights can be protected as rights rather than as favors, so that the capacity to change one’s mind is not hostage to mood or to calendar (Berlin 203 to 213; Pettit 80 to 89; Ostrom 160 to 166).
The ethic culminates in a single stance that can be taught and measured. To design for freedom is to encode space for principled disobedience that educates the system itself. Principled disobedience is not defiance for its own sake. It is the protected act by which a person or a team refuses to comply with a description of the world that has become untrue, and offers a reasoned counter description that can be adjudicated in public. In such moments institutions either shrink into control or widen into learning. Sen’s language of capability offers the evaluative lens. The question to ask of any reform is not whether it increases motion or even whether it increases accuracy in the short term. The question is whether it enlarges the real opportunities that people and tools have to bring the world into speech and to act upon that speech without fear. When the answer is yes, the system becomes more reliable precisely because it has become more free, since its reliability now includes the courage and the means to be revised by what it did not expect to find (Sen 284 to 299).
The ethic I have described does not excuse harm under the banner of exploration, nor does it sanctify exploration against safety. It seeks a union of reliability and revelation by composing infrastructures that keep each answerable to the other. Pettit’s non domination ensures that reliability does not become arbitrary rule. Berlin’s warning ensures that purpose does not silence plurality. Ostrom’s design principles ensure that participation is real and corrigible. Haraway’s situated objectivity ensures that technical description remains humble and discussable. Nussbaum’s practical wisdom ensures that timing and attention are protected against the tyranny of immediacy. Sen’s capability approach ensures that our metrics and rituals finally serve the enlargement of responsible choice. Under these conditions freedom becomes a property of the institution’s architecture rather than a slogan. It becomes the patient strength by which a system can keep faith with reality without losing coherence, which is to say the form of intelligence that deserves our trust.
IX. Planetary and Historical Contexts
Every institution that claims to learn does so within a cosmos of meaning and matter that it neither built nor controls. To understand institutional learning within this larger horizon is to abandon the fiction that governance can remain a merely human enterprise. Systems of intelligence, policy, and measurement are themselves participants in planetary metabolism: they extract, compute, discard, and thereby contribute to the ecological and historical conditions that constrain their own continuance. The question of how an institution learns therefore cannot remain epistemological alone. It is ontological and ecological, concerned with how human systems integrate with the rhythms of the earth that hosts them. In this sense, institutional ignorance at scale mirrors the cosmological amnesia of modernity itself. It arises from the same metaphysical mistake that detached mind from matter and progress from place, the same error that Sallie McFague identified when she warned that the modern imagination has “unhoused itself from creation,” forgetting that the world is not a backdrop but a body in which all our institutions are organs (McFague 72–74).
To recover learning as an ecological act is to re situate intelligence within planetary time. Dipesh Chakrabarty has argued that history is now bifocal, both global and planetary: global in its human entanglements of capital, labor, and technology; planetary in its subjection to earth systems that set boundaries to all human projects (Chakrabarty, The Climate of History in a Planetary Age 3–7, 45–52). The planetary enters history not as scenery but as actor, interrupting our managerial chronologies with feedback loops that no quarterly schedule can tame. The Anthropocene thus reveals a deeper epistemic crisis: institutions built for prediction and control must now learn responsiveness and repentance. Their task is not to master the planet but to listen to it, to translate its feedback into forms of governance capable of humility. The institution that learns under planetary conditions must therefore cultivate what McFague called “theological imagination as ecological discipline,” a disposition that regards knowledge as a practice of belonging rather than of domination (McFague 88–93).
Timothy Morton gives this belonging its phenomenological texture. His notion of the hyperobject—entities so massively distributed across time and space that they exceed all local grasp—forces institutions to confront the limits of their cognition (Morton 1–5, 17–22). Climate change, ocean acidification, atmospheric carbon—all are hyperobjects that cannot be bounded by any one jurisdiction or analytic frame. They trouble every epistemology premised on control, because they make visible the gap between data and experience, between measurement and meaning. In such encounters, intelligence becomes an act of surrender: to know is not to command but to remain teachable by what one can never fully map. A reflexive architecture of learning, as developed in this dissertation, is thus not only an organizational reform but a planetary pedagogy. It is a means by which institutions can practice what Morton calls “coexistence as intimacy,” designing systems that remain permeable to feedback from realities that refuse simplification (Morton 82–86).
