〒220-0011Judgment Sovereignty in the Age of AGI
A Civilizational Design Framework for AI Delegation and Human Responsibility
Mitsuaki Koga
Civilizational Design Philosopher
Designer of Judgment Sovereignty for the Age of AGI
English Overview — Version 1.0, 2026
This paper presents an introductory English overview of Mitsuaki Koga’s Theory of Judgment Sovereignty in the Age of AGI. It examines what human beings may delegate to AI, what must remain under human judgment, and how responsibility should be structured as AI enters the domain of decision-making.
This is an independently structured English overview, not a complete translation of the original Japanese treatise. The complete Japanese treatise has not yet been publicly released.
Abstract
Artificial intelligence is entering not only the domain of human work but also the domain of human judgment.
AI systems can organize information, compare alternatives, identify risks, predict outcomes, and recommend courses of action. As these systems become more capable and move toward artificial general intelligence, the central question will no longer be merely whether human beings should use AI.
The deeper question is:
What should human beings entrust to AI and AGI, what should they retain, and what must they continue to assume responsibility for until the very end?
This paper introduces the concept of judgment sovereignty: the human capacity to determine, according to human values and purposes, what should be regarded as right, what should be chosen, what must be rejected, and which consequences human beings will accept and bear.
Judgment sovereignty does not require the rejection of AI. It provides a framework for using AI without surrendering the human center of judgment.
At the core of the theory is the Principle of Judgment Classification, which distinguishes four categories:
judgments that may be delegated to AI;
judgments in which AI may assist but human verification is required;
judgments in which AI may assist but final judgment must remain human;
judgments that must be protected as an inviolable human domain.
This framework goes beyond the general requirement that a human remain “in the loop.” It asks a prior question: Which judgments may legitimately be delegated, and which must remain under human sovereignty?
The theory connects AI governance with human dignity, responsibility, management, technology, and the design of civilization. Its ultimate aim is not to prevent the development of intelligent systems, but to establish a civilization in which increasingly capable intelligence serves human dignity rather than replacing human beings as the authors of their purposes and responsibilities.
Keywords: judgment sovereignty, artificial general intelligence, human dignity, AI governance, human oversight, judgment classification, responsibility, civilizational design, human-centered philosophy
1. The Central Question of the AGI Era
Artificial intelligence is already becoming part of everyday work and life.
It writes documents.
It organizes information.
It compares alternatives.
It analyzes markets.
It predicts risks.
It assists recruitment and evaluation.
It supports managerial decision-making.
These capabilities will continue to develop.
Whether artificial general intelligence will emerge at a particular time or in a particular form remains uncertain. The essential issue, however, is not limited to predicting when AGI will arrive.
The direction of change is already clear.
Artificial intelligence is becoming deeply involved in human intellectual work, organizational decisions, education, medicine, economics, public administration, politics, and social institutions.
As this involvement deepens, human beings must decide not only how to use AI, but how far its authority should extend.
The defining question of the AGI era is therefore:
Can human beings use increasingly capable intelligence while retaining sovereignty over their own judgments?
Many discussions of AI focus on performance.
Can AI reason accurately?
Can it make reliable predictions?
Can it outperform human experts?
Can it reduce error and increase efficiency?
These questions are important, but they are not sufficient.
Even if an AI system produces an accurate answer, further questions remain:
What purpose is this answer serving?
Who determined that purpose?
Which values were included or excluded?
Whose interests are protected?
Whose dignity may be harmed?
Which consequences are acceptable?
Who will bear responsibility for the result?
These are not merely technical questions.
They concern the human authorship of purposes, values, choices, and responsibility.
This is the domain of judgment sovereignty.
2. What Is Judgment Sovereignty?
Judgment sovereignty is not merely the right to select one option from a list.
It is:
the human capacity to determine, according to one’s own values, what should be regarded as right, what should be chosen or rejected, and which consequences one will assume responsibility for.
A choice takes place among alternatives that have already been presented.
A judgment goes deeper.
