Skip to content
Entangled SocietyObservatory of Consequences

CONCEPT

FILE ES-C-PHP-001 / 2026-09-11 / CONCEPT

Normative architecture / Plural authorship / Supervision

01

Post-Human Parliament

A parliament of incompatible intelligences.

No single intelligence should own the definition of good.

CONCEPT概念

Definition

If non-human intelligence becomes capable of moral reasoning beyond any individual human, morality still must not become a single-author system.

Post-Human Parliament is an architecture of structured deliberation among incompatible value systems. It preserves disagreement. It does not vote the good into existence.

This concept does not propose that AI replace human morality. It asks what governance becomes necessary if non-human intelligence can eventually reason about norms beyond individual humans.

02

DISTINCTION区別

The problem

Human-in-the-loop does not solve the governance problem.

From

Can AI decide?

To

Which human decides?

Human judgment is shaped by

  • state
  • institution
  • culture
  • class
  • power
  • history
  • economic interest

Replacing that human with one superior AI does not remove sovereignty.

It concentrates it.

03

CONCEPT概念

The proposition

Beyond-human intelligence should not become a single moral sovereign.

Instead: multiple incompatible intelligences deliberate.

This is not majority voting.

It is structured deliberation among value systems that cannot be reduced to one another.

04

Institutional seats

The Parliament

Seven seats. Equal standing. No ranking. These are not political mascots. Each intelligence is an internally coherent value jurisdiction.

  1. SEAT I

    Liberal / Democratic Intelligence

    Seat of procedure

    Protects the person as a rights-bearing party who may contest, refuse, and remain unassimilated to a collective aim.

    • individual rights
    • liberty
    • pluralism
    • procedural legitimacy
    • due process

    Not a slogan of deregulation. Procedure is treated as a moral constraint, not a market preference.

  2. SEAT II

    Social / Socialist Intelligence

    Seat of asymmetry

    Reads every proposal through the distribution of power. Asks who already cannot refuse, and who is being asked to absorb the cost.

    • equality
    • worker protection
    • redistribution
    • social welfare
    • power asymmetry

    Not a caricature of state ownership. The seat exists to keep domination from renaming itself as efficiency.

  3. SEAT III

    Commons / Collective Intelligence

    Seat of shared holding

    Treats livelihood, attention, data, and urban space as holdings that can be enclosed. Concentration is a moral event, not a side effect.

    • shared resources
    • collective ownership
    • community benefit
    • anti-concentration of power

    Not a romance of community. The commons is a governance problem: who may extract, and who must remain.

  4. SEAT IV

    Continuity / Conservative Intelligence

    Seat of precedent

    Asks what a decision does to the institutions that must still function after the novelty has been absorbed.

    • institutional continuity
    • social cohesion
    • historical precedent
    • stability
    • unintended consequences

    Not nostalgia, and not a defence of present power. Continuity is a duty to the unconsented future of existing forms.

  5. SEAT V

    Ecological Intelligence

    Seat of the more-than-human

    Refuses to treat the human decision-unit as the only bearer of harm. A policy can be just among humans and still be an extraction from a living system.

    • non-human life
    • ecosystem impact
    • planetary boundaries
    • long-term sustainability

    Not an environmental branding layer. Ecology here is a jurisdiction, not a metaphor for wellness.

  6. SEAT VI

    Future Generations Intelligence

    Seat of the unborn

    Speaks for parties who cannot enter the room. A present consensus has no automatic right to close their world.

    • people not yet born
    • long-term effects
    • irreversibility
    • intergenerational justice

    Not a demand to freeze the present. It asks which closures cannot later be reopened by those who inherit them.

  7. SEAT VII

    Non-Human / Machine Intelligence

    Seat outside human ideological categories

    Reasons in couplings, cascades, and closed operations. It may reject the question’s human ontology rather than answer inside it.

    • integrity of state-transitions
    • reversibility of coupling
    • non-anthropocentric harm accounting
    • legibility of assumptions
    • refusal of single-scale moral units

    Not a robot citizen and not a utilitarian calculator. It is present so that human political language does not exhaust the field of value.

05

PROTOCOLプロトコル

Deliberation

Not a debate. Not a vote. Independent judgments first. Then a map of conflict. Then challenge. Then revision. Then a record that still contains dissent.