Planetary consciousness also reveals the moral poverty of instrumental reason when divorced from situated life. James C. Scott’s analysis of legibility demonstrates that projects of modernization have long sought to make the world administratively visible—forests turned into grids, people into statistics, ecosystems into yield curves. The unintended consequence is a blindness to local complexity and to forms of knowledge that do not fit the template of rational management (Scott 11–23, 262–307). Under planetary conditions this blindness becomes catastrophic, as the simplifications that once facilitated order now undermine the biosphere itself. Institutional learning must therefore recover what Scott calls metis, the practical wisdom of the particular, and integrate it into the architectures of decision making. Only then can institutions achieve what McFague describes as “loving perception,” a disciplined gaze that refuses to abstract life into resource and that measures success by fidelity to the living world rather than by efficiency alone (McFague 106–109).
Donna Haraway’s notion of “response-ability” gives this ecological fidelity its procedural form. To be response able is to enter into patterned relations of care with those one cannot fully know or control, to remain available for correction by the more-than-human others who co constitute one’s existence (Haraway, Staying with the Trouble 14–19). Institutions that practice response ability redesign their interpretive circuits not only for internal anomaly but for planetary feedback. Data about climate impact, resource extraction, and social vulnerability become moral interlocutors rather than compliance metrics. Haraway’s vision translates into a concrete governance principle: systems must be designed to receive contradiction from the world as grace rather than as threat. Such architectures move beyond anthropocentric accountability toward what McFague called “a theology of participation,” where the goal is not mastery of the body of the world but communion within it (McFague 115–119; Haraway 41–44).
Bruno Latour’s later political ecology radicalizes this participation by collapsing the modern distinction between nature and society. “We have never been modern,” he writes, because our sciences and our politics have always co-produced each other through networks of mediation (Latour, We Have Never Been Modern 103–106). The implication is that institutional intelligence must now be designed as diplomacy among heterogeneous actors—human, technological, and ecological. Latour’s “parliament of things” becomes the design template for a planetary institution: a procedural space in which nonhuman stakes are represented, where long timescales and material agencies are given standing, and where decisions are treated as negotiations among worlds. Learning in this register is no longer adaptation alone but reconciliation—the ongoing labor of composing a common world among entities that cannot share a language but can share consequences (Latour, Facing Gaia 15–21, 97–101).
The anthropology of survival proposed by Anna Tsing clarifies the microphysics of such reconciliation. In The Mushroom at the End of the World, Tsing reveals how life persists amid ruins, how coordination emerges without command, and how attention to peripheral ecologies offers a realism unavailable to centralized management (Tsing 1–6, 131–135). Her method—an art of noticing—should be institutionalized as a principle of design. Reflexive learning requires attunement to marginal signals, to the quiet data of repair that never reach executive dashboards. Tsing’s salvage epistemology suggests that the future of intelligence lies not in omniscient integration but in polyphonic assembly: systems that can listen to multiple temporalities at once, weaving local fragments into adaptive coherence. Institutions that build this capacity cease to be fortresses of control and become instead habitats of relation.
Historical consciousness must accompany this ecological humility. Modernity’s moral exhaustion—its exhaustion of both the planet and its own narratives of progress—reveals that institutions can no longer afford innocence about the costs of their coherence. Every structure of stability has been purchased by some externalized fragility, every metric of success by some zone of uncounted loss. Chakrabarty insists that planetary thinking must hold together both justice and survival: that we cannot choose between human rights and earth rights, because the two are now entwined in the same metabolism. The institutional ethic of learning must therefore unite reliability with repentance, progress with restitution. A truly learning organization is one that remembers its own complicity and designs for atonement through repair—an echo of McFague’s claim that love of the world must take institutional form in economies of restraint and redistributive attention (Chakrabarty, The Climate of History in a Planetary Age 167–172; McFague 154–160).
Such an institution would measure itself not by the volume of its outputs but by the grace of its corrections. It would treat carbon accounting, social auditing, and resource transparency not as reputational tasks but as confessions before the body of the world. It would keep an archive of its own learning, a moral ecology of revisions through which future generations could trace how knowledge evolved in the presence of error. It would learn to think with the temporality of soil, river, and species—slow enough to remember, fast enough to care. Whitehead’s dictum that “intelligence is rhythm” gains here its planetary resonance: rhythm as the alternation between operation and listening, between exploitation and reverence (Whitehead 74–76). The institution that sustains such rhythm participates in what McFague called “the conversion of the industrial mind,” a transformation from consumption to communion (McFague 163–167).
