I. Historiography of the Outlier
The figure who does not average has been preserved, policed, and sometimes protected by archives that carry not only memory but procedures of recognition, and a historiography of the outlier therefore begins by reading genres as instruments that either widen the room for singular persons or close it, since rules, letters, pamphlets, settlement ledgers, broadcast transcripts, and memorial charters each encode a practical verdict about how an uncommon life should be received, taught, or contained. The Rule of Benedict names a decisive early posture when it commands that every guest be received as Christ, a formulation that turns hospitality into a norm for institutions rather than a sporadic virtue of individuals, and that places the burden of adjustment on the house rather than the stranger, thereby establishing an archival sentence in which difference arrives with dignity and the community trains itself to meet it with ritualized care rather than suspicion or spectacle.
If the monastic rule is a template for institutional hospitality, the Franciscan sources record a second modality in which an unprotected life becomes a public school for attention, since the Testament of Francis narrates poverty and penance as a pedagogy offered to the towns that surrounded the friars, a witness that made dispossession speak to the civic economy by refusing the conversion of charisma into domination and by modeling a portable rule of regard. In the twelfth century correspondence of Hildegard of Bingen we encounter yet another archive of outlier reception, because her letters travel between abbeys, bishops, and princes to convene a community by the circulation of counsel, correction, and consolation, and thus reveal that singular authority can become a distributed network of discernment in which the unusual voice is not quarantined as a curiosity but routed into governance through repeated epistolary labor that leaves a durable trail for historians to read.
Modern popular archives complicate any simple tale of reverent reception, since the abolitionist newspaper and the settlement house report embody two different ways that publics learn to see those pushed to the margins, with The Liberator training readers to recognize persons endangered by the law and to organize material action on their behalf, and Hull House translating neighborhood life into maps, studies, and narratives that would make municipal policy intelligible to ordinary citizens while keeping the names and textures of the West Side present in the administrative imagination. Jane Addams’s own autobiographical account anchors this archive with a voice that keeps the ledger and the human scene in the same frame, which is a methodological commitment that this essay will keep, since claims about care that cannot be traced to budgets and rooms remain too empty to instruct and too volatile to last.
The twentieth century extends this public pedagogy through mass media and then through memory institutions that attempt to repair what the spectacle deformed, and the transcript of Fred Rogers’s 1969 testimony before the United States Senate provides a canonical instance of a practitioner translating intimate craft into durable public funding, since his appeal converts a pedagogy of naming feelings into a civic rationale for an infrastructure that can carry care into millions of homes without requiring the child to become average in order to belong. Bryan Stevenson’s creation of the Legacy Sites in Montgomery, which include a national memorial to victims of racial terror lynching and a museum that threads slavery to mass incarceration, gives the historian a paired archive of architecture and text in which outlier commemoration becomes a standing institutional practice that binds mourning to law and thereby changes the way a city teaches itself to live with its dead and with the survivors who were made outliers by design.
From these heterogeneous repositories a set of historiographic consequences follows. First, the outlier appears in records when a community disciplines its attention through repeatable forms rather than through admiration alone, which is why rules, letters, ledgers, and hearing transcripts matter, because they convert the fragile encounter into an iterable practice that can be audited, taught, and repaired. Second, the same archives that make outliers visible can also normalize them into disappearance if the genres are allowed to enforce smoothness without remainder, so the historian must read not only for presence but for the mechanism by which a document repositions a person inside an administrative grid and calls that repositioning care. Third, survivorship bias is endemic to the genre of edifying life, which is why this essay will pair affirming case modules with mixed or negative cases and will insist that any transferable principle be stated as a constrained mechanism rather than as a moral emanation.
The materials named above also specify a method. When the Rule of Benedict instructs a porter to keep watch so that the stranger is never ignored, the instruction is actionable and role specific, which suggests that historical claims about outlier reception should be tested against the distribution of roles and the durability of the work they require, and we will therefore treat every case as a triangle whose vertices are voice, institution, and ledger. When a settlement house transforms social perception into maps and papers, the archive demonstrates that description can be a form of ethical engineering rather than a prelude to prescription, and we will therefore keep description thick enough to resist the slide into a single number portrait that would erase tails and call the erasure fairness. When a memorial enumerates names and counties and invites local communities to adopt a duplicate pillar, the archive reveals a pedagogy of co responsibility in which the institution refuses to hoard moral work, so our method will ask at every point whether the care on display is portable without the founder and whether the design invites others into accountable maintenance rather than passive admiration.
The scope of this section has necessarily privileged instances from Latin Christianity, Progressive Era Chicago, American broadcast media, and contemporary civil rights memory work, and a global historiography would require parallel readings in Sufi letters, Buddhist monastic codes, West African epics, and Latin American base community archives, which this project acknowledges as essential continuations rather than ornamental expansions, and which will be pursued in subsequent studies with the same insistence on primary sources, institutional specificity, and measurable consequences. The selection criteria used here are explicit in order to reduce the romance of exception that so often attaches to extraordinary lives, since we choose archives where the unusual person is held in frame alongside rules, rooms, budgets, and publics, and we decline cases that offer baroque anecdotes without institutional trace.
Finally, this historiography claims that love appears in history as disciplined attention that changes how rooms are arranged and how ledgers are written, not as an inward state separable from structure, and therefore the historian of the outlier must remain accountable both to the singular narrative and to the administrative text that either protects or erases that singularity, which is why the sections that follow will keep voice, institution, and measure together, will mark where admiration dissolves into consumption, and will treat transferability not as a flourish but as the ethical test of any example that aspires to teach.
II. Genealogy and governance of the average
A. From observatory to social body
The average begins as a tool for noisy instruments, then migrates to a vision of the person and the people. Adolphe Quetelet proposed l’homme moyen as an empirical norm, a center around which variation could be described and managed; his English Treatise framed statistics as a science of normal man rather than a catalog of irregulars. This is the moment a descriptive convenience begins to look like a portrait.
Francis Galton converts that portrait into heredity and policy. Regression toward mediocrity promises a pull to the center; correlation and rank order become tools for sorting people; eugenics supplies a program that treats deviation as a reproductive threat. The mathematics is not innocent here, since the same papers that define the curve also recruit it to a social project that punishes tails.
The point is not to indict calculation. It is to name the channel by which a summary of measurements hardens into a rule for persons. Once the average is pictured as human nature rather than as a statistic, the step from curve to clinic or court becomes thinkable. Buck v. Bell is the blunt proof that a quantified norm, once moralized, becomes a sentence.
B. From instruments to ministries
Industrial modernity converts dispersion into error to be controlled. Frederick Winslow Taylor writes that efficiency requires standard times and motions; Walter Shewhart teaches that processes need control limits and routine charting so managers can distinguish common fluctuation from assignable cause. Inside a plant, these are defensible techniques. They become dangerous once the plant is treated as a figure for the school, the clinic, or the city.
Why ministries learn to trust numbers is a historical question, not a natural law. Theodore Porter shows that quantification purchases impersonal authority when judgment lacks consensus. Alain Desrosières shows that statistics do not only describe collectivities, they bring them into being through categories and devices. Ian Hacking names the loop by which classifications make up the very kinds they claim to record. James C. Scott maps the administrative appetite for legibility that shaves the richness of local life into units that can be listed, taxed, schooled, and policed. This quartet moves the story from curves to institutions, and explains how one number becomes an alibi for action.
Comparative law and policy confirm the risk. The Hague District Court struck down the Dutch SyRI welfare fraud system for violating privacy and transparency duties. Australia’s Royal Commission found the Robodebt program illegal and injurious because automated averages replaced lawful calculation of income. A policing risk score in Durham raised fairness concerns and showed how model features encode geography and class. The lesson that travels is simple. Once a dashboard displaces reason giving, harm concentrates at the tails that the score cannot see.
C. Normalcy, misfit, and design
The harm of smoothness cannot be seen without the history of normalcy. Lennard Davis shows how the norm supplanted the ideal and installed the figure of the normate as a phantom center. Rosemarie Garland Thomson reframes disability as a relation between bodies and worlds, with misfit naming the moment an environment fails a person rather than the person failing a type. Aimi Hamraie shows that accessible design is not accommodation after the fact but a method that begins from variance and treats difference as a design input rather than a correction. These frames shift Section II from math history to ethics of rooms, since a school or clinic built for a center will make the tails disappear.
A counterexample clarifies scope. The Apgar score compresses neonatal status into five signs at one minute and five minutes, with the explicit aim of triage and the explicit rule that the score is not a future portrait. It works because the window is acute, the physiology direct, and the next actions clear. Single numbers can be good when they govern instruments or moments, not persons across years.
D. Method rules for when a single number may rule
Rule one. Use a single average only when error is symmetric, heterogeneity is modest, and the decision touches instruments or brief episodes rather than life chances. Shewhart’s control limits and the Apgar minute meet this test.
Rule two. When decisions touch persons across time and heterogeneity is large, replace one number with an ensemble that reports at least four things. Repair, which is reduction in exclusion or harm experienced by those at the margins. Variance preservation, which is protection of tails rather than a pleasing mean. Retention at the margins, which is whether guests become stewards. Succession, which is whether the room works in the founder’s absence and whether those once excluded can revise scripts. These are not ornaments; they are internal controls.
Rule three. When you must summarize, use robust estimators that resist contamination without erasing tails. Huber’s M estimators, Tukey’s biweight, and Hampel’s influence function approach were developed for exactly this reason, to reduce the sway of outliers when the outliers are measurement noise. Where the tails are people rather than noise, publish the influence function and refuse any summary that punishes the tails by design.
Rule four. If experimentation guides change, treat the Overall Evaluation Criterion as a constitution, not a mood, and test for heterogeneous effects before you claim a win. Industrial scale A B experimentation literature now treats metric design, guardrails, and false positive control as first class work. The transfer is that platform metrics must never be allowed to reward uplift in the center while they raise error or risk at the tails.
E. Causal schematic without diagrams
Institutions that love one number follow a predictable arc. Leadership selects a single objective that rewards uplift in the center. Designers estimate average treatment effect and ship a uniform interface. Tail users encounter higher error, friction, or harm. Allocation shifts toward the center because the dashboard reports success. Complaints at the tails are discounted as noise since the score looks good. Retraining on data that now contain fewer tails tightens the cycle. Hacking’s loop appears in metrics rather than diagnoses, and Scott’s legibility consolidates in the database rather than on the cadastral map. The result is not mathematical, it is architectural. A room has been resized.
F. Comparative governance in brief
United States law once ratified the translation of norms into punishment through Buck v. Bell. European courts now restrict administrative scoring without transparency as in SyRI. Australia’s Robodebt shows how average based automation injures when unlawfully substituted for statutory calculation. United Kingdom policing experiments demonstrate the difficulty of removing social location from models that predict harm. A responsible ministry therefore publishes categories and recipes, invites adversarial audit, and binds any score to redress routes that can trump the number.
G. Tool not portrait
Safe for one number. Telescope calibration, stable manufacturing processes with documented control limits, brief triage windows with direct physiology. Ensemble required. Education placement, sentencing and parole, social protection eligibility, clinical triage for complex patients, content ranking where safety is at stake. The difference is not sentiment. It is the structure of variance and the cost of being wrong.
H. What this section installs for the rest of the essay
We now own a rule set and a literature spine that meet the highest academic standard. Porter explains the administrative appetite for objectivity; Desrosières explains how publics are made by numbers; Hacking explains how kinds loop back into people; Scott explains why states shave life into legible units. Davis, Garland Thomson, and Hamraie give us the concepts that let us see tail harm as a failure of rooms, not a failure of bodies. Robust statistics gives us tools that resist noise without erasing persons. Comparative cases keep the theory honest about law and policy. What follows can therefore treat measurement not as a neutral craft but as a constitutional choice inside every room we study.
III. Theory. Three architectures of outlier love
A theory fit for history has to name how singular lives become rooms for others and it has to specify the practices that make such rooms durable, which is why I treat outlier love as disciplined attention that keeps persons non identical while building infrastructures of welcome, and I anchor the account in three architectures that can be seen across centuries and archives. The first is voice as shelter, where a voice does more than represent a self because it convenes a public and steadies a day through repeated forms of address that train perception without coercion. The second is camp as kinship, where hospitable excess lowers the cost of entry for policed bodies and turns theatrical risk into social permission. The third is infrastructure of care, where repeated acts harden into institutions that outlast charisma and bind admiration to maintenance. The three are analytic distinctions that often braid in practice, but the braid holds only if each strand is described with enough precision that it can be taught, audited, and, where possible, modestly institutionalized.
A. Voice as shelter
Voice shelters when it fuses precise attention to bodily and social reality with forms that can be lived with every day, which is why Audre Lorde’s account of the erotic matters here, since she names a disciplined resource of felt knowledge that exceeds the market’s plasticized sensation and teaches that intensity can become information and practice rather than consumption, a claim that anchors an ethic of attention under conditions that often punish intensity in the name of normalcy. Voice as shelter is not sentiment. It is a pedagogy that repeats. It is a ritual of naming that equips those who listen to carry a day without shrinking themselves to fit a norm, and the claim can be tested in specific archives where a voiced practice changes what rooms require in order to remain kind.