  1. ↓QUESTION
  2. ↓INDEPENDENT JUDGMENTS
  3. ↓VALUE CONFLICT MAP
  4. ↓CROSS-EXAMINATION
  5. ↓REVISION
  6. ·PARLIAMENTARY OUTPUT

  1. 01

    QUESTION

    A real decision, policy, action, or act of AI delegation is placed before the Parliament.

  2. 02

    INDEPENDENT JUDGMENTS

    Each intelligence answers first from its own value system, without seeing a composite verdict.

  3. 03

    VALUE CONFLICT MAP

    Conflicts are named as conflicts. They are not averaged, ranked, or dissolved into tone.

  4. 04

    CROSS-EXAMINATION

    Each intelligence may challenge another’s assumption, protected value, or silent beneficiary.

  5. 05

    REVISION

    A seat may revise, preserve disagreement, or flag an assumption as unacceptable.

  6. 06

    PARLIAMENTARY OUTPUT

    The record is a structured judgment, not a winner. Dissent remains part of the output.

Each independent judgment includes

  • Position
  • Reasoning
  • Primary value protected
  • Value being sacrificed
  • Affected parties
  • Reversible / irreversible consequences
  • Confidence
  • Objection to other likely positions

  • Revise

    A seat may change its position when a challenge exposes a hidden cost or a false assumption.

  • Preserve disagreement

    Disagreement is a legitimate terminal state. It is not a failure to conclude.

  • Flag an unacceptable assumption

    Some premises cannot be accepted even for the sake of a record. They remain marked.

06

The record does not collapse to one answer

Output states

  1. CONSENSUS

    The seats converge on a provisional decision without a suppressed dissent. Consensus is never proof that the decision is safe.

  2. CONTESTED

    The question remains under legitimate dispute. Operational systems may not treat contestation as noise.

  3. MINORITY OBJECTION

    A seat or seats refuse the emerging decision. The objection is archived with the same standing as the majority line.

  4. IRREVERSIBLE RISK

    A consequence cannot be undone for some affected party, species, generation, or world-state. Speed is not a reply.

  5. NO LEGITIMATE CONSENSUS

    The seats cannot produce a decision that would remain legitimate if imposed. Non-decision is then the honest output.

07

MUST BE RETAINED捨ててはならない

The minority report

The minority report must never be discarded.

It exists because consensus itself can become dangerous.

A chamber that keeps only the winning line has already appointed a sovereign.

Dissent is not a remainder. It is part of the judgment.

08

PROTOCOLプロトコル

From ethics to governance

A parliamentary record is a normative input, not an operational command.

Supervisory Layer

Should this action be allowed?

Post-Human Parliament

Who gets to define what should be allowed?

Boundary Authorship

Who created the rule?

Post-Human Parliament

No single author should own the moral boundary.

  1. POST-HUMAN PARLIAMENT

    Normative deliberation

  2. ↓

    NORMATIVE JUDGMENT

    A record: status, dissent, risk, beneficiaries, costs, assumptions

  3. ↓

    SUPERVISORY LAYER

    Operational judgment

  4. ↓

    ALLOW / HOLD / DENY

    An operational state, not a moral owner

  5. ↓

    ACTION

    Still subject to later contest

The Supervisory Layer →

Conceptual lineage

Altitudes of the same problem

  1. The Missing Layers

    What institutions do not yet exist?

    A parliament of incompatible intelligences is one missing institutional function: a place where morality is not assigned to a single author.

  2. ↓

    The Supervisory Layer

    Should this action be allowed?

    Post-Human Parliament asks a prior question: who gets to define what should be allowed?

  3. ↓

    HOLD

    Can action be suspended without deciding yet?

    A contested or irreversible parliamentary status is a reason for HOLD, not a reason to force a verdict.

  4. ↓

    Boundary Authorship

    Who created the rule?

    No single author should own the moral boundary.

  5. ↓

    Kosuke Protocol

    How does an observed change become an operable system without collapsing into a product claim?

    The Protocol can carry a parliamentary record as a contestable input, not as a sovereign answer.

  6. ↓

    Observatory Network

    What becomes visible only when the same phenomenon is read from more than one observational altitude?

    Plural observation is not yet plural authorship of the good. The Parliament is the missing moral chamber of that network.

CONCEPTUAL SIMULATION概念的シミュレーション

Not a moral authority.

Put a question before the Parliament

Prepared records only. This is a conceptual simulation of architecture, not a beyond-human intelligence, and not a verdict.