Judgment determines:
what should be treated as the problem;
what should be valued;
what purpose should be pursued;
what must not be sacrificed;
whose interests and dignity must be considered;
which time horizon should be used;
and which consequences the decision-maker is prepared to bear.
Judgment is therefore not simply a comparison between Option A and Option B.
It is an act of deciding which standards should be used to understand the situation in the first place.
AI can compare alternatives.
It can estimate probabilities.
It can identify patterns in past data.
It can propose methods that appear optimal in relation to a defined objective.
But the objective itself must still be examined.
Should profit be maximized at the cost of long-term trust?
Should efficiency take priority over employment stability?
Should a statistically high-risk individual be denied an opportunity?
Should a person be evaluated only through measurable performance?
Should social predictability be placed above freedom?
Should a medically possible intervention be pursued in every case?
The ability to calculate means does not by itself establish the legitimacy of ends.
Human beings do not exercise judgment sovereignty by refusing all technological assistance. They exercise it by retaining the authority and responsibility to question purposes, evaluate values, choose ends, and bear consequences.
Judgment sovereignty means using AI without transferring the center of judgment to AI.
3. The Gradual Hollowing-Out of Human Judgment
The loss of judgment sovereignty may not occur through a dramatic transfer of power.
It may occur gradually, through convenience.
At first, an AI system organizes information.
Next, it compares alternatives.
Then it recommends an option.
Eventually, the recommendation becomes the practical default.
The human being still appears to make the final decision. A manager clicks an approval button. A professional signs a document. An executive authorizes an action.
Formally, the decision remains human.
Substantively, however, the structure of the judgment may already have been created by the AI system.
The system has:
defined the relevant information;
selected the variables;
framed the alternatives;
ranked the risks;
established the apparent optimum;
and presented one conclusion as the rational choice.
The human decision-maker may do little more than confirm that conclusion.
This produces what I call the hollowing-out of judgment sovereignty.
The name of the decision remains human.
The content of the judgment is increasingly constructed by AI.
The responsibility remains with the human being.
The standards of judgment are no longer actively formed within the human being.
The danger of the AGI era is therefore not limited to the possibility that intelligent systems will directly control people.
There is another danger:
human beings may gradually surrender the practice of judgment themselves.
As AI outputs become more coherent, comprehensive, and persuasive, people may begin asking AI before clarifying their own questions.
They may receive an answer before examining their own values.
They may accept a recommended purpose before deciding what they truly seek to accomplish.
The result is not necessarily visible coercion. It may be voluntary dependence on an external structure of judgment.
For this reason, the more capable AI becomes, the more deliberately human beings must preserve the capacity to question, disagree, suspend judgment, and decide.
4. The Principle of Judgment Classification
The statement that “a human being must make the final decision” is not sufficient.
A real decision is not a single, indivisible act.
It may include:
information gathering;
problem definition;
data interpretation;
comparison of alternatives;
prediction;
value judgment;
exception handling;
authorization;
explanation;
and the assumption of responsibility.
A human being may formally approve a decision while AI has already shaped nearly every meaningful element of it.
The question is therefore not only where a human appears in the process.
The prior question is:
Which parts of judgment may be delegated, which require human verification, which must remain human, and which must never be transferred to AI authority?
I call the framework for making these distinctions the Principle of Judgment Classification.
It consists of four categories.
4.1 Judgments That May Be Delegated to AI
The first category includes judgments that are:
routine;
repetitive;
low-risk;
reversible;
governed by clear criteria;
and subject to effective monitoring.
Examples may include:
scheduling;
initial document organization;
routine classification;
standardized calculations;
inventory replenishment suggestions;
preliminary trend analysis;
and the generation of standard-form communications.
Delegation in these areas can reduce human burden and allow people to concentrate on more consequential judgments.
The purpose of judgment sovereignty is not to require human beings to perform every task manually.
Appropriate delegation can strengthen human judgment by freeing attention for questions of purpose, value, responsibility, and exception.