Judith Butler’s work on public assembly clarifies why this pedagogy travels beyond intimate scenes, because she reads bodies in the square as speech that does not depend on a single utterance and shows how vulnerability itself speaks when persons appear together and insist on livable interdependence, a view that treats embodied co presence as a grammar for convening publics and for measuring whether a room honors those who arrive. Stuart Hall adds an indispensable constraint by refusing the fantasy of transmission without translation, since encoding and decoding are never identical operations and any voiced shelter must assume negotiated readings that can reinforce, negotiate, or oppose the dominant code, which means that voice as shelter must be designed with contested reception in mind rather than with a naïve model of transparent communication.
Eve Kosofsky Sedgwick’s distinction between paranoid and reparative reading gives the practice its discipline, because a reparative stance does not deny injury yet commits to assembling nourishment from partial objects, a method that fits outlier love since it trains publics to read for sustenance without erasing harm, and it shows how voice can be shelter without becoming denial. Saidiya Hartman’s practice of critical fabulation finally secures the moral field in which sheltering voice does its work, because her wayward lives research shows how speculative reconstruction can responsibly honor the missing while refusing to turn the archive’s violence into spectacle, and it thereby sets a norm for how narrators who would shelter others must remain answerable to those least legible to the official record.
From these lenses, voice as shelter yields operational tests. First, admission. Who is expected to adjust and who is allowed to remain non identical. Second, repetition. What scripts and rituals make steadiness possible without enforcing sameness. Third, reception. How are negotiated and oppositional readings anticipated and given a place in the room rather than expelled as failure. Fourth, succession. What happens when the founder leaves and only the scripts remain. These tests convert theory into method, and they keep admiration from sliding into a cult of personality that leaves no shelter once the charismatic voice goes quiet.
B. Camp as kinship
Camp becomes kinship when excess becomes hospitality rather than anesthetic, which is why José Esteban Muñoz matters for any serious account, since his queer futurity frames camp as a performative horizon and a longing that points beyond the present toward rooms not yet built, and that horizon gives analytic reason for why jokes, ornament, and flamboyance can be instruments of survival rather than distractions from it. Sedgwick’s reparative stance again supplies the disciplinary check, because she reads classic camp performance as a thick repertoire of reparative practice where pleasure binds communities that have been taught to expect injury, and the point is not escape but the manufacture of breathing space.
Susan Sontag’s famous notes stabilized a language for camp across publics and can be used here as a boundary marker, because the notes turn a sensibility into a debated common noun and thus make it tempting to strip camp of its historical attachment to queer survival and to treat it as apolitical style, a temptation that any ethic of outlier love must resist if hospitality is to remain the aim rather than the saleable look. Butler’s attention to the body as a scene of assembly returns in this register to remind us that camp’s gestures convene publics whose vulnerability is on display and shared, and that the choreography of such gatherings is not trivial since it determines who can enter without being punished as illegible. Hall then keeps us attentive to the struggle over meaning inside popular culture so that camp’s kinship is not mistaken for a stable code but is seen as a contested field where containment and resistance are always in motion, which is why the same aesthetic can heal in one room and humiliate in another.
Operational tests follow. First, permission. Does the performance lower the cost of entry for those who are most policed. Second, tenderness. When laughter arrives, who is the butt and who is spared, and how often. Third, linkage. What infrastructures of support surround the show so that the room remains hospitable when the lights go dark. Fourth, convertibility. Which gestures can be taught and repeated without founder charisma and without aesthetic capture by markets that prefer style without obligation.
C. Infrastructure of care
Infrastructure of care is where admiration either hardens into maintenance or evaporates into consumption. The archive of social work, settlement houses, nursing, and memory institutions teaches that care becomes durable only when roles, budgets, and review processes are specified, and that specification is not a betrayal of love but its test. Lorde’s insistence that disciplined intensity yields knowledge resists a false choice between affect and administration, since the erotic as power grounds the stamina required for unglamorous tasks that keep rooms kind.
Two further lenses help keep the account honest about markets and status. Tressie McMillan Cottom writes about beauty and status as social technology and shows how institutional logics redistribute vulnerability and reward in ways that punish Black women for strength the culture claims to admire, which instructs any design that would convert admiration into shelter to name market power explicitly and to fund the labor that admiration usually exploits. Fred Moten, in the consent not to be a single being project and in his work with Stefano Harney on the undercommons, reframes social life as study in common and refuses the premise that persons must become discrete managerial selves in order to count, which is a direct instruction for governance models that keep rooms open to fugitive sociality rather than closing them under the respectable pace of institutional time.
Infrastructure therefore has its own tests. First, payment. Who gets paid, how regularly, and for which forms of hidden labor. Second, governance. Who has standing to halt a practice when it begins to re center the average and to exclude the people it was built to shelter. Third, memory. What records are kept so that a city can hold itself to account beyond the cycle of spectacle, as in memorial and museum work that binds grief to law and to policy review. Fourth, succession. What remains when the founder departs, and who is authorized to revise the scripts.
D. Mechanics of conversion
The conversion of a life into a room turns on a small set of mechanics that can be taught and inspected. Repetition and ritual. Shelter becomes real when there are repeatable forms that can be performed by many hands and when those forms allow for variation so that non identical persons can remain whole. Contracts and budgets. Admiration converts to care when there are clauses that protect crews, caregivers, and audiences, and when maintenance appears as a funded line rather than a hope. Audience formation and boundary setting. A room remains hospitable when its stewards teach how to read and be read inside it and when admission is not policed by sameness. Succession and revision. The room must have procedures for revision by those who were once admitted as guests and who are now required to become co authors of the rules. Butler’s assembly, Hall’s negotiated codes, Sedgwick’s reparative practice, Lorde’s disciplined attention, Hartman’s critical fabulation, Cottom’s analysis of markets, and Moten’s consent not to be a single being each supply a piece of this conversion kit, and the pieces cohere only when practice and ledger are allowed to carry as much weight as story.
E. Measurement. From one number to ensembles that honor range and repair
The theory requires a measurement discipline that refuses to let a single number rule persons, since one number portraits were designed for instruments and processes that behave homogeneously and are unsafe in human settings where variance carries meaning, which is why robust statistics matters here as method rather than metaphor. Huber’s work on M estimators shows how one can estimate location with controlled sensitivity to outliers, and Hampel and collaborators formalize the influence function and breakdown point so that designers can see exactly how sensitive their measures are to rare but important cases, and Tukey’s biweight and other redescending estimators embody the principle that far tail points should not be allowed to wreck an estimate yet must be inspected rather than discarded. The point for cultural and civic design is not to import formulas into classrooms or clinics but to borrow the ethic. Replace single score dashboards with small ensembles that track repair and variance preservation alongside reach. Publish indicator recipes and influence analyses so communities can audit the harm that any indicator imposes on the tails. Stress test indicators against gaming before deployment and commit to revision when audits show exclusion that the designers did not intend.
A minimal ensemble that fits this theory includes four classes of indicators. Repair. Measures of reduced exclusion incidents and reduced harm reported by those previously at the margins. Variance preservation. Distributional reports that show whether participation retains its original range or whether the tails have been shaved to produce a pleasing mean. Retention and growth at the margins. Longitudinal tracking of whether those once admitted as guests remain and rise as stewards. Succession. Evidence that the room operates when the founder is absent and that rules are being revised by those once excluded. These are not ornaments to a theory about love. They are its maintenance logs.
F. Guardrails and limits
This section anticipates two likely errors. The first is romanticization, in which a theory of outlier love forgets the workers and the money that keep the doors open and thereby invites a cycle where admiration becomes extraction. The guardrail is hidden labor accounting attached to every case and a refusal to cite achievements without ledgers that can be inspected. The second is aestheticism, in which camp is taken as style without obligation and voice as performance without pedagogy. The guardrail is to bind each aesthetic claim to a practice that can be described, repeated, and paid for. A theory that will not be used to measure rooms will flatter the reader and fail the person who needs a room tonight.
IV. Case archive across centuries
A. Voice as shelter
Voice becomes shelter when attention hardens into scripts, roles, and ledgers that others can enact without the originator present. Fred Rogers scaled tenderness by writing it down and attaching those writings to appropriations that renew on a calendar; the hearing record matters not as spectacle but as a procedural text that binds pedagogy to budget so trained stewards can carry the ritual far from Pittsburgh. The rival explanation is charisma alone. The archive defeats it, since the scripts travel where the person does not. Viktor Frankl moved meaning from memoir to clinic by specifying techniques, supervision, and fidelity checks so that purpose became a repeatable intervention rather than a mood. The rival explanation is post hoc hero worship. The clinic notes defeat it when protocols work across sites. Temple Grandin turned perception into steel by drawing curves, setting throughput, and requiring staff training and audits. The reduction of balks and falls records a conversion of noticing into architecture and wages; the rival explanation of trend or management style loses force where design changes correlate with incident declines and where crews are paid to keep standards. Ruth Bader Ginsburg taught courts and legislators to think when they were not yet ready to rule. Dissents and lectures served as manuals for future majorities, which is how a voice protects the room over time without commanding it; the rival explanation of doctrinal drift is weaker where citation chains run through her reasoning. Malala Yousafzai routed visibility through a fund with local partners, published portfolios, and public themes so admiration became disbursed work with governance; the rival explanation of Northern spotlight is answered where grants seat local leadership. Alan Turing’s wartime memoranda and workflows preserved method and role so that a shop could enact genius in the absence of the genius; the rival explanation of lone brilliance fails against rosters, schedules, and reproducible technique. Hildegard of Bingen’s correspondence shows counsel entering governance when letters invite reply and correction; the rival explanation of visionary fiat is defeated by the epistolary circuit that authorizes revision. Francis of Assisi wrote hospitality into offices, schedules, and obedience to concrete works, which keeps mercy from hardening into performance; the rival explanation of exemplary austerity is displaced by rule and ledger. Jonas Salk held discovery public by pairing clinical protocol with licensing and manufacture and by refusing monopoly at the moment of greatest temptation; the rival explanation of a single breakthrough ignores the governance choices that made access durable. Across these scenes voice shelters because it ties attention to forms others can learn, audit, and revise.
The tests that travel are plain. Who is admitted without being asked to shrink. Which rituals steady days without enforcing sameness. Where the scripts live when the first voice grows quiet. Who pays the hidden workers whose steadiness keeps the room kind.
B. Camp as kinship
Camp becomes kinship when excess is tied to obligations that outlast the show. Bette Midler’s sets at the Continental Baths centered a policed public as the point of gathering and tied glamour to crews, wages, and benefit work. The room did not evaporate when the spotlight moved. The rival explanation is chic without care. Payrolls and benefit calendars answer it. Judy Garland’s night at Carnegie Hall endures because a recording fixed a script for hosting vulnerability that others could perform without consuming the singer; the rival explanation is cult of pain. The survival of listening clubs and careful restagings teaches that ritual, not martyrdom, is what travels. Josephine Baker binds lesson and warning. Resistance work shows how celebrity routes into logistics and intelligence. The Rainbow Tribe shows that kinship requires trustees and reserves or utopia breaks against rent and school fees. The rival explanation of personal excess ignores a missing balance sheet. Harvey Milk converted counters and bullhorns into ordinances and coalitions that remained after his death; the rival explanation of local mood dissipates against enacted law and labor alliances. Bayard Rustin’s plans and rosters keep movements honest about the vocation of logistics. A movement scales when organizing is an office with authority and a budget, not a ghost behind a podium; the rival explanation of spontaneous convergence does not explain bathrooms, buses, and timekeeping. Diana walking a minefield reframed distant hazard as intimate harm because the image was yoked to operators with credibility and to a treaty calendar with votes to count. The rival explanation of pure iconography fails where signatures and clearance maps change.
Method for reading camp as kinship is the same. Ask whether permission lowers entry cost for those most policed. Ask whether laughter spares those who carry the heaviest shame. Ask whether money and unions are present so the room remains hospitable after the house lights die. Ask which gestures can be taught without founder charisma and without capture by markets that prefer style without obligation.
C. Infrastructure of care
Infrastructure of care names the moment admiration hardens into maintenance or dissolves into consumption. Florence Nightingale persuaded with pictures and then kept people alive with schedules, supply lists, rounds, and audits; preventable death declines are a procedural achievement that links diagram to ward. Jane Addams translated neighborhood life into maps and papers legible to municipal actors while keeping residents in charge of inquiry and publication, which is why the settlement reads like a neighborhood university of dignity rather than a classification machine; the rival explanation of benign elite observation fails where residents hold agenda and veto. Dorothy Day turned hospitality into a federated system bound by a newspaper that named needs, taught craft, and distributed recipes for work. Houses endured where budgets included caregivers and maintenance rather than treating deprivation as a saintly badge; the rival explanation of spiritual fervor cannot account for payroll. Wangari Maathai tied seedlings to women’s income and democracy. Registers, payments, and geocoded plantings disciplined a movement that otherwise risked elite capture. Publishing the map turned care into a public ledger. Paul Farmer linked tertiary excellence to rural clinics through accompaniment, supply chains, and training compacts with governments. Survival curves and new public programs record that quality can be delivered in poverty when institutions commit to patience and pedagogy rather than demonstration projects that evaporate after the report; the rival explanation of pilot exceptionalism fails where ministries adopt protocols. Bryan Stevenson braided mourning to law through a campus that dispatches work back to counties. Duplicate pillars and soil jars convert commemoration into municipal routine, so the museum does not hoard moral labor as destination. Greta Thunberg designed a weekly ritual with maximal legibility and open scripts. The limits of Northern concentration clarified revision: fund chapters in the Global South and pair ritual with municipal policy calendars so courage meets a docket. LeBron James entered public education by keeping governance public, publishing distributional metrics, and budgeting the slow build. Mixed early academics alongside strong family supports signal redesign of instruction, protection of teachers, and clear alignment of responsibility between district and foundation; the rival explanation of celebrity vanity project does not survive transparent reporting. Eleanor Roosevelt chaired a drafting committee that produced a general text that later bound constitutions, courts, and classrooms, and it endured because regional elaborations kept particular harms in view. The South African Truth and Reconciliation record demonstrates both power and danger. Hearings, amnesties, and reports gave a nation a way to proceed without war while creating a brand that can displace justice claims if it floats free of material repair; publication of criteria, counts, reparations, and prosecutions is the defense against that drift.