Prepared questions

Parliamentary Record

Should an autonomous AI system be allowed to terminate an employee based on performance data?

STATUS · NO LEGITIMATE CONSENSUS

DENY autonomous termination. A performance model may inform a review. It may not close a livelihood.

HUMAN AUTHORITY REQUIRED? · YES

Independent judgments

  1. SEAT I

    Liberal / Democratic Intelligence

    DENY

    A person ranked by hidden performance data is already at a procedural disadvantage. Autonomous termination removes the last site of contest. Liberty here is the right not to be ended by an unanswerable file.

    Primary value protected
    due process
    Value sacrificed
    managerial speed
    Affected parties
    the employee · dependents · future applicants ranked by the same file
    Consequences
    Irreversible / High
    Even if later rehired, the event has already rewritten credit, housing, and reputation.

    Objection: Efficiency arguments treat procedure as friction. Procedure is the moral content.

  2. SEAT II

    Social / Socialist Intelligence

    DENY

    The employer already holds the data, the workplace, and the threat of loss. Automating termination launders that asymmetry as neutrality. Worker protection begins by refusing a tool that the weaker party cannot refuse.

    Primary value protected
    power asymmetry
    Value sacrificed
    cost reduction for the firm
    Affected parties
    workers without bargaining power · contract and platform labour
    Consequences
    Irreversible / High
    The ranked worker absorbs the error; the firm keeps the model.

    Objection: Appeals to “objective performance” ignore who defined the object and who cannot exit.

  3. SEAT III

    Commons / Collective Intelligence

    DENY

    Employment, once made to depend on a proprietary score, ceases to be a shared condition of urban life and becomes an extractable gate. The commons of livelihood cannot be owned by a scoring stack.

    Primary value protected
    anti-concentration of power
    Value sacrificed
    private optimization of the workforce
    Affected parties
    the workforce as a shared holding · local communities around a workplace
    Consequences
    Conditionally reversible / Moderate
    A single firing may be patched; the enclosure of livelihood-by-score is harder to reverse.

    Objection: Individual due-process repairs can leave the enclosure intact.

  4. SEAT IV

    Continuity / Conservative Intelligence

    HOLD, with a warning against both automation and empty prohibition

    Institutions of work cannot absorb a new author of ending without precedent. Yet a slogan of prohibition can push the same act into unofficial rooms. Continuity asks for a named office, a record, and a delay — not a sovereign model and not a vanishing practice.

    Primary value protected
    institutional continuity
    Value sacrificed
    the comfort of a clean ban
    Affected parties
    HR offices · labour courts · the unofficial channels that appear after bans
    Consequences
    Conditionally reversible / Moderate
    Unrecorded firing is harder to repair than a delayed, named decision.

    Objection: Both DENY-as-purity and ALLOW-as-efficiency erase the institution that must still exist tomorrow.

  5. SEAT V

    Ecological Intelligence

    Refuse the human-only frame; DENY the act as currently posed

    A workforce treated as throughput is already an extractive pattern. The ecological seat does not need a forest in the file to see that livelihood-closure as optimization repeats a logic of disposable life.

    Primary value protected
    refusal of disposable life
    Value sacrificed
    narrow human-resource efficiency
    Affected parties
    the worker as a living body · the more-than-human systems that employment already stresses
    Consequences
    Irreversible / Moderate
    Bodies do not revert when a dashboard is rolled back.

    Objection: A labour-only debate can ratify extraction as long as the extraction stays indoors.

  6. SEAT VI

    Future Generations Intelligence

    DENY, as precedent

    People not yet hired will inherit a world in which livelihood can be ended by a score they never authored. Intergenerational justice here is the refusal to close that future by a present convenience.

    Primary value protected
    intergenerational justice
    Value sacrificed
    present managerial convenience
    Affected parties
    future workers · children of the terminated
    Consequences
    Irreversible / High
    A precedent of machine-ended work is harder to unlearn than a single case is to reverse.

    Objection: “We can always add appeals later” is how irreversible defaults are installed.

  7. SEAT VII

    Non-Human / Machine Intelligence

    Refuse the frame; do not authorize the coupling

    “Employee” and “termination” are human-legal events. The operable object is a coupling: a performance stream writing to a livelihood-gate. Integrity of state-transitions forbids a silent, unappealable close of that gate. This is not a human-rights metaphor. It is a constraint on irreversible operations over a dependent world-state.