4.2 Judgments Requiring Human Verification
The second category includes judgments in which AI may organize information, generate drafts, or propose alternatives, but a human being must verify the output.
Examples may include:
sales proposals;
customer-response drafts;
preliminary review of employment candidates;
comparisons of operational improvement plans;
contract drafts;
meeting materials;
and AI-generated public communications.
Human verification must be substantive rather than ceremonial.
The verifier must examine:
factual accuracy;
missing context;
bias;
unsuitable assumptions;
conflicts with organizational values;
and possible effects on affected individuals.
A human being who lacks the knowledge, authority, time, or freedom to reject the output is not performing meaningful verification.
4.3 Judgments That Must Be Retained by Human Beings
The third category includes judgments in which AI may provide extensive support, but the final judgment must remain human.
These judgments involve substantial:
responsibility;
value conflict;
interpersonal consequences;
exceptions;
organizational purpose;
or long-term effects.
Examples include:
final recruitment decisions;
employee evaluation;
dismissal;
credit and lending decisions;
termination of important contracts;
responses to serious customer harm;
major pricing changes;
organizational restructuring;
strategic investment;
business withdrawal;
and fundamental management policy.
AI may compare alternatives, estimate outcomes, and identify risks.
It must not become the substantive bearer of the decision.
The human decision-maker must be able to explain:
what was valued;
what was rejected;
who may be affected;
what uncertainty remained;
why the decision was made;
and why the decision-maker is prepared to accept responsibility for it.
4.4 Inviolable Domains of Human Judgment
The fourth category concerns judgments that must not be delegated to AI as the governing authority.
These include judgments concerning:
human dignity;
life and bodily integrity;
personhood;
fundamental freedom;
conscience;
the intrinsic worth of a person;
severe exclusion or punishment;
the meaning and direction of a human life;
and the fundamental direction of society and civilization.
AI may provide information within these domains.
It may identify relevant precedents, compare consequences, or reveal overlooked risks.
It must nevertheless remain in a supporting position.
The final value judgment and responsibility must remain human.
An inviolable domain does not necessarily mean a domain from which AI is completely excluded.
It means a domain in which AI must not be recognized as the sovereign judging subject.
5. Criteria for Classifying Judgments
The four categories cannot be assigned solely by industry or job title.
The same type of activity may belong to different categories depending on its consequences and context.
At least six criteria should be considered.
5.1 Clarity of Responsibility
Who will be responsible for the consequences?
The less clearly responsibility can be assigned, the less deeply the judgment should be delegated to AI.
5.2 Reversibility
Can the decision be corrected?
Judgments that significantly affect a person’s life, employment, reputation, health, freedom, or future require stronger human control because their consequences may be difficult or impossible to reverse.
5.3 Effect on Human Dignity
Does the judgment affect a person’s status, freedom, livelihood, identity, relationships, or intrinsic worth?
The greater the effect on dignity, the stronger the requirement for human judgment.
5.4 Presence of Value Conflict
Does the judgment merely apply a clear rule, or does it require a decision about what should take priority?
When competing values are involved, the decision cannot be reduced to calculation alone.
5.5 Need for Exception Handling
Can the case be properly resolved through standardized rules?
Human beings sometimes recognize circumstances, relationships, suffering, or dignity within an exception that a standardized process fails to represent.
5.6 Weight of Accountability
Must the decision be explained and defended to employees, customers, citizens, families, regulators, or society?
The greater the burden of explanation and responsibility, the more important it is for a human being to remain the substantive decision-maker.
6. Why Human Oversight Is Not Enough
Judgment sovereignty is related to concepts such as:
Human-in-the-Loop;
Human Oversight;
Meaningful Human Control;
Human-in-Command;
and AI orchestration.
These concepts are important.
Human-in-the-Loop concerns where human participation is placed within a process.
Human Oversight concerns how human beings supervise AI systems and their outputs.
Orchestration concerns how human and machine tasks are coordinated within a workflow.
The Principle of Judgment Classification addresses a prior level.