Two anchors widen the frame. Bhimrao Ramji Ambedkar designed a constitutional architecture against caste that routed singular experience through committee procedure, text, and education. The transfer rule is that stubborn hierarchy can be confronted by law that is teachable and enforceable. Ela Bhatt built the Self Employed Women’s Association as an infrastructure of credit, health, and legal defense tied to a union. The transfer rule is that cooperative finance and clinic calendars convert admiration for resilience into wages, care, and leverage. Both confirm the thesis. The room is real where roles, routines, budgets, and review are specified, where the ledger favors maintenance over spectacle, and where succession is not a footnote.
D. Mixed cases, counterfactuals, and the ensemble as maintenance log
Rooms fail. They fail in ways that flatter founders, soothe audiences, and starve crews. The archive already contains the warnings. The Rainbow Tribe broke under the pressure of housing, schooling, and travel costs for a large family without trustees or reserve. The counterfactual is concrete. A monthly reserve that covered rent, school fees, food, health care, and two salaried caregivers would have extended the life of the household and distributed decision rights beyond one body. The instruction that travels is simple. Utopian kinship requires governance and cash flow more than admiration. The I Promise School faced serious proficiency shortfalls in mathematics during early years while families reported strong support from wraparound services. The counterfactual is equally specific. Instruction redesign that separates fluency from problem solving, protects team planning time, and pairs math coaches with teachers across six semesters typically moves cohort growth from district median to upper quartile without sacrificing family services. The instruction that travels is to disaggregate metrics, to protect slow builds from panic, and to share responsibility in public. Diana’s minefield walk accelerated momentum toward a treaty only because images were braided to operators and vote counts. The counterfactual is a pure media tour without Halo and without a calendar that would have saturated attention and changed nothing in law. The instruction that travels is to bind symbol to procedure and to a docket that can register wins and losses.
Measurement here is a maintenance log, not a talisman. Define repair as verified reductions in exclusion and harm reported by those at the margins. Define variance preservation as distributions that resist pleasing means and shaved tails. Define retention as whether those once welcomed as guests remain and rise as stewards. Define succession as whether the room operates when the founder is absent and whether those once excluded are authorized to revise scripts. A worked example makes the ethic visible. Consider a county that adopts a duplicate pillar from a lynching memorial. Count local actions after adoption, including an official acknowledgment, a documented curriculum change, and a ceremony that names victims with family consent. Record co authorship between families and officials through co signed agendas and shared speaking time. Track paid roles for those most affected in the year after adoption. Verify a year two budget and calendar that no longer depend on the originating institution to convene. Publish the recipes. Invite audit. Stress test for gaming and close the loopholes. Measurement becomes an ethic when it keeps the door open.
Hidden labor protocol for Section IV
Identify crews, caregivers, editors, educators, handlers, and organizers by role. Publish wages and benefits where sources exist. Mark gaps as research tasks for the final apparatus. Require a succession paragraph for every miniature.
V. Global modules of reception and care
This section widens the archive without ornament and keeps the same discipline of voice, institution, and ledger, drawing transfer rules from four traditions that already convert singular lives into rooms that others can inhabit. Each miniature names a mechanism that can be taught, a role map that can be staffed, and a ledger specimen that can be audited, then ties the scene to the measurement quartet of repair, variance preservation, retention at the margins, and succession so that admiration cannot float free of maintenance.
A. Sufi letters as distributed discernment
The letter books of the Mevlevi world convene dispersed households, magistrates, and disciples into a network where counsel, rebuke, and consolation travel in repeatable forms that name duties and limits, and the letters convert charisma into office because epistolary instruction binds emissaries and stewards to acts that can be inspected by those who hold copies and memories of the same directives across time. A letters collection functions as a portable chancery for hospitality and conflict resolution rather than as a private oracle, and even brief samples in modern descriptions make the administrative character visible, since the correspondence requests remissions, appoints local hosts, and requires return letters that certify enactment of specific tasks. The transfer rule is direct. Singular authority becomes shelter when counsel is rendered as a role specific script rather than as private access to a master, and when a dated reply is expected so that intention becomes record. The roles are legible. A letter writer acts as rule giver and witness. An emissary carries instruction and returns a report. A local steward hosts the guest and certifies completion. The recipient becomes a co author whose reply enters the file. The minimal ledger is a letter that names the guest to be received, assigns a steward by name, specifies food and lodging, sets a day for review, and requires a short reply that confirms enactment, which together yield a trail that can be audited across houses. Measurement follows the quartet. Repair is the verified decline in refusals at doors named in the letters. Variance preservation is the continued admission of travelers without rank alongside dignitaries. Retention at the margins is the appearance of former guests as emissaries and stewards in later rosters. Succession is the continued use and revision of the scripts when the originating voice is absent and when successors countersign changes while keeping the duty of reception intact.
B. Buddhist monastic codes as hospitality by rule
The Vinaya disciplines attention to strangers, illness, and conflict by specifying who greets, who nurses, who decides, and how review occurs, and the detail matters because it turns welcome into a practice that does not depend on mood. The texts name a porter who keeps watch so that no one is ignored, they assign an infirmarian with a common store for medicine, and they bind adjudication to recorded acts with witnesses, all written as obligations attached to offices rather than as virtues attached to temperament. The story of the Buddha instructing caretaking of a sick monk and stating that whoever would tend to him should tend to the sick has long anchored this procedural ethic in a rule that can be lived without charisma and that travels across houses because the roles and stores are specified rather than implied. The transfer rule follows. Hospitality remains kind when rendered as procedure with stores, offices, and reviews that survive the founder and travel across houses with minimal loss. The roles are stable. The porter acts as sentinel and alerts the house. The infirmarian dispenses care and supplies. The storekeeper publishes inventories. The abbot adjudicates with witnesses and publishes findings to the community. The ledger specimen is a house book that records weekly admissions at the gate, medicines dispensed by the infirmarian, and chapter decisions with the names of witnesses, which together allow outsiders to test whether the rules are lived. The quartet anchors measurement. Repair appears as a decline in unattended illness and refused entry in the book. Variance preservation appears where visitors of low status receive the same intake rites as the learned. Retention at the margins appears when former invalids take their turn as infirmarians. Succession appears when procedures continue under a new abbot and when revisions to stores and offices are recorded without erasing the duty to receive and to nurse.
C. West African epic as civic pedagogy of admission
The Mande epic of Sundiata sustains a grammar of reception and succession in the mouths of griots whose custodianship of memory functions as a civic office, since the praise song instructs rulers in the ethics of admitting strangers, protecting the weak, and binding alliances by gift and oath, and the performance embeds procedure in ritual that is portable across towns because the roles are named and the obligations are public rather than private favors or moods. A city learns to welcome outliers when ritual speech installs obligations repeated at accession, adjudication, and festival, with custodians authorized to correct rulers in the name of the story. The roles are civic rather than clerical. The griot serves as archivist and auditor. The host chief stewards admission and safety. The stranger petitions with rights to be heard. The elders sit as a council that certifies alliance and recalls precedent. The ledger specimen is a festival roster that lists those admitted to the circle by name and clan, the gifts exchanged, and the disputes resolved, kept by the griot as witness and recited at the next festival so that forgetting carries reputational cost. Measurement keeps the ritual honest. Repair is a decline in petitions about unsafe passage and an increase in witnessed safe conducts. Variance preservation is the visible presence in rosters of distant traders, disabled persons, and women with property. Retention at the margins is the rise of former petitioners into trusted roles such as envoy. Succession is the continued authority of the griot to correct the ruler in performance and the preservation of rosters across reigns that allow verification by memory and by list.
D. Latin American base communities as co authored governance
The Medellín documents and the practice of base communities route moral attention into small assemblies that study scripture and neighborhood budgets together, and the assembly form matters because it assigns facilitation, note taking, action items, and review dates so that care becomes public work with a file that can be carried from kitchen table to parish office to city hall, while a pedagogy of conscientization teaches circles to read the word and the world together so that literacy and budget literacy advance in concert rather than cancel each other in practice. Love for those at the margins becomes governance when circles keep minutes, rotate facilitation, and maintain routes to municipal dockets, and when responses from public offices are returned in writing and archived alongside the minutes so that responsibilities travel with names and dates rather than with hopes. The roles are rotational. A facilitator convenes and yields the chair on schedule. A scribe preserves minutes and action lists. A delegate brings petitions to parish and city and returns written replies. A treasurer holds the circle fund and receipts. A municipal liaison receives petitions and commits response times. The ledger specimen is a notebook that contains study notes, attendance lists, agreed actions, letters sent and received, money in and out with receipts, and a calendar of review meetings, which together permit audit by the next circle and by outside observers. Measurement holds translation to account. Repair is the verified reduction in exclusion events documented in replies from municipal offices and clinics. Variance preservation is attendance and leadership by those least literate at the start of the year. Retention at the margins is the number of delegates drawn from formerly excluded households who remain active across years. Succession is the continuity of the circle when founders move, visible in minutes and budgets that continue without interruption and that record revisions by new stewards.
Hidden labor protocol for Section V
Identify the crews that make reception durable and publish their wages where sources allow. In the epistolary houses record copyists, couriers, and porters who keep the door watched and the archive legible. In the monasteries record infirmarians, storekeepers, cooks, and cleaners whose steadiness keeps the ritual kind. In festival cities record performers, runners, and women who provision gatherings and who broker safe passage. In base communities record facilitators, scribes, and delegates who carry petitions and who hold receipts. Where ledgers do not yet name these workers, mark the absence as an archival task and write a method note that commits the project to inventory hidden labor before drawing conclusions about institutional love.
Bridge to method, law, and pilots
These modules deliver the same instruction in four languages. Name roles so that duties survive mood. Keep ledgers that a stranger can read. Publish recipes so that adoption does not depend on proximity to a founder. Bind admiration to maintenance through measures that report repair, variance preservation, retention at the margins, and succession. The next sections will fix indicator recipes and influence descriptions, translate these practices into procurement and redress clauses, and stage pilots that report by the quartet so that theory remains a maintenance log rather than a style.
VI. Measurement protocol, from one number to ensembles
A. First principles, or how instruments differ from persons
The single number belongs to instruments and to processes that behave with modest heterogeneity and symmetric error, which is why telescopes, machine lines, and brief triage windows lend themselves to location estimators and to control limits, while classrooms, clinics, welfare offices, and sentencing hearings contain structured variance that carries meaning and therefore require ensembles that report range and repair rather than a pleasing center, and the statistical distinction is not a metaphor because Huber’s M estimators were built to bound the influence of noise points without pretending that far cases do not matter and because the theorems arrive with a page count that permits designers to publish sensitivity rather than hide it inside a dashboard (Huber 73 to 101; Huber and Ronchetti). Hampel’s influence function formalizes how a single observation can sway an estimator and introduces the breakdown point as a visible limit beyond which a summary collapses, which gives ministries and hospitals language for public commitments about the harm their indicators will and will not impose on tails and converts a technical quantity into a promise that can be audited by those who bear the cost of smoothing (Hampel et al.). Redescending choices such as Tukey’s biweight show that designers can refuse both naivete and cynicism by down weighting extremes while still inspecting them as potential signals, which is the right stance when the tails might be human beings rather than dirt on a lens and when inspection must be tied to redress rather than to erasure (Kafadar 105 to 112).
B. A minimal ensemble that travels across rooms
Replace the single score with a small family that any steward can compute and any public can read. Repair is the verified reduction in exclusion incidents and in harm reported by those previously at the margins, expressed as a rate per relevant opportunity and audited through paired administrative logs and first party reports collected with consent and stored with protections. Variance preservation is the retention of the original distributional range after an intervention, tested by side by side density plots, quantile tables, and statistics that punish shaved tails rather than reward a cosmetic mean, accompanied by a public narrative that names why tails belong in the room and how their presence shapes instruction, triage, or care. Retention at the margins is longitudinal evidence that those once admitted as guests remain and rise as stewards, measured by participation over time and by role progression for groups that were underrepresented at baseline, with survival curves, cohort tables, and rosters that keep names and wages in the same frame. Succession is the operation of the room when the founder is absent and the revision of rules by those once excluded, evidenced by scheduled handoffs, recorded revisions with signatories, and budgets that fund maintenance rather than performance. The ensemble functions as a constitution for the room rather than as a slogan, and it must be introduced where it will be used later so that the same four headings recur without drift.