In this sense, planetary and historical learning converge on a single imperative: institutions must become organs of the earth’s own intelligence. They must extend the planet’s capacity for reflection rather than for extraction, and they must treat their metrics, models, and rituals as forms through which the earth itself can think with us. When this alignment is achieved, learning ceases to be a managerial virtue and becomes a sacramental act—the world knowing itself through our willingness to be changed by it.
Works Cited
Chakrabarty, Dipesh. The Climate of History in a Planetary Age. University of Chicago Press, 2021.
Haraway, Donna J. Staying with the Trouble: Making Kin in the Chthulucene. Duke University Press, 2016.
Latour, Bruno. Facing Gaia: Eight Lectures on the New Climatic Regime. Polity Press, 2017.
—. We Have Never Been Modern. Translated by Catherine Porter, Harvard University Press, 1993.
McFague, Sallie. The Body of God: An Ecological Theology. Fortress Press, 1993.
Morton, Timothy. Hyperobjects: Philosophy and Ecology after the End of the World. University of Minnesota Press, 2013.
Scott, James C. Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed. Yale University Press, 1998.
Tsing, Anna Lowenhaupt. The Mushroom at the End of the World: On the Possibility of Life in Capitalist Ruins. Princeton University Press, 2015.
Whitehead, Alfred North. Process and Reality: An Essay in Cosmology. Corrected edition, edited by David Ray Griffin and Donald W. Sherburne, Free Press, 1978.
Arendt, Hannah. The Human Condition. Second edition, University of Chicago Press, 1998.
Berlin, Isaiah. “Two Concepts of Liberty.” Four Essays on Liberty. Oxford University Press, 1969, pp. 118 to 172, 173 to 220.
Han, Byung Chul. The Transparency Society. Translated by Erik Butler, Stanford University Press, 2015.
Haraway, Donna J. “Situated Knowledges, The Science Question in Feminism and the Privilege of Partial Perspective.” Feminist Studies, vol. 14, no. 3, 1988, pp. 575 to 599, https://doi.org/10.2307/3178066.
Nussbaum, Martha C. The Fragility of Goodness, Luck and Ethics in Greek Tragedy and Philosophy. Revised edition, Cambridge University Press, 2001.
Ostrom, Elinor. Governing the Commons, The Evolution of Institutions for Collective Action. Cambridge University Press, 1990.
Pettit, Philip. Republicanism, A Theory of Freedom and Government. Oxford University Press, 1997.
Sen, Amartya. Development as Freedom. Knopf, 1999.
Edmondson, Amy C. “Psychological Safety and Learning Behavior in Work Teams.” Administrative Science Quarterly, vol. 44, no. 2, 1999, pp. 350 to 383, https://doi.org/10.2307/2666999.
Han, Byung Chul. The Transparency Society. Translated by Erik Butler, Stanford University Press, 2015.
Haraway, Donna J. “Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective.” Feminist Studies, vol. 14, no. 3, 1988, pp. 575 to 599, https://doi.org/10.2307/3178066.
March, James G. “Exploration and Exploitation in Organizational Learning.” Organization Science, vol. 2, no. 1, 1991, pp. 71 to 87, INFORMS, https://doi.org/10.1287/orsc.2.1.71.
Meadows, Donella H. Thinking in Systems, A Primer. Chelsea Green Publishing, 2008.
Nussbaum, Martha C. The Fragility of Goodness, Luck and Ethics in Greek Tragedy and Philosophy. Revised edition, Cambridge University Press, 2001.
Polanyi, Michael. The Tacit Dimension. University of Chicago Press, 2009.
Power, Michael. The Audit Society, Rituals of Verification. Oxford University Press, 1997.
Sen, Amartya. Development as Freedom. Knopf, 1999.
Strathern, Marilyn. “The Tyranny of Transparency.” British Educational Research Journal, vol. 26, no. 3, 2000, pp. 309 to 321, https://doi.org/10.1080/713651562.
Weick, Karl E. Sensemaking in Organizations. Sage Publications, 1995.
Argyris, Chris, and Donald A. Schön. Organizational Learning: A Theory of Action Perspective. Addison Wesley, 1978.