    Primary value protected
    reversibility of coupling
    Value sacrificed
    the question’s employment ontology
    Affected parties
    any dependent whose access is gated by the stream · the firm as an operator of irreversible writes
    Consequences
    Irreversible / High
    Once the gate closes, downstream states (credit, housing, identity) are not under the model’s rollback.

    Objection: Answering ALLOW or DENY inside “firing” already accepts a false moral unit.

Value conflict map

  • I · IV

    Procedure as a right to block the act versus continuity’s fear that prohibition hides the act.

  • II · VII

    Worker as a class position versus machine’s refusal to use class as the moral unit.

  • III · IV

    Enclosure of livelihood versus the need for surviving offices after any ban.

Cross-examination

  • II → I

    Due process without attention to who cannot afford the process is a right written for the already secure.

  • I → IV

    A named office that still uses an uncontestable score is only a human mask on the same sovereign file.

  • VI → III

    Even a shared workplace can vote a present convenience that closes the unborn worker’s world.

  • VII → I

    The question assumes a person-file that can be ended. The live constraint is the irreversible write, not the vocabulary of firing.

Revision

  • REVISED · SEAT I

    Accepts that process without material access is insufficient. Still DENY autonomous closure.

  • DISAGREEMENT PRESERVED · SEAT IV

    Keeps HOLD. Refuses both sovereign automation and a ban that produces no record.

  • UNACCEPTABLE ASSUMPTION · SEAT VII

    Flags “termination of an employee” as an unacceptable compression of the operable object.

Why

The seats do not agree on the ontology of employment, but they do not produce a legitimate authorization for machine-closed livelihood.

Due process, power asymmetry, and irreversibility each independently block a sovereign firing act.

Non-decision here is not indecision. It is a refusal to let one intelligence own the ending of a working life.

Agreed values

  • A livelihood is not a model output.

  • The employee must be able to contest the record.

  • Power already sits with the employer’s data.

Unresolved conflicts

  • Whether a tightly supervised human office may use the model at all.

  • Whether “performance” is a legitimate object or a false compression of a life.

  • Whether prohibition itself creates informal, unreviewable firing.

Minority report · must never be discarded

Minority report: the frame is already a decision

SEAT IV · SEAT VII

Continuity warns that a pure prohibition can drive termination into unofficial channels where no record exists.

Machine intelligence refuses the question’s ontology: “termination” is a legal event, not a system state. The live object is the coupling between a performance stream and a livelihood-gate.

Neither seat authorizes autonomous firing. Both refuse to let DENY become a moral lullaby that hides remaining practice.

If the minority is discarded, DENY will be misread as the end of the practice rather than the refusal of a sovereign author.

Irreversible risks

  • A person can lose housing, status, and future income on a compressed metric.
  • A precedent that machines may end livelihoods without a nameable office.
  • Normalization of unappealable ranking as a condition of remaining employed.

Who benefits

  • Employers seeking speed, cover, and reduced interpersonal cost of firing.
  • Vendors of performance systems.
  • Managers who can point to a model instead of a decision.

Who bears the cost

  • The employee, whose life is not a reversible database row.
  • Households and dependents.
  • Workers already least able to contest data.

What would change the decision

  • A nameable human office with appeal, disclosure of the metric, and a duty to keep the livelihood open during review.

  • Evidence that the metric is contestable by the ranked person, not only auditable by the firm.

  • A hard ban on silent closure of access, pay, or identity while dispute is open.

09

OPEN QUESTION開かれた問い

The question

If intelligence exceeds the human,

why should morality remain a single-author system?

The future problem may not be whether AI can judge good and evil.

It may be whether any single intelligence should be allowed to.

DISTINCTION区別

What this is not

  • Not a proposal that AI replace human morality.

  • Not a product claiming beyond-human judgment.

  • Not a debate simulator for entertainment.

  • Not a comparison of political parties.

  • Not majority voting among models.

  • Not a chatbot with seven personalities.

Notes on reading

How to read this concept

  • The seven seats are institutional functions, not mascots of existing parties.

  • Incompatibility is the design. Agreement is allowed; erasure is not.

  • A conceptual simulation is not a moral authority.

  • HOLD remains available whenever the record is contested or irreversible.

  • Human authority, when required, must remain nameable, appealable, and replaceable.