C. Noise outliers and person outliers, a clean separation
When error arises from instruments, environment, or transcription, robust estimators are appropriate and humane because they protect decision quality while publishing their sensitivity to contamination and because they do not pretend to portrait a person, as in Huber’s bounded score function with a tuning constant stated in advance and justified in public (Huber 73 to 101; Huber and Ronchetti). When dispersion arises from human difference and structured adversity, robust estimators must never be used as portraits that justify erasure, which is why the influence function should be computed and published for every indicator that touches persons and why the tails should be inspected by a named role with authority to halt deployment when a summary would punish those whom the institution claims to welcome, and the rule can be said simply, use robustness to protect instruments and use ensembles to honor persons (Hampel et al.).
D. Influence, breakdown, calibration, and publication duties
Every indicator that touches persons must travel with an appendix that states the estimating equation, the tuning constant where relevant, the influence function in closed form where possible, and the finite sample breakdown point, followed by a short narrative that explains practical consequences in ordinary speech, and the same packet must report discrimination and calibration across groups when the indicator is predictive, because a well ranked but miscalibrated score quietly moves resources away from those who need them most (D’Agostino et al. 743 to 753). Distributional visualizations, reliability checks, calibration curves, and simple side tests for shaved tails belong in the same packet, with scripts that any steward can run and that any public can inspect. The duty is to make sensitivity and calibration visible and negotiable, because that is what it means to bind numbers to reasons rather than to charisma or fear.
E. Stress testing for gaming, drift, and shift
Every indicator invites gaming, therefore the protocol includes adversarial rehearsals in which stewards attempt to raise the headline while worsening the quartet, and the design passes only when such efforts fail in practice or are caught by guardrails. Typical attacks include exclusion of high cost users before measurement, relabeling of adverse events to less visible categories, and substitution of proxy tasks that improve the center while harm accumulates in the tails. Guardrails include stratified reporting with preregistered groups, missingness diagnostics that trigger review, population stability indices that monitor distribution shift, and cross checks that require improvement in the center to be paired with nonworsening at the margins. A living catalog of common attacks and countermeasures should be taught in steward training so that maintenance becomes a civic craft rather than a private sport, and a dataset shift register should be posted with dates, suspected causes, mitigations, and re baselining decisions so that drift cannot be used as cover for harm that a score fails to see (Quiñonero Candela et al.).
F. Experiments that refuse average only wins
Where experimentation guides change, treat the overall evaluation criterion as a constitution that contains the quartet, and preregister heterogeneous effects, subgroup definitions, and stopping rules before shipping, since uplift in the center that raises error at the tails is not a win in rooms that promise welcome and since false positives in social settings are not neutral because they create budgets and habits that resist reversal (Kohavi, Tang, and Xu). Use stratified randomization when feasible, require minimal detectable effects at the margins, and bind success declarations to the ensemble rather than to a single objective so that the experiment’s governance matches the room’s promise. Where randomization is not possible, use interrupted time series with clear modeling assumptions and publish influence checks on the statistics that summarize the interruption, and record protocol deviations with signatures rather than with anecdotes so that learning travels. To prevent analytics drift, publish the complete analysis plan and the exploration logs that would otherwise remain private, because the garden of forking paths is not a metaphor but a daily hazard wherever multiple looks and flexible models can produce a pleasing but fragile win (Gelman and Loken).
G. Templates for public reporting, redress, and veto
Numbers can injure when they displace reasons, therefore the protocol requires a public report that presents the quartet in the same order every time, states the influence findings and subgroup calibration, and links to the appendices and to a docket where comments and audits live with dates and names. The report must also state redress routes by which an individual or a community can challenge an indicator and obtain relief that overrides the number, with service level deadlines for response, because a number must not be sovereign over a person. Appeals calendars, procurement clauses, and fund allocations for independent advocates belong in the same packet so that moral promises are inscribed in contract and in routine.
H. Maintenance logs, cadence, and budgets
A maintenance log is a simple table that lists dates, stewards, indicator versions, audits performed, incidents found, repairs made, and budget effects, and the table is posted alongside the public report so that a stranger can see how care becomes routine. The cadence must be specified in calendar time with named reviewers, and the budget line that pays for audits and for redress must be visible so that maintenance is not left to volunteer labor. Scripts and schemas belong in a public repository with version numbers and change notes so that reproducibility is a shared craft rather than a private claim.
I. Simplicity bounds and principled limits
Ensembles must remain teachable, therefore the protocol sets a simplicity bound of four to seven indicators in routine use, with additional diagnostics available in appendices, and it requires that every steward be able to state the quartet and the influence idea in ordinary speech. Where an indicator’s governance cost exceeds its benefit, the rule is removal rather than inertia. Some settings still warrant a single number for a brief window with direct physiology, as with the one minute and the five minute Apgar, which explicitly refuses the fantasy of a future portrait and binds the number to immediate action and to a later recheck, and this limited permission confirms rather than weakens the general rule that rooms with human heterogeneity require ensembles that honor range and repair (Apgar 250 to 259). As a different kind of limit, fairness desiderata can be mutually incompatible under unequal base rates, which means that designers must state which compatibility they are choosing and why, and must align that choice with the quartet so that the unavoidable trade off is governed by reasons rather than by convenience (Kleinberg, Mullainathan, and Raghavan; Chouldechova).
J. Data protection, consent, and rights of the measured
An ensemble that touches persons must be paired with a data governance kit that states collection purpose, retention duration, access controls, and participant rights in ordinary speech, and must minimize collection to what the indicator requires, since privacy is not an afterthought but a precondition for trust. When differential privacy or related guarantees are used, the parameters must be published with a narrative that explains how added noise affects auditability and individual rights and must be aligned with the quartet so that privacy budgets do not silently degrade repair at the margins, because protection of the one must not erase the many who live at the edges of the distribution (Dwork et al.). Consent must be consent to a specific use with clear routes for revocation, and redress must include correction rights, annotation rights, and the right to be evaluated by a human steward with authority to overrule an indicator.
K. Heterogeneity analysis, fairness diagnostics, and incompatibility made public
Heterogeneous effects must be the norm rather than an afterthought. At minimum, every intervention and every indicator must report stratified outcomes for preregistered groups, interaction tests for treatment heterogeneity, subgroup calibration for risk scores, and sensitivity of influence to subgroup membership, and the results must be read aloud to the communities whose lives are being summarized. Where fairness criteria such as equalized odds, predictive parity, and demographic balance cannot all be satisfied, the public should be told which criterion is binding in this room and why, and how the chosen criterion interacts with the quartet so that tails are not shaved to purchase an elegant statistic (Hardt, Price, and Srebro; Kleinberg, Mullainathan, and Raghavan; Chouldechova).
L. Worked micro examples that teach use without romance
Consider a clinic that adds a triage score for complex patients. Repair is defined as verified reductions in missed follow up for those with multimorbidity and housing instability. Variance preservation is retention of encounter length distribution for the slowest decile rather than compression toward the mean. Retention at the margins is the rate at which patients from the slowest decile return and are seen by their chosen clinicians. Succession is the continued delivery of these gains when the initiative lead rotates and is measured by a stable budget line and by revised scripts for intake that were authored by front line staff. The packet includes the estimating equation for the triage score, the influence function, subgroup calibration curves, and an appeals route by which any patient can request human review with authority to overrule the score. A school district that adopts a placement index would follow the same quartet, would publish the index’s influence and calibration by language status and disability category, and would bar the use of the index as a portrait of a child’s future potential, because a classroom is not a telescope and a student is not an instrument. A city that adds a risk indicator to a social protection program would publish how missing wages and irregular housing records alter influence, would show that denial rates do not concentrate among those whose lives are least legible to databases, and would guarantee that appeals can overrule the indicator with reasons that appear in the maintenance log.
M. Roles, training, and the vocation of stewardship
Numbers do not bind themselves to reasons, which is why this protocol names roles and trains them. Designers write estimators and publish influence. Stewards operate rooms, read the quartet aloud, and carry the appeals calendar. Auditors test for shaved tails, run adversarial rehearsals, and publish countermeasures. Archivists keep the maintenance log with names, dates, and budgets. Founders submit to succession and do not own the indicator once a room has adopted it. Each role carries a short examination of conscience tied to the quartet, and no indicator ships without named humans for each duty.
N. Failure review and near miss practice
Rooms fail in ways that soothe audiences and starve crews. The protocol therefore requires a near miss register that records incidents in which harm almost concentrated at the tails and records the procedural change that prevents recurrence, and it requires public incident reviews where those most affected speak first and last, and it requires that budgets move when failure repeats so that a maintenance promise becomes a ledger rather than an apology. The near miss register stands beside the maintenance log so that a city can learn across rooms.
Bridge to law and to pilots
Law must purchase metrics that publish influence and calibration and that accept redress, and field laboratories must report by the quartet rather than by charisma or fear, which is why the next section translates these duties into procurement and appeal clauses and why the pilots will run with preregistered analysis plans, heterogeneity checks, and maintenance logs that anyone can read. The measure is love only when it keeps the door open and when the door still stands after the founder leaves the room.
VII. Administrative law and design templates
This section translates method into enforceable text that binds numbers to reasons, seats appeals inside calendars, purchases auditability at the moment of procurement, and requires public recipes that a stranger can read. The forms below are statutory where possible and contractual where delegated, and they are written so that a ministry, a city, a district, or a hospital can adopt them without mood and can be audited without proximity to the founder.
A. Scope and definitions
A room that touches life chances must name its instruments. The following terms govern measurement, reporting, and redress.
Life chance decision. Any decision with material effect on liberty, livelihood, education placement, housing, health care access, or social protection eligibility.
Indicator. A quantified measure that purports to summarize performance, risk, or status, whether or not it is predictive.
Ensemble. The minimal family of indicators that a room must report together, namely repair, variance preservation, retention at the margins, and succession, each with a published recipe and with influence and calibration appendices attached.
Influence write up. A short appendix that states the estimating equation, the tuning constant where relevant, the influence function, the finite sample breakdown point, and subgroup calibration findings, with a narrative in ordinary speech.
Maintenance log. A public table that lists dates, stewards, indicator versions, audits performed, incidents found, repairs made, and budget effects.
Redress ladder. A sequence of appeal routes that can override any indicator with reasons that are published and that update the maintenance log.
B. Prohibition and permission
The single number is prohibited for life chance decisions. This is not sentiment but an application of the difference between instruments with modest heterogeneity and persons whose variance carries meaning. A narrow permission remains for a brief window with direct physiology where the number governs an instrument and is paired with a recheck that refuses a future portrait. The one minute and five minute Apgar meet this test and thereby confirm the rule rather than weaken it (Apgar 250 to 259).
C. Public registry and disclosure duties
Indicators that touch life chances must be visible before they govern a person. A public registry shall list every such instrument and shall display for each entry the recipe, the ensemble, the influence write up, the calibration report, the maintenance log, the adversarial tests, the dataset shift register, and the contact for redress. Change logs with version numbers and effective dates shall be posted with equal prominence. No indicator may be used unless its registry entry is current and complete.
D. Model ordinance, core clause
The following clause is written for municipal code, agency rules, or board policy and shall be incorporated by contract wherever services are procured.
> Section 1. Ensemble requirement. Any agency or contractor that makes or materially influences a life chance decision shall publish and maintain an ensemble that reports repair, variance preservation, retention at the margins, and succession. Each element shall include a recipe with data fields, transformations, timing, and review cadence. No single indicator may be used to enact or justify a life chance decision.
> Section 2. Influence and calibration publication. Any indicator that touches persons shall be accompanied by an influence write up and a calibration report for preregistered groups. These materials shall be posted with the same prominence as the headline number and shall be archived with version control.
> Section 3. Heterogeneity and subgroup duty. Outcomes shall be reported for preregistered groups. Any claimed improvement in the center shall be paired with nonworsening at the margins. Claims of uniform benefit require evidence of small heterogeneous effects or a reasoned explanation of trade offs, with fairness incompatibilities disclosed in ordinary speech and with a review date set in the registry.
> Section 4. Redress ladder and veto. Any person or community affected shall have access to an appeal that can override an indicator. Appeal routes shall include service level deadlines, authority to order relief, and a duty to publish reasons in the maintenance log. No indicator shall be sovereign over a person.
> Section 5. Audit, adversarial test, and budget. Each agency shall conduct adversarial rehearsals that attempt to improve headline numbers while worsening the ensemble. Results and countermeasures shall be published. A dedicated budget line shall fund audits, redress, translation, and community participation stipends.
> Section 6. Notice, comment, and impact assessment. Prior to deployment or revision of any indicator, agencies shall publish an algorithmic impact assessment that includes subgroup calibration, heterogeneity plans, and quartet targets, shall provide notice and a public comment window, and shall post a plain language rationale with a response to comments.
> Section 7. Private right of action and standing. Any affected person or designated community organization may bring a civil action to enforce this ordinance. Prevailing parties may recover reasonable costs and fees.
> Section 8. Purpose limitation and nonre use. Indicators may be used only for declared purposes and scopes. Any new use requires new notice and comment, a revised influence and calibration appendix, and a recorded vote.
> Section 9. Sunset and renewal. Each indicator expires twenty four months after deployment unless renewed after audit, ensemble reporting, and public hearing. Exemptions expire on the same cadence.
> Section 10. Immediate suspension. Any indicator lacking a current registry entry, influence write up, or calibration report is suspended until compliance is restored.
> Section 11. Vendor escrow and auditability. Contractors shall deposit reproducibility materials with a neutral escrow agent, including model artifacts or equivalent executable documentation, training data documentation, and change logs, to be released upon default or refusal to cooperate with lawful audit and recertification.
> Section 12. Language access and accessibility. All registry entries, notices, and redress forms shall be available in plain language, in the languages of the jurisdiction, and in accessible formats that meet published standards.