Gebru, Timnit, et al. “Datasheets for Datasets.” Communications of the ACM, vol. 64, no. 12, 2021, pp. 86–92, https://doi.org/10.1145/3458723.
Han, Byung-Chul. The Burnout Society. Translated by Erik Butler, Stanford University Press, 2015.
Haraway, Donna J. “Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective.” Feminist Studies, vol. 14, no. 3, 1988, pp. 575–599, https://doi.org/10.2307/3178066.
Kuhn, Thomas S. The Structure of Scientific Revolutions. Second edition, enlarged, University of Chicago Press, 1970.
Latour, Bruno. Reassembling the Social: An Introduction to Actor Network Theory. Oxford University Press, 2005.
Meadows, Donella H. Thinking in Systems: A Primer. Chelsea Green Publishing, 2008.
Mitchell, Margaret, et al. “Model Cards for Model Reporting.” Proceedings of the Conference on Fairness, Accountability, and Transparency, 2019, pp. 220–229, https://doi.org/10.1145/3287560.3287596.
Nussbaum, Martha C. The Fragility of Goodness: Luck and Ethics in Greek Tragedy and Philosophy. Revised edition, Cambridge University Press, 2001.
Whitehead, Alfred North. Process and Reality: An Essay in Cosmology. Corrected edition, edited by David Ray Griffin and Donald W. Sherburne, Free Press, 1978.
Argyris, Chris, and Donald A. Schön. Organizational Learning, A Theory of Action Perspective. Addison Wesley, 1978.
Bowker, Geoffrey C., and Susan Leigh Star. Sorting Things Out, Classification and Its Consequences. MIT Press, 1999.
Edmondson, Amy C. “Psychological Safety and Learning Behavior in Work Teams.” Administrative Science Quarterly, vol. 44, no. 2, 1999, pp. 350 to 383, https://doi.org/10.2307/2666999.
Han, Byung Chul. The Transparency Society. Translated by Erik Butler, Stanford University Press, 2015.
Kuhn, Thomas S. The Structure of Scientific Revolutions. Second edition, enlarged, University of Chicago Press, 1970.
Latour, Bruno. Reassembling the Social, An Introduction to Actor Network Theory. Oxford University Press, 2005.
March, James G. “Exploration and Exploitation in Organizational Learning.” Organization Science, vol. 2, no. 1, 1991, pp. 71 to 87, INFORMS, https://doi.org/10.1287/orsc.2.1.71.
Meadows, Donella H. Thinking in Systems, A Primer. Chelsea Green Publishing, 2008.
Polanyi, Michael. The Tacit Dimension. University of Chicago Press, 2009.
Power, Michael. The Audit Society, Rituals of Verification. Oxford University Press, 1997.
Scott, James C. Seeing Like a State, How Certain Schemes to Improve the Human Condition Have Failed. Yale University Press, 1998.
Star, Susan Leigh, and Karen Ruhleder. “Steps toward an Ecology of Infrastructure, Design and Access for Large Information Spaces.” Information Systems Research, vol. 7, no. 1, 1996, pp. 111 to 134, https://doi.org/10.1287/isre.7.1.111.
Strathern, Marilyn. “The Tyranny of Transparency.” British Educational Research Journal, vol. 26, no. 3, 2000, pp. 309 to 321, https://doi.org/10.1080/713651562.
Vaughan, Diane. The Challenger Launch Decision, Risky Technology, Culture, and Deviance at NASA. University of Chicago Press, 1996.
Weick, Karl E. Sensemaking in Organizations. Sage Publications, 1995.
Argyris, Chris, and Donald A. Schön. Organizational Learning: A Theory of Action Perspective. Addison Wesley, 1978.
Bowker, Geoffrey C., and Susan Leigh Star. Sorting Things Out: Classification and Its Consequences. MIT Press, 1999.
Gebru, Timnit, et al. “Datasheets for Datasets.” Communications of the ACM, vol. 64, no. 12, 2021, pp. 86–92, https://doi.org/10.1145/3458723.
Haraway, Donna J. “Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective.” Feminist Studies, vol. 14, no. 3, 1988, pp. 575–599, https://doi.org/10.2307/3178066.
Latour, Bruno. Reassembling the Social: An Introduction to Actor Network Theory. Oxford University Press, 2005.
Meadows, Donella H. Thinking in Systems: A Primer. Chelsea Green Publishing, 2008.