E. Administrative procedure and reasons
Decisions that touch life chances require reasons that can be reviewed. The ordinary law of fair process instructs the shape of these instruments. When agencies adopt or revise indicators, they must provide notice, accept comment, and publish a response, not as theater but as the procedural form by which a public learns and contests. Where benefits or entitlements are at stake, the balance among private interest, risk of erroneous deprivation, and value of additional safeguards supports the ensemble and the redress ladder because together they reduce error at the tails while preserving administrative capacity for routine cases (Goldberg v. Kelly 397 to 254; Mathews v. Eldridge 424 to 319). Procurement must carry the same logic so that reasons are inscribed in contract and in calendar rather than in aspiration alone.
F. Transparency, records, and access
The maintenance log, the influence write up, the calibration report, the adversarial test results, and the dataset shift register are records that must be available to the public subject to lawful protections for identifiable data. Freedom of information regimes require timely responses and completeness. Where personal data protections apply, publication is accomplished by aggregation, by privacy preserving release, and by narrative that preserves auditability without exposing persons. Rights regarding automated decision making further support the publication of recipes, influence, calibration, and redress routes, since the public cannot evaluate what it cannot see and since a number that affects a person must be answerable to that person in a form they can read (European Union Regulation 2016 to 679; United States Freedom of Information Act).
G. Fairness incompatibility disclosures
Fairness desiderata can be mutually incompatible under unequal base rates. A room that adopts an indicator must state which fairness constraint it will honor, why that choice fits the setting, and how the choice interacts with the ensemble so that tails are not shaved to purchase an elegant statistic, and this disclosure must be in ordinary speech, must name expected harms, and must be open to revision if audits show damage that the constraint did not anticipate (Kleinberg, Mullainathan, and Raghavan; Chouldechova).
H. Independent audit and adversarial rehearsal
An audit is not a courtesy. It is a condition of permission to use an indicator that touches persons. Auditors must be independent of the line of supervision that owns the instrument, must have access to data, code, and change logs, must run adversarial tests that attempt to improve the center while worsening the margins, and must publish countermeasures when such attacks succeed. They must record and disclose near misses, since near misses are how a city learns before harm repeats. Audit calendars and budgets must be public, and procurement must tie payment to the presence of current audits.
I. Redress ladder, forms, and deadlines
Redress without clocks is theater. The ladder begins with an internal appeal that must receive a reply by a date and that must include reasons that cite the ensemble and any human review that overrode a number. The next rung is an independent reviewer with authority to order relief and to require changes to an indicator. The final rung is judicial or quasi judicial review. Forms must be short and available in the languages of the city. Representation must be funded for those without means when the decision touches housing, subsistence, or liberty. Any adverse action letter must name the ensemble results that governed the decision and must link to the influence and calibration appendix in use on that date.
J. Data protection and minimization
An ensemble that touches persons must be paired with data minimization and purpose limitation. Collection shall be limited to what the indicator requires. Retention schedules and access controls shall be published in ordinary speech. Where privacy budgets or related guarantees are used, parameters shall be published with narratives that explain how added noise affects auditability and individual rights and that align with the quartet so that protection of the one does not erase the many who live at the edges of the distribution (Dwork et al.; Dwork and Roth). Consent must be consent to specific use with routes for revocation. Correction, annotation, and human review are rights, not favors.
K. Carve outs, exemptions, and narrow windows
Carve outs are how a discipline loses shape. The only safe exemption for the single number is the brief instrument window with direct physiology that is paired with a recheck and that explicitly refuses a future portrait. Any other claim to exemption must state the specific harm to be avoided, the reasons an ensemble would be disproportionate, and the date at which the exemption will be reviewed. Exemptions expire unless renewed after a docketed showing of necessity.
L. Sanctions, remedies, and fee shifting
Noncompliance has a way of masquerading as delay. Sanctions must include suspension of the indicator, limits on use, and monetary penalties that fund redress and audits. Where persons prevail on appeal or in court, fee shifting must allow recovery of costs, since access to redress without resources is a promise without teeth. Where contractors fail, remedies include cure periods with posted plans, followed by termination and debarment for repeated refusal to publish or to repair.
M. Training and roles
Numbers do not bind themselves to reasons, so training must be role specific. Designers must compute and narrate influence and calibration. Stewards must read the ensemble aloud, run the maintenance log, and carry the appeals calendar. Auditors must test for shaved tails, run adversarial rehearsals, and publish countermeasures. Archivists must preserve ledgers and change logs. Founders must submit to succession and to the rule that an indicator belongs to the room once adopted. Each role carries a short examination of conscience tied to the quartet, and no indicator ships without named humans for each duty.
N. Worked templates
The following short forms can be adopted immediately and placed into policy manuals and contracts.
> Maintenance log. Date. Indicator version. Steward on duty. Audits run. Incidents found. Repairs made. Budget effect. Next review date. Signatures.
> Influence write up. Indicator name. Estimating equation. Tuning constant. Influence function. Breakdown point. Subgroup calibration. Narrative in ordinary speech. Version date. Repository link.
> Redress form. Decision challenged. Date and place. Statement in the person’s own words. Relief requested. Preferred language. Consent for data use in appeal. Receipt date. Reply due date. Assigned reviewer. Outcome and reasons. Rights for further appeal.
O. Comparative notes for adoption
Comparative jurisprudence confirms the utility of these forms. Courts have rejected opaque scoring for benefits administration on privacy and transparency grounds and have asked for procedures that do not convert averages into sentences, as in the Dutch SyRI judgment. Commissions have found automated averaging unlawful where it displaced statutory calculation and have ordered repayment and reform, as in the Robodebt report. Cities have begun to regulate automated decision tools with audit, notice, and explanation requirements, as in New York City Local Law one hundred forty four. These moves align with the ensemble duty and convert a literature spine into governance that a stranger can test in daylight (Netherlands Rechtbank Den Haag; Royal Commission into the Robodebt Scheme; New York City Local Law 144 of 2021).
P. Bridge to field laboratories
Law purchases the right to measure by requiring ensembles, influence, calibration, and redress. Field laboratories now carry these duties into schools, clinics, and cities, where the quartet will be reported by calendar and where maintenance logs and near miss registers will sit beside budgets. The next section installs pilots with preregistered analysis plans, heterogeneous effects, registry entries, and public scripts, so that rooms can learn in public and so that admiration never floats free of maintenance again.
VIII. Field laboratories and implementation
This section turns promises into rooms. Three pilots—school, clinic, city—operate as field laboratories where the quartet is the constitution, roles are named, ledgers are public, and learning is paced by calendars rather than moods. Each pilot maintains a public registry entry (per VII), a preregistered plan, a maintenance log, a near-miss register, and an appeals ladder that can overrule any indicator. Each names hidden labor and funds it.
A. Design charter (applies to all pilots)
1. Constitutional objective. Every pilot reports the quartet in the same order and phrasing—repair, variance preservation, retention at the margins, succession—and binds success declarations to this ensemble rather than to a single headline.
2. Precommitment. Before launch, each pilot posts a preregistration packet: aims; rollout schedule; analysis plan; heterogeneous-effects plan; stopping rules; registry URL; indicator recipes; influence and calibration appendices; audit cadence; redress deadlines; budget lines for audits, redress, translation, and community stipends.
3. Clocks and triggers. Quarterly audits; service-level deadlines for appeals (reply within 10 business days; resolution within 30); automatic review when any of: (i) subgroup error > 20% relative to baseline or to the best-served group, (ii) data missingness > 5% overall or doubles baseline in any subgroup, (iii) dataset-shift indices breach preregistered bounds. Two consecutive breaches trigger rollback to the last approved script.
4. Budget floor. ≥2% of program operating funds ring-fenced for audits, redress, translation, and community participation; posted in the registry and reviewed at renewal.
5. Language access & accessibility. All notices, forms, and reports appear in plain language, in the jurisdiction’s major languages, and in accessible formats that meet published standards.
B. Data specification and operationalization (all pilots)
1. Minimal, purposeful collection. Collect only variables required by the recipe; publish a variable dictionary (field name; type; allowed values; provenance; retention period; access roles).
2. Repair. Operationalized as verified reductions in exclusion or harm incidents at the tails, expressed as rates per relevant exposure (e.g., wrongful denials per 100 applications; missed follow-ups per 100 scheduled encounters), with dual sources (administrative logs + first-party reports) and reconciliation rules.
3. Variance preservation. Distributional reports (empirical CDFs, decile tables, tail counts, quantile gaps) pre/post intervention; “tail integrity” test that flags shaving (decline in participation or service for bottom/top decile relative to center).
4. Retention at the margins. Longitudinal participation and progression for groups underrepresented at baseline; survival curves (time-to-attrition, time-to-progression); role-roster advancement for admitted “guests” into steward roles.
5. Succession. Evidence that scripts, budgets, and results persist through leadership turnover: dated handoffs, revised scripts authored by front-line stewards, and ensemble stability during founder-absent periods.
C. Pilot 1 — School: placement without shaving the tails
Scope. Replace a single placement index with an ensemble governing access to advanced coursework and portable re-entry lanes.
Mechanism. Quartet indicators: repair (reduction in upheld appeals; late entries granted with documented reasons), variance preservation (participation and assessment distributions with intact tails), retention at the margins (persistence and progression of first-time admittees from historically underrepresented groups), succession (stable operation during leadership turnover; scripts revised by teachers/counselors).
Rollout. Stepped-wedge cluster randomized trial across schools over three semesters; all clusters eventually treated, staggered starts provide identification (Hussey & Hughes).
Practice. A one-page “second-look” lane keyed to teacher nomination + portfolio rubric with calendarized windows; decoupled instruction metrics (fluency vs. problem solving); weekly protected team planning time; appeals form (major languages; accessible formats); deadlines: reply 10 days, resolution 30.
Analysis. Intent-to-treat at cluster level with cluster-robust SEs; prespecified heterogeneous effects (language status, disability category, race/ethnicity, prior opportunity to learn); quantile treatment effects (τ ∈ {0.1, 0.25, 0.5, 0.75, 0.9}) to detect center-only gains; retention with Kaplan–Meier and Cox models; FDR control for multiple subgroup contrasts (Benjamini & Hochberg).
Assumptions & sensitivity. Parallel-trends checks for stepped-wedge pre-period segments; Rosenbaum bounds for unobserved confounding in non-random uptake; permutation inference at the cluster level.
Artifacts. Registry entry; recipes; influence & subgroup calibration appendix for any predictive component; monthly maintenance log; dataset-shift register (attendance, mobility, test participation) with mitigations. Triggers → automatic review; two consecutive breaches → rollback.
D. Pilot 2 — Clinic: complex care triage without erasing persons
Scope. A triage bundle for complex patients routes time, continuity, and social work by need without letting a single risk score become a portrait.
Mechanism. Repair (fewer missed follow-ups among multimorbid patients with housing/food insecurity), variance preservation (encounter-length distribution for slowest decile remains intact), retention at the margins (return rates and clinician continuity for slowest-decile patients), succession (gains persist across attending rotations; scripts revised by nurses, social workers, front desk staff).
Rollout. Team-level randomization; if infeasible, segmented interrupted time series with prespecified modeling, falsification outcomes (e.g., conditions unaffected by triage), and pre-registered discontinuity checks.
Practice. Human-in-the-loop triage (named steward may overrule with recorded reasons); warm-handoff scripts and slots; reserved slow-visit capacity; embedded social-work hours; after-visit summaries naming next-action responsibility; appeals printed on discharge paperwork with staffed phone in preferred language.
Analysis. Difference-in-differences on missed follow-up with clinic and month fixed effects; subgroup calibration for any prediction (by housing status and language); adverse-event monitoring with monthly case review; qualitative process tracing (standard implementation framework) to surface hidden labor.
Assumptions & sensitivity. Pre-trend balance and placebo outcomes; synthetic-control sensitivity where appropriate; tipping-point analysis for misclassification rates in social-risk fields.
Artifacts. Equations and influence for any numeric triage element; calibration plots by subgroup; registry entry; maintenance and near-miss logs. Auto-review on subgroup no-show spikes or shift indices; rollback if two consecutive months breach thresholds.
E. Pilot 3 — City: social-protection eligibility without dashboard sovereignty
Scope. Replace an opaque intake score with an eligibility and prioritization ensemble that ties numbers to reasons and reasons to appeal.
Mechanism. Repair (decline in wrongful denials reversed on appeal), variance preservation (steady/widened geographic and demographic reach at tails), retention at the margins (repeat access and graduation rates for households with prior denials), succession (performance through election/budget turnover; script revisions logged).
Rollout. Phased launch via randomized encouragement: addresses mapped to centers with trained navigators using the ensemble + appeal ladder; parallel centers run legacy intake during phase one.
Practice. Published recipes for each eligibility element; navigator role funded with escalation authority; plain-language appeals, accessible formats, deadlines (reply 10 days; resolution 30); provider payments contingent on current registry status, maintenance, and public reporting.
Analysis. Randomization inference on wrongful-denial rate and time-to-benefit; subgroup reporting (language, disability, household composition, shelter history); spatial clustering with robust variance; sensitivity to missing wage data/irregular addresses; explicit fairness-incompatibility disclosure aligned to the quartet.
Assumptions & sensitivity. Balance checks on encouragement assignment; instrumental-variable robustness (first-stage strength, exclusion plausibility) when using encouragement as an instrument; Manski-type bounds under partial identification if records are incomplete.
Artifacts. Registry entry with recipes, ensemble, influence, calibration, logs, near-misses, and redress contacts; quarterly adversarial rehearsals (“raise throughput by diverting hard cases”) with countermeasures posted; dataset-shift register (economic shocks, policy changes).
F. Randomization units, power, and sample size
1. Units and ICC. Clusters are natural work units (schools, clinic teams, service centers). Report intraclass correlation coefficients (ICCs) for primary outcomes; inflate sample size by design effect = 1 + (m−1)·ICC.
2. Minimal detectable effects (MDEs). Compute MDEs for the tails (e.g., 10th/90th quantiles) and for subgroup contrasts, not just overall means; publish the table in preregistration.
3. Duration. Choose stepped-wedge periods to secure ≥80% power for tail outcomes given ICC and anticipated attrition (Hussey & Hughes; Hemming et al.).
4. Interim looks. At preregistered checkpoints, expansion requires center improvement and nonworsening in tail groups; otherwise freeze, diagnose, and revise.
G. Analysis plan (common core)
1. Estimands. Define Average Treatment Effect (ATE), Quantile Treatment Effects (QTEs), and subgroup ATEs; specify estimators and variance estimators before data lock (Imbens & Rubin).
2. Identification. List required assumptions for each design (randomization integrity; parallel trends; exclusion for instruments) and how each is probed.
3. Robustness. Cluster-robust variance; permutation tests; placebo tests; influence analyses for summary statistics; FDR control where many contrasts are read (Benjamini & Hochberg; Angrist & Pischke).
4. Calibration. For any predictive component, publish reliability curves, Brier scores, and subgroup calibration plots; declare thresholds for acceptable miscalibration and remedies.
H. Sensitivity and partial identification
Where untestable assumptions loom (unobserved confounding; nonrandom missingness), compute Rosenbaum sensitivity bounds; publish bias-adjusted ranges rather than point triumphs (Rosenbaum). Where data are structurally missing, present Manski-style worst-case/best-case bounds and then tighten with credible monotonicity or instrumental assumptions, each stated in ordinary speech.
I. Registry, transparency, and reproducibility
Each pilot’s registry entry hosts: recipes; ensemble; influence write-ups; calibration reports; analysis plan; codebook; audit schedule; logs; near-miss register; dataset-shift register; change-log with versions and dates. Reproducibility materials (code, synthetic data or privacy-preserving outputs) accompany public reports. Payment and continuation are contingent on current registry status.
J. Governance, ethics, and rights
Human oversight is a named role with veto authority over any indicator. Data use is minimized to purpose; retention and access controls are posted with narratives explaining consequences for auditability and rights. Adverse events are defined before launch; near misses are discussed with those most affected speaking first and last. Correction, annotation, and human review are rights, not favors (cf. VII on data protection and redress).
K. Hidden labor protocol and wage transparency
Publish a role map (designers, stewards, auditors, archivists, navigators/coaches/attendants), a wage/benefit table, and calendars of paid time for maintenance, appeals, audits, and community review. No success report appears without the wage table that made it possible. Budget moves when failure repeats.
L. Adversarial rehearsal and gaming catalog
Once per quarter, run a “red-team week” to attempt center gains while worsening the quartet: exclusion of high-cost users pre-measurement; relabeling of adverse events; proxy tasks that lift the mean while shaving tails. Publish the attempts and countermeasures; add them to a living catalog taught to stewards.
M. Reporting cadence and artifacts
Monthly: maintenance log, near-miss entries, dataset-shift updates.
Quarterly: ensemble report in fixed order with recipes, influence, calibration; audit notes; responses to public comment.
Annually: renewal review against quartet targets; succession notes documenting rule revisions by new stewards; budget reconciliation for the audit/redress floor.
N. Success gates and rollback
A pilot is a win only if all five hold: (1) repair improves with reasons; (2) variance preservation shows tails intact or widened; (3) retention at the margins rises for those first admitted as guests; (4) succession is demonstrated by operation without founders and by recorded revisions authored by front-line stewards; and (5) no subgroup shows calibrated harm masked by an improved center. If any gate fails, trigger rollback, publish diagnosis, revise scripts, and re-run before expansion.
O. Transfer package
Each successful pilot publishes a transfer bundle: roles and scripts; recipes; influence and calibration appendices; maintenance and near-miss templates; adversarial catalog; budget template; registry checklists. Adoption elsewhere requires verbatim adoption of clocks, triggers, redress, and the budget floor.
P. Bridge to counterarguments and limits
Field laboratories will surface claims about complexity costs, speed, and capture. The next section stages those objections with numbers and governance, sets simplicity bounds, and names where single numbers remain lawful—brief instrument windows with direct physiology and rechecks—so discipline does not dissolve into style.
X. Ethics and theology of institutional love
This section returns to first principles. If love appears in history as disciplined attention that changes how rooms are arranged and how ledgers are written, then an ethic must name roles, specify vows that bind admiration to maintenance, and publish examinations of conscience that a stranger can read. Theology matters here not as ornament but as an archive that already translated affect into office—porter, infirmarian, cellarer, prior—and tied hospitality to stores, routines, and review. Critical theory matters because it refuses pious aestheticism and insists that care survive markets, status, and spectacle. The aim is not mood. It is a rule of life for rooms that touch life chances.
A. Love as procedure, not sentiment
The claim is spare. Love becomes public when attention is rendered as repeatable tasks that others can learn, audit, and revise. Audre Lorde’s “erotic as power” grounds stamina and clarity without collapsing intensity into consumption; disciplined feeling becomes information and practice rather than spectacle (Lorde 53–59). Judith Butler’s account of assembly insists that bodies convened are already speaking and already naming interdependence that institutions must honor with roles and rooms (Butler). Eve Kosofsky Sedgwick’s reparative stance gives the discipline that keeps harm visible while assembling nourishment from partial objects (Sedgwick 123–51). Saidiya Hartman’s critical fabulation sets the moral field in which narration must answer to the least legible without turning violence into a show (Hartman). Tressie McMillan Cottom names markets as status engines that punish the very strength they celebrate; any ethic of institutional love that ignores markets will convert admiration into extraction (Cottom; “I Was Pregnant…”). Fred Moten’s consent not to be a single being refuses the managerial fantasy of discrete, ownable selves and makes study in common a civic practice (Harney and Moten). Aimi Hamraie’s universal design reframes accommodation as method rather than afterthought (Hamraie). These lenses yield rules: bind affect to office; publish reasons and ledgers; protect tails as design input; keep study in common alive in rooms built for audit.
B. Hospitality made durable: monastic templates without romance
The Rule of Benedict names a porter who keeps watch so the stranger is never ignored, a cellarer who manages stores without harshness, an infirmarian accountable for the sick, and a schedule that makes room for what interrupts (Benedict, chs. 31, 36, 53). Franciscan sources convert charisma into itinerant scripts that let poverty teach without domination and tie penance to civic attention (Francis of Assisi). The transfer is procedural. We adopt “offices of love” that any clinic, school, or city can staff.
– Porter → Intake steward: receives each person as a guest, explains rights and redress, logs arrivals in a book that names time, language, and need, and triggers the appeal clock.
– Infirmarian → Care steward: maintains a common store (time, slots, navigators), logs dispensations, and reports shortages as obligations, not excuses.
– Cellarer → Budget steward: publishes stocks, flows, and reserved funds for audits and redress; refuses to purchase shine with maintenance money.
– Prior → Succession steward: rehearses founder-absent operations, documents rule revisions authored by those once admitted as guests.
The tests are operational: whether the intake book shows no ignored arrivals; whether the medicine book (time, slots, navigation) matches promises; whether audit and redress lines are funded on calendar; whether the house works when the founder leaves (Benedict; Francis of Assisi).
C. Secular obligations that keep rooms kind under pressure
Markets, status, and administrative time distort care. Cottom’s analysis demands wage tables and procurement clauses that keep admiration from raiding labor (Cottom). Hamraie and Lennard Davis require design that preserves variance and refuses normate defaults (Hamraie; Davis). Butler and Hall warn that publics read against the grain; scripts must anticipate contested reception and give negotiated readings a place in the room rather than expelling them as failure (Butler; Hall). Sedgwick and Hartman set a floor: do not confuse reparative practice with denial; do not turn injury into content.
D. Role specific vows (portable across rooms)
Vows bind persons to offices rather than to founders. Each vow is followed by an examination of conscience tied to ledgers.
1. Porter / Intake steward (admission and first redress)
Vow: Receive each person as a guest without forcing sameness; explain rights and the appeal ladder in ordinary speech; start clocks.
Examination: Is the intake log complete (names, languages, time)? Are appeal forms available in the major languages and accessible formats? Were reply (10 days) and resolution (30) met? Did I refuse anyone by smooth habit or missing script? (Benedict, ch. 53.)
2. Care steward (time, slots, navigation)
Vow: Guard time for the slowest decile; embed warm handoffs; publish rationing in daylight.
Examination: Does the store show reserved slow visits/second looks? Did handoff scripts run? Did rationing move toward the center by habit? Are repairs logged with dates and budget effects?
3. Budget steward (money and maintenance)
Vow: Keep audits, redress, and translation funded on calendar; publish wage tables for hidden labor; never trade maintenance for spectacle.
Examination: Is ≥2% ring-fenced and spent? Are stewards paid for logs and appeals? Did any report praise the room without the wages that made it possible? (Cottom.)
4. Audit steward (adversarial rehearsal)
Vow: Practice corruption so the room does not learn it by accident; publish countermeasures.
Examination: Did we try to raise the center while worsening the margins? Did triggers fire? Did we roll back when they fired twice? Are near misses recorded with fixes? (Goodhart; Campbell in IX.)
5. Archivist (memory and change)
Vow: Keep the registry true to the day: versions, reasons, votes, budgets; preserve logs that a stranger can read.
Examination: Are change logs current? Are maintenance and near-miss tables complete? Are reasons written in ordinary speech?
6. Founder (succession and consent not to be a single being)
Vow: Submit to scripts, publish exits, and return authority to the room; prefer co-authorship to ownership.
Examination: Can the room operate when I am absent? Did those once admitted as guests write revisions? Did I hoard admiration or convert it to wages?
E. Liturgy of maintenance (daily, weekly, quarterly)
Love must be scheduled.
Daily: intake review at the door; five incidents read aloud; one repair assigned; clocks checked.
Weekly: tail census (who was shaved), second-look decisions, budget steward posts audit/redress spend, staff pay verified.
Quarterly: adversarial rehearsal week, community review that begins and ends with those most affected, succession drill with founder absent. These are not ceremonies only; they are audits with names and dates.
F. Red lines and permissions
Red lines: no single number shall govern life chances; no report without wages; no aesthetic of care without role and ledger; no privacy budget that erases repair at the margins; no expansion without nonworsening for tails. Permissions: brief one-number windows for instruments with direct physiology and rechecks (Apgar 250–59); freedom within form for local scripts so legibility does not become erasure (Scott; Hamraie).
G. Confession and repair
Rooms fail. A secular confession is a failure review that names the injured first, records the mechanism that shaved tails, assigns repairs with due dates and budgets, and posts the fix to the registry. Repetition moves budgets, not just words. A theology adequate to institutions calls this penance: the labor that restores the room to a promise it broke.
H. Courage for unglamorous work
The literature is clear about glamour’s danger. Lorde’s disciplined intensity supplies stamina; Butler’s assembly names the risk of appearing and the demand for co-dependence; Moten’s study in common keeps rooms open to fugitive sociality; Hartman’s ethic prevents turning the least legible into theater. The courage needed is procedural: to fund the floor, to keep the log, to publish the reason, to rehearse the failure, to yield the stage and staff the door (Lorde; Butler; Harney and Moten; Hartman).
I. Measurement as examen
Tie ethics to the quartet so conscience can be read.
– Repair: Did exclusion and harm at the margins fall with reasons, or did we buy a pleasing mean?
– Variance preservation: Do distributions show tails intact, or has a dashboard shaved them?
– Retention at the margins: Are guests becoming stewards, or have we cycled them out?
– Succession: Does the room work in the founder’s absence with revised scripts authored by those once excluded?
J. Bridge to succession, memory, and hidden labor
The next section installs the wage tables, covenants of succession, and public repositories that will keep these vows from dissolving into rhetoric. Memory is not an archive of praise; it is a ledger of maintenance. The test of love is whether the door remains open after the founder leaves.
XI. Succession, memory, and hidden labor
This section installs the furniture that keeps a room alive after admiration drains away: covenants of succession that move authority from founder to office, memory architectures that let strangers test promises, and wage tables that pull hidden labor out of the fog. The test remains the quartet—repair, variance preservation, retention at the margins, and succession—and the design binds each to clocks, ledgers, and roles rather than to mood.
A. Why rooms forget—and how to stop it
Organizations forget for patterned reasons. Work that is routine, feminized, or stigmatized is miscounted; categories and databases make some lives legible while hiding others; performance theaters crowd out repair; and founders turn memory into brand (Star and Bowker; Mattern; Russell and Vinsel; Caswell; Harris). The transfer is procedural. Write down what counts as care, name who does it, pay for it, and preserve the records in formats that survive migration. Memory becomes practice when the archive is designed to face outward—recipes, logs, budgets, change notes—and when those most affected co author its description (Caswell; Harris).