Mitchell, Margaret, et al. “Model Cards for Model Reporting.” Proceedings of the Conference on Fairness, Accountability, and Transparency, 2019, pp. 220–229, https://doi.org/10.1145/3287560.3287596.
Polanyi, Michael. The Tacit Dimension. University of Chicago Press, 2009.
Schön, Donald A. The Reflective Practitioner: How Professionals Think in Action. Basic Books, 1983.
Sen, Amartya. Development as Freedom. Knopf, 1999.
Smith, Dorothy E. Institutional Ethnography: A Sociology for People. AltaMira Press, 2005.
Suchman, Lucy A. Plans and Situated Actions: The Problem of Human Machine Communication. Cambridge University Press, 1987.
Argyris, Chris, and Donald A. Schön. Organizational Learning: A Theory of Action Perspective. Addison Wesley, 1978.
Bateson, Gregory. Steps to an Ecology of Mind. University of Chicago Press, 2000.
Bowker, Geoffrey C., and Susan Leigh Star. Sorting Things Out: Classification and Its Consequences. MIT Press, 1999.
Han, Byung Chul. The Transparency Society. Translated by Erik Butler, Stanford University Press, 2015.
Haraway, Donna J. “Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective.” Feminist Studies, vol. 14, no. 3, 1988, pp. 575–599, https://doi.org/10.2307/3178066.
Kuhn, Thomas S. The Structure of Scientific Revolutions. Second edition, enlarged, University of Chicago Press, 1970.
March, James G. “Exploration and Exploitation in Organizational Learning.” Organization Science, vol. 2, no. 1, 1991, pp. 71–87. INFORMS, https://doi.org/10.1287/orsc.2.1.71.
Meadows, Donella H. Thinking in Systems: A Primer. Chelsea Green Publishing, 2008.
Nussbaum, Martha C. The Fragility of Goodness: Luck and Ethics in Greek Tragedy and Philosophy. Revised edition, Cambridge University Press, 2001.
Suchman, Lucy A. Plans and Situated Actions: The Problem of Human Machine Communication. Cambridge University Press, 1987.
Weick, Karl E. Sensemaking in Organizations. Sage Publications, 1995.
Whitehead, Alfred North. Process and Reality: An Essay in Cosmology. Corrected edition, edited by David Ray Griffin and Donald W. Sherburne, Free Press, 1978.
Argyris, Chris, and Donald A. Schön. Organizational Learning: A Theory of Action Perspective. Addison-Wesley, 1978.
Han, Byung-Chul. The Transparency Society. Translated by Erik Butler, Stanford University Press, 2015.
Haraway, Donna J. Staying with the Trouble: Making Kin in the Chthulucene. Duke University Press, 2016.
Kuhn, Thomas S. The Structure of Scientific Revolutions. Second edition, enlarged, University of Chicago Press, 1970.
March, James G. “Exploration and Exploitation in Organizational Learning.” Organization Science, vol. 2, no. 1, 1991, pp. 71–87. INFORMS, https://doi.org/10.1287/orsc.2.1.71.
Meadows, Donella H. Thinking in Systems: A Primer. Chelsea Green Publishing, 2008.
Polanyi, Michael. The Tacit Dimension. University of Chicago Press, 2009.
Sen, Amartya. Development as Freedom. Knopf, 1999.
Weick, Karl E. Sensemaking in Organizations. Sage Publications, 1995.
Haraway, Donna J. Staying with the Trouble: Making Kin in the Chthulucene. Duke UP, 2016.
Han, Byung-Chul. The Transparency Society. Translated by Erik Butler, Stanford UP, 2015.
Latour, Bruno. We Have Never Been Modern. Translated by Catherine Porter, Harvard UP, 1993.
Meadows, Donella H. Thinking in Systems: A Primer. Chelsea Green Publishing, 2008.
Nussbaum, Martha C. The Fragility of Goodness: Luck and Ethics in Greek Tragedy and Philosophy. Revised ed., Cambridge UP, 2001.
Sen, Amartya. Development as Freedom. Knopf, 1999.
Simondon, Gilbert. On the Mode of Existence of Technical Objects. Translated by Cecile Malaspina and John Rogove, Univocal Publishing, 2017.
Whitehead, Alfred North. Process and Reality: An Essay in Cosmology. Corrected ed., edited by David Ray Griffin and Donald W. Sherburne, Free Press, 1978.
Jonathan Johnson-Swagel 
Leave a comment