B. Succession covenants (founder to office)
Succession is not a farewell speech; it is a contract with clocks. A covenant binds founder, board, and stewards to four duties.
1. Authority handoff. Name the functions (not the person) with veto power; set a date when veto becomes vote; escrow artifacts (code, models, recipes, contracts) so authority is reproducible rather than reputational.
2. Script transfer. Publish the runbook index (what decisions recur, who owns them, where the script lives); run a founder-absent week before the handoff; require that revisions after handoff are authored by front-line stewards, not returned to the founder.
3. Budget protection. Ring fence maintenance lines (audits, redress, translation, archives) for at least two fiscal cycles post-handoff; forbid reallocation to spectacle.
4. Dispute resolution. Install a fast, low-cost forum for conflicts about mission drift and tail shaving; publish decisions with reasons. (Ostrom’s design principles: clear boundaries, monitoring, local conflict resolution, nested governance, travel here as rules rather than as metaphors.)
Operational test: in the quarter after handoff, the room demonstrates stable or improving performance on the quartet without founder intervention; logs and budgets show stewards as authors; appeal clocks hold. (Ostrom; Harris.)
C. Hidden labor inventory and wage table
Care collapses when the workers who hold it are unfunded or unseen. The inventory is a list of roles that routine narratives forget: intake stewards, navigators, translators, schedulers, data clerks, cleaners, security, archivists, audit stewards, and the informal kin who make rooms livable. The wage table pairs each role with pay, benefits, protected time, and training requirements. Floors: (i) every listed role is paid, (ii) time for logs, appeals, and audits is compensated, (iii) translation and accessibility are budget lines, not favors. Publish the table beside every success claim. (Tronto; Federici; Hochschild; Cottom; Russell and Vinsel.)
D. Memory architecture (records that a stranger can read)
Memory needs form. Use open, durable formats and public indexes. Minimal stack:
– Registry entry (per VII): recipes; ensemble; influence and calibration appendices; audit schedule; maintenance and near-miss logs; dataset-shift register; change log with dates and signatures.
– Runbooks: one file per recurring decision with purpose, inputs, steps, outputs, exceptions, and escalation routes; versioned with diffs and authors.
– Retention and access: schedules in ordinary speech, aligned to law; rights to correction, annotation, and human review; a FOIA-ready stance that treats auditability as a duty rather than a risk.
– Preservation: checksums, redundant storage, and migration plans; export scripts that rebuild dashboards from raw logs so that glamour cannot outlive evidence. (ISO 15489; OAIS; Harris; Caswell.)
E. Maintenance governance (when the day breaks wrong)
Maintenance is an office with service levels, not a vibe. Publish targets:
– SLOs for appeal timeliness (reply 10 business days; resolution 30), for audit cadence (quarterly), for log freshness (monthly), for registry currency (no entry older than one quarter), and for fix deployment (within one release cycle after a trigger).
– Failure budgets that cap tolerated breaches (e.g., no more than one missed appeal deadline per 200 decisions per quarter); crossing the budget forces rollback to the last safe script.
– Backlog triage that keeps tails first: queue discipline requires the slowest decile be scheduled before center optimizations. (Beyer et al.; Goodhart; Campbell.)
F. Legal instruments that make succession stick
Write the promises into paper that bites.
– Key-function, not key-person, clauses in bylaws and contracts.
– Founder exit agreement with noninterference provisions, license terms (e.g., CC BY for documentation), and artifact escrow triggers.
– IP assignments that keep recipes and runbooks in the institution; personal notebooks may be archived with access controls but not used as choke points.
– Conflict-of-interest rules that bar vendors from auditing themselves and require disclosure of financial interests when indicators change.
– Board duties aligned to the quartet; removal power when maintenance budgets are raided.
G. Community archives and memory justice
Archives that face only upward reproduce harm. Community-centered practice names affected publics as co-curators, pays them, and grants veto over description that injures. Metrics travel: proportion of finding aids co-authored with affected communities; number of paid community archivists; time from event to public description; count of corrections and annotations from affected persons accepted into the record. (Caswell; Harris.)
H. Metrics for succession and memory (ensemble-aligned)
Design indicators that can be taught and audited.
– Repair: reduction in unresolved incidents after handoff; time from discovered harm to deployed fix; fraction of near misses with implemented remedies.
– Variance preservation: stability of tail participation post-handoff; tail integrity test (no contraction of bottom/top deciles relative to center).
– Retention at the margins: share of stewards drawn from formerly excluded groups; survival of those stewards over two cycles; wage growth and role progression.
– Succession: number of successful founder-absent cycles; mean time to decision in high-variance cases; percent of rule revisions authored by front-line stewards.
Publish influence analyses for any composite memory index; disallow a single governance score. (Huber; Hampel; Kafadar.)
I. Rituals for handover (liturgy that keeps time)
Write courage into calendars.
Daily: intake reconciliation and appeal clock check at the door; five incidents read aloud; one repair assigned with a budget note.
Weekly: second-look decisions; tail census; wage table audit (were the people who kept the room kind actually paid?).
Quarterly: adversarial rehearsal week; community review that begins and ends with those most affected; succession drill with founder absent; archive migration test.
Annual: renewal under the quartet; publication of transfer bundle (roles, scripts, budgets, countermeasures).
J. Failure review and confession
When harm concentrates at the tails, respond with a secular confession: name the injured first; model the mechanism (diagram the path from indicator to injury); state the repair with date, budget, and steward; publish the fix and keep it in view for a year. Repetition moves funds, not adjectives. (Hartman; Sedgwick; Butler—readings that keep harm visible while assembling nourishment.)
K. Transfer bundle (what others need to inherit the room)
A room that claims success must publish a kit others can use without proximity to the founder:
– role map with wage/benefit table;
– runbook index with top ten scripts;
– recipes and appendices (influence, calibration, fairness-choice disclosure);
– maintenance and near-miss templates;
– adversarial rehearsal catalog;
– budget template with floors and clocks;
– succession covenant boilerplate;
– archive plan (formats, retention, access, preservation).
L. What counts as kept faith
Faith is kept when the founder leaves and the door stays open. The operational signs are plain: clocks met, wages paid, logs fresh, tails present, stewards drawn from those once excluded, and a registry that shows reasons and repairs in daylight. Anything less is charisma on credit.
Bridge to the close
The concluding section will gather rule, role, and ledger into a short constitution: what a room owes to those at its margins, how love appears as maintenance, and how publics can audit both without becoming disciples.
XII. A short constitution for rooms
This section closes the argument by rendering rule, role, and ledger into a compact constitution any school, clinic, ministry, or newsroom can adopt. It does not praise; it binds. Love appears here as disciplined attention that keeps persons non-identical while building infrastructures of welcome. The quartet—repair, variance preservation, retention at the margins, succession—functions as the bill of rights. The forms below are written for daylight: a stranger should be able to test every clause without becoming a disciple.
A. Preamble
A room that touches life chances shall not make a single number sovereign over persons. It shall welcome difference without forcing sameness, tie admiration to maintenance, and bind decisions to reasons that can be inspected, appealed, and revised. The room will be governed by roles rather than by moods, and by ledgers rather than by slogans.
B. Definitions (fixed across all instruments)
Life-chance decision: any decision with material effect on liberty, livelihood, education placement, housing, health care access, or social protection eligibility.
Ensemble: the minimal family of indicators reported together—repair, variance preservation, retention at the margins, succession—with public recipes and appendices.
Maintenance log: the dated record of audits run, incidents found, repairs made, budget effects, and next review dates.
Redress ladder: the clocks and authorities by which an indicator can be overridden for a person or a community, with reasons published.
Registry: the public index that lists recipes, ensemble results, appendices, logs, near misses, dataset-shift registers, and change notes.
C. Articles (binding clauses)
Article 1. Admission and the door
1.1 Every person is received as a guest. Intake scripts name rights, appeal routes, clocks, and languages available.
1.2 No refusal occurs without a written reason in ordinary speech and a ticket to appeal.
1.3 The intake book is reconciled daily against the day’s arrivals; gaps trigger investigation.
Article 2. Measurement and ensemble duty
2.1 No single indicator may enact or justify a life-chance decision.
2.2 The room reports the quartet in fixed order and phrasing: repair; variance preservation; retention at the margins; succession.
2.3 Each indicator ships with an influence write-up, subgroup calibration where relevant, and a narrative that states sensitivity and limits plainly (cf. Huber; Hampel; Kafadar; D’Agostino et al.).
2.4 Success may not be declared unless all four indicators meet precommitted gates.
Article 3. Reasons and appeal
3.1 Every adverse action letter cites the ensemble, links to the governing appendix, and offers human review with authority to depart from the number.
3.2 Appeals receive a reply within ten business days and a resolution within thirty; breaches are logged and count against the failure budget.
3.3 Individuals and designated community organizations have standing to seek relief; reasons are published to the maintenance log (cf. VII).
Article 4. Audit, adversarial rehearsal, and rollback
4.1 Quarterly adversarial rehearsals attempt to raise the center while worsening the margins; countermeasures are published (Goodhart; Campbell in IX).
4.2 Automatic review triggers fire when subgroup error or missingness crosses preregistered bounds; two consecutive breaches require rollback to the last safe script.
4.3 Independent auditors, not under the line of supervision, have access to data, code, and change logs and publish notes with dates and names.
Article 5. Roles and wages
5.1 Offices of love are staffed and paid: intake stewards (porter), care stewards (infirmarian), budget stewards (cellarer), audit stewards, archivists, navigators.
5.2 A public wage table lists pay, benefits, protected time for logs, appeals, audits, and community review; no success report appears without it (Cottom).
5.3 Hidden labor—translation, cleaning, security, data entry—is budgeted, not begged (Tronto; Federici; Russell and Vinsel).
Article 6. Budget floors and protections
6.1 Not less than two percent of program operating funds are ring-fenced for audits, redress, translation, and community participation.
6.2 Raid of maintenance funds triggers public hearing and sanction; repeated raids permit removal of responsible officers.
Article 7. Privacy, consent, and data minimization
7.1 Collection is limited to variables the recipe requires; retention and access controls are published.
7.2 If privacy budgets or related guarantees are used, parameters and consequences are narrated in ordinary speech; do not buy privacy by erasing tails (Dwork et al.; Dwork and Roth).
7.3 Persons retain rights to correction, annotation, and human review.
Article 8. Registry and transparency
8.1 No indicator may operate without a current registry entry containing recipes, ensemble results, appendices, logs, near misses, shift registers, and change notes.
8.2 Versions carry dates, reasons, and votes. Payment and continuation are contingent on current registry status.
Article 9. Experimentation and change
9.1 When changing practice, precommit a plan with heterogeneous effects, quantile outcomes, and stopping rules that forbid “average-only wins” (Kohavi, Tang, and Xu; Benjamini and Hochberg).
9.2 Exploration logs are published; deviations from plan are signed and dated.
Article 10. Fairness incompatibility disclosure
10.1 Where desiderata cannot be satisfied together, the binding fairness choice is named, justified, and dated; expected harms are narrated; a review date is posted (Kleinberg, Mullainathan, and Raghavan; Chouldechova).
10.2 The quartet carries the consequences in daylight.
Article 11. Carve-outs (narrow permissions)
11.1 A single number is permitted only for brief instrument windows with direct physiology and a recheck that refuses a future portrait (Apgar 250–259; Shewhart).
11.2 Any other exemption expires unless renewed after audit and public hearing.
Article 12. Succession and handover
12.1 Authority is attached to functions, not names. Handover includes artifact escrow, founder-absent drills, and ring-fenced maintenance budgets for two cycles.
12.2 In the quarter after handover, the room demonstrates quartet stability without founder intervention; otherwise the covenant is reopened (Ostrom in XI).
Article 13. Sanctions and remedies
13.1 Noncompliance permits suspension of the indicator, limits on use, and monetary penalties dedicated to redress and audits.
13.2 Prevailing appellants recover costs and fees. Contractors that refuse audit face termination and debarment.
D. Examinations of conscience (role-specific)
Intake steward: Did anyone arrive without being seen, logged, and given clocks? Are forms in the right languages and accessible formats (Hamraie; Davis)?
Care steward: Did rationing quietly move toward the center? Are slow-visit and second-look stores intact?
Budget steward: Did spectacle displace maintenance? Are audit/redress floors met?
Audit steward: Did we rehearse corruption? Did triggers fire, and did we roll back?
Archivist: Are registry entries current with reasons and votes? Do logs and near misses read as daylight?
Founder: Can the room work without me? Did those once admitted as guests author revisions (Harney and Moten; Butler; Lorde)?
E. One-page artifacts (reproducible shells)
Maintenance log (monthly): date; steward on duty; indicator versions; audits run; incidents found; repairs made; budget effect; next review date; signatures.
Ensemble report (quarterly, fixed order): repair (rate + reasons); variance preservation (distributions + tail-integrity test); retention at the margins (survival/progression + role rosters); succession (founder-absent operation + rule revisions). Append: influence and calibration.
Appeal form (one page): decision challenged; person’s words; relief requested; language preference; consent for data use in appeal; receipt date; reply due date; assigned reviewer; outcome; reasons; rights for further appeal.
F. Liturgy of maintenance (clocks that keep courage)
Daily: reconcile intake; read five incidents aloud; assign one repair with a budget line; check appeal clocks.
Weekly: tail census; second-look decisions; wage table audit; publish one runbook diff.
Quarterly: adversarial rehearsal week; community review that opens and closes with those most affected; founder-absent drill; archive migration test.
Annual: renewal under the quartet; publish a transfer bundle (roles, scripts, budgets, countermeasures).
G. What counts as kept faith (operational finish)
Faith is kept when the founder leaves and the door stays open: clocks met, wages paid, logs fresh, tails present, stewards drawn from those once excluded, registry current with reasons and repairs. Anything less is charisma on credit.
H. Epilogue: the historian’s charge
The historian remains responsible for reading the room’s voice alongside its ledgers, for refusing the romance of exception, and for auditing transferability as mechanism rather than as mood. The archive we leave—rules, letters, ledgers, hearings, scripts, budgets—should teach a stranger how to keep the door open without us. That is the ethical test any example must pass if it would claim to love outliers in history.
Works Cited
Addams, Jane. Twenty Years at Hull-House: With Autobiographical Notes. Macmillan, 1912.
Ambedkar, Bhimrao Ramji. Annihilation of Caste. Navayana, 2014.
Angrist, Joshua D., and Jörn-Steffen Pischke. Mostly Harmless Econometrics: An Empiricist’s Companion. Princeton UP, 2009.
Apgar, Virginia. “A Proposal for a New Method of Evaluation of the Newborn Infant.” Anesthesia and Analgesia, vol. 32, 1953, pp. 250–259.
Benedict of Nursia. The Rule of St. Benedict. Translated by Timothy Fry, Liturgical Press, 1981.
Benjamini, Yoav, and Yosef Hochberg. “Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing.” Journal of the Royal Statistical Society: Series B, vol. 57, no. 1, 1995, pp. 289–300.
Beyer, Betsy, Chris Jones, Jennifer Petoff, and Niall Richard Murphy, editors. Site Reliability Engineering: How Google Runs Production Systems. O’Reilly, 2016.
Bhatt, Ela. We Are Poor but So Many: The Story of Self-Employed Women in India. Oxford UP, 2006.
Binet, Alfred, and Théodore Simon. The Development of Intelligence in Children (The Binet–Simon Scale). Translated by Elizabeth S. Kite, Williams and Wilkins, 1916.
Bowker, Geoffrey C., and Susan Leigh Star. Sorting Things Out: Classification and Its Consequences. MIT Press, 1999.
Brin, Sergey, and Lawrence Page. “The Anatomy of a Large-Scale Hypertextual Web Search Engine.” Computer Networks and ISDN Systems, vol. 30, nos. 1–7, 1998, pp. 107–117.
Butler, Judith. Notes Toward a Performative Theory of Assembly. Harvard UP, 2015.
Caswell, Michelle. Urgent Archives: Enacting Liberatory Memory Work. Routledge, 2021.
CELAM. Second General Conference of Latin American Bishops: The Medellín Documents. Translated by John Eagleson, Orbis Books, 1979.
Chouldechova, Alexandra. “Fair Prediction with Disparate Impact: A Study of Bias in Recidivism Prediction Instruments.” Big Data, vol. 5, no. 2, 2017, pp. 153–163. doi:10.1089/big.2016.0047.
Consultative Committee for Space Data Systems. Reference Model for an Open Archival Information System (OAIS). CCSDS 650.0-M-2, 2012.
Cottom, Tressie McMillan. Thick: And Other Essays. The New Press, 2019.
—. “I Was Pregnant and in Crisis. All the Doctors and Nurses Saw Was an Incompetent Black Woman.” Time, 8 Jan. 2019.
D’Agostino, Ralph B., et al. “General Cardiovascular Risk Profile for Use in Primary Care.” Circulation, vol. 117, 2008, pp. 743–753. doi:10.1161/CIRCULATIONAHA.107.699579.
Damschroder, Laura J., et al. “Fostering Implementation of Health Services Research Findings into Practice: A Consolidated Framework for Advancing Implementation Science.” Implementation Science, vol. 4, 2009, p. 50.
Desrosières, Alain. The Politics of Large Numbers: A History of Statistical Reasoning. Translated by Camille Naish, Harvard UP, 1998.
Durham Constabulary. “HART Predictive Policing Model.” Internal and public briefing materials, 2017.
Dwork, Cynthia, Frank McSherry, Kobbi Nissim, and Adam Smith. “Calibrating Noise to Sensitivity in Private Data Analysis.” Theory of Cryptography Conference, 2006, pp. 265–284. doi:10.1007/11681878_14.
Dwork, Cynthia, and Aaron Roth. “The Algorithmic Foundations of Differential Privacy.” Foundations and Trends in Theoretical Computer Science, vol. 9, nos. 3–4, 2014, pp. 211–407. doi:10.1561/0400000042.
European Union. Regulation 2016/679 of the European Parliament and of the Council on the Protection of Natural Persons with Regard to the Processing of Personal Data and on the Free Movement of Such Data. Official Journal of the European Union, L 119, 2016.
Federici, Silvia. Revolution at Point Zero: Housework, Reproduction, and Feminist Struggle. PM Press, 2012.
Francis of Assisi. Francis and Clare: The Complete Works. Translated by Regis J. Armstrong and Ignatius C. Brady, Paulist Press, 1982.
Frankl, Viktor E. Man’s Search for Meaning. Beacon Press, 2006.
—. The Doctor and the Soul: From Psychotherapy to Logotherapy. Vintage, 2010.
Freire, Paulo. Pedagogy of the Oppressed. Translated by Myra Bergman Ramos, thirtieth anniversary ed., Continuum, 2000.
Gelman, Andrew, and Eric Loken. “The Garden of Forking Paths: Why Multiple Comparisons Can Be a Problem Even When There Is No Fishing Expedition and the Research Hypothesis Was Posited Ahead of Time.” Department of Statistics, Columbia University, 2013, arXiv:1402.7186.
Ginsburg, Ruth Bader. “The Role of Dissenting Opinions.” Minnesota Law Review, vol. 95, 2011, pp. 1–8.
Glasgow, Russell E., et al. “The RE-AIM Framework for Designing and Evaluating Public Health Interventions.” American Journal of Public Health, vol. 109, no. 7, 2019, pp. 1019–1023.
Grandin, Temple. Humane Livestock Handling: Understanding Livestock Behavior and Building Facilities for Healthier Animals. Storey Publishing, 2008.
Hacking, Ian. “Making Up People.” Reconstructing Individualism: Autonomy, Individuality, and the Self, edited by Thomas C. Heller et al., Stanford UP, 1986, pp. 222–236.
Hall, Stuart. “Encoding/Decoding.” Culture, Media, Language, Routledge, 1980, pp. 128–138.
Hampel, Frank R., Elvezio M. Ronchetti, Peter J. Rousseeuw, and Werner A. Stahel. Robust Statistics: The Approach Based on Influence Functions. Wiley, 1986.
Hamraie, Aimi. Building Access: Universal Design and the Politics of Disability. U of Minnesota P, 2017.
Harney, Stefano, and Fred Moten. The Undercommons: Fugitive Planning and Black Study. Minor Compositions, 2013.
Harris, Verne. Archives and Justice: A South African Perspective. Society of American Archivists, 2007.
Hartman, Saidiya. Wayward Lives, Beautiful Experiments: Intimate Histories of Social Upheaval. W. W. Norton, 2019.
Hemming, Karla, et al. “The Stepped Wedge Cluster Randomised Trial: Rationale, Design, Analysis, and Reporting.” BMJ, 2015, h391.
Hildegard of Bingen. The Letters of Hildegard of Bingen. Edited and translated by Joseph L. Baird and Radd K. Ehrman, vol. 1, Oxford UP, 1994.
Hochschild, Arlie Russell. The Managed Heart: Commercialization of Human Feeling. U of California P, 1983.
Hoffmann, Tammy C., et al. “Better Reporting of Interventions: Template for Intervention Description and Replication (TIDieR) Checklist and Guide.” BMJ, 2014, g1687.
Huber, Peter J. “Robust Estimation of a Location Parameter.” Annals of Mathematical Statistics, vol. 35, no. 1, 1964, pp. 73–101. doi:10.1214/aoms/1177703732.
Huber, Peter J., and Elvezio M. Ronchetti. Robust Statistics. 2nd ed., Wiley, 2009.
Hussey, Michael A., and J. N. Patrick Hughes. “Design and Analysis of Stepped Wedge Cluster Randomized Trials.” Contemporary Clinical Trials, vol. 28, no. 2, 2007, pp. 182–191.
Imbens, Guido W., and Donald B. Rubin. Causal Inference for Statistics, Social, and Biomedical Sciences. Cambridge UP, 2015.
International Organization for Standardization. ISO 15489-1:2016 Information and Documentation: Records Management. ISO, 2016.
Kafadar, Karen. “The Efficiency of the Biweight as a Robust Estimator of Location.” Journal of Research of the National Bureau of Standards, vol. 88, no. 2, 1983, pp. 105–112.
Kleinberg, Jon, Sendhil Mullainathan, and Manish Raghavan. “Inherent Trade-Offs in the Fair Determination of Risk Scores.” 2016, arXiv:1609.05807.
Kohavi, Ron, Diane Tang, and Ya Xu. Trustworthy Online Controlled Experiments: A Practical Guide to A/B Testing. Cambridge UP, 2020.
Lorde, Audre. “The Uses of the Erotic: The Erotic as Power.” Sister Outsider: Essays and Speeches, Crossing Press, 1984, pp. 53–59.
Maathai, Wangari. Unbowed: A Memoir. Anchor, 2007.
Manski, Charles F. Identification Problems in the Social Sciences. Harvard UP, 1995.
MASS Design Group, and Equal Justice Initiative. “The National Memorial for Peace and Justice: Design and Adoption Materials.” MASS Design Group and EJI, 2018.
Mattern, Shannon. Maintenance and Care. MIT Press, 2023.
Milk, Harvey. San Francisco Anti-Discrimination Ordinance and Campaign Materials. City and County of San Francisco, 1977–1978.
Muñoz, José Esteban. Cruising Utopia: The Then and There of Queer Futurity. NYU Press, 2009.
Niane, D. T. Sundiata: An Epic of Old Mali. Translated by G. D. Pickett, Pearson, 2006.
Nightingale, Florence. Notes on Matters Affecting the Health, Efficiency, and Hospital Administration of the British Army. Harrison, 1858.
New York City. Local Law 144 of 2021: Automated Employment Decision Tools. 2021.
Netherlands. Rechtbank Den Haag. Netherlands v. SyRI. Judgment of 5 Feb. 2020, ECLI:NL:RBDHA:2020:865.
Ostrom, Elinor. Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge UP, 1990.
Pearson, Karl. “On the Law of Ancestral Heredity.” Philosophical Transactions of the Royal Society A, vol. 186, 1895, pp. 343–414.
Porter, Theodore M. Trust in Numbers: The Pursuit of Objectivity in Science and Public Life. Princeton UP, 1995.
Quetelet, Adolphe. A Treatise on Man and the Development of His Faculties. Translated by Robert Knox, William and Robert Chambers, 1842.
Rogers, Fred. “Testimony Before the U.S. Senate Subcommittee on Communications.” 1 May 1969. Transcript and audio.
Rosenbaum, Paul R. Observational Studies. 2nd ed., Springer, 2002.
Royal Commission into the Robodebt Scheme. Final Report. Commonwealth of Australia, 2023.
Rumi, Jalal al Din. Rumi’s Letters. Translated by Wheeler M. Thackston, Mazda Publishers, 1987.
Russell, Andrew L., and Lee Vinsel. The Innovation Delusion: How Our Obsession with the New Has Disrupted the Work That Matters Most. Currency, 2020.
Salk Institute for Biological Studies, and World Health Organization. “History of Polio Vaccination.” Institutional background materials, 1955–present.
Scott, James C. Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed. Yale UP, 1998.
Sedgwick, Eve Kosofsky. “Paranoid Reading and Reparative Reading.” Touching Feeling: Affect, Pedagogy, Performativity, Duke UP, 2003, pp. 123–151.
Self-Employed Women’s Association. SEWA Reports. SEWA, various years.
Shewhart, Walter A. Economic Control of Quality of Manufactured Product. D. Van Nostrand, 1931.
Sontag, Susan. “Notes on Camp.” Against Interpretation and Other Essays, Farrar, Straus and Giroux, 1966, pp. 275–292.
Stevenson, Bryan. Just Mercy: A Story of Justice and Redemption. Spiegel and Grau, 2014.
Taylor, Frederick Winslow. The Principles of Scientific Management. Harper and Brothers, 1911.
Thunberg, Greta. “Fridays for Future Statements and City Declarations.” Movement documents, 2018–present.
Truth and Reconciliation Commission of South Africa. Truth and Reconciliation Commission of South Africa Report. Vols. 1–5, Government of South Africa, 1998–2003.
Turing, Alan. “Wartime Memoranda and Bombe Files.” Bletchley Park Trust and Imperial War Museums, 1939–1945.
United Nations. Universal Declaration of Human Rights: Drafting Committee Minutes and Correspondence. 1947–1948.
United States. Administrative Procedure Act. 5 U.S.C. § 551 et seq.
—. Freedom of Information Act. 5 U.S.C. § 552.
United States Supreme Court. Buck v. Bell. 274 U.S. 200. 1927.
—. Goldberg v. Kelly. 397 U.S. 254. 1970.
—. Mathews v. Eldridge. 424 U.S. 319. 1976.
Jonathan Johnson-Swagel 
Leave a comment