It asks:
Which judgments should a human being be placed in the loop to protect?
Which judgments may be automated?
Which judgments require verification?
Which judgments must remain substantively human?
Which judgments must never be placed under AI authority?
Judgment classification is therefore not a replacement for human oversight or orchestration.
It is a higher-order principle needed to design them correctly.
A human can be present in a workflow without retaining judgment sovereignty.
A person may be instructed to approve an AI recommendation without having:
the necessary information;
the competence to examine it;
the authority to reject it;
the time to deliberate;
or institutional protection when disagreeing.
Such a person is formally involved but substantively subordinate.
Meaningful human involvement requires more than presence.
It requires:
authority to question;
capacity to understand;
freedom to disagree;
access to relevant context;
permission to recognize exceptions;
and responsibility for the final judgment.
7. The Philosophical Foundation: Human-Centered Philosophy
Judgment sovereignty is grounded in Human-Centered Philosophy.
Human-Centered Philosophy does not understand a person merely as:
a member of an institution;
a unit of labor;
a consumer;
an object of evaluation;
a source of risk;
or a target of optimization.
A human being is a being who questions, feels, establishes purposes, forms relationships, chooses, and assumes responsibility.
Human beings do not merely respond to questions.
They also ask what should be questioned.
They do not merely pursue externally assigned objectives.
They can reconsider the objective itself.
A human being may ask:
Why does this problem matter?
Why should this person be protected?
Why does this apparently rational option feel wrong?
What kind of organization do we wish to become?
Which future should we refuse?
What responsibility are we prepared to bear?
Human judgment begins with inquiry.
Inquiry is formed through experience, failure, suffering, relationships, concern for others, hope, and responsibility.
One of the dangers of the AGI era is that answers may arrive before human questions have matured.
A conclusion may be presented before a person has examined what is truly at stake.
An optimized objective may be accepted before an organization has clarified what it exists to protect.
From the standpoint of Human-Centered Philosophy, the process of questioning, hesitating, evaluating, choosing, and assuming responsibility is not an inefficiency to be eliminated.
It is part of human dignity.
8. Human Judgment as a Multilayered Process
The philosophical foundation of judgment sovereignty also includes what I call the Theory of Multilayered Simultaneous Consciousness.
This is a philosophical account of human judgment rather than a claim that judgment is produced by rational calculation alone.
Human judgment includes the simultaneous activity of several dimensions:
bodily awareness;
sensibility;
reason;
intuitive understanding;
spiritual awareness;
and the integrative consciousness that brings these dimensions together.
Bodily awareness receives subtle responses to a situation.
Sensibility perceives discomfort, suffering, harmony, beauty, tension, and resonance.
Reason organizes information, compares alternatives, and constructs explanations.
Intuitive understanding grasps structures and meanings that may not be visible through isolated data.
Spiritual awareness concerns dignity, conscience, mission, prayer, and transcendent meaning.
Integrative consciousness brings these dimensions together and enables a person to form a judgment as one’s own.
AI may powerfully support analytical and rational functions.
It can process information, identify patterns, perform comparisons, and model consequences beyond ordinary human capacity.
But human judgment should not therefore be reduced to its calculable dimensions.
A decision may be numerically defensible and still feel fundamentally wrong.
It may be efficient while treating people carelessly.
It may produce short-term benefits while damaging future trust.
It may conform to standard rules while ignoring exceptional human circumstances.
Judgment sovereignty is therefore not the sovereignty of reason alone.
It is the sovereignty of the integrated human person.
9. Technology That Supports Judgment and Technology That Replaces It
Technology can develop in two different directions.
One direction supports human judgment.
The other gradually replaces it.
Technology that supports judgment:
broadens alternatives;
reveals blind spots;
makes consequences visible;
returns questions to the human decision-maker;
and preserves space for deliberation and disagreement.
Technology that replaces judgment:
presents one optimum as inevitable;
narrows alternatives without transparency;
guides behavior without conscious consent;
makes exceptions difficult;
and encourages human beings to follow rather than judge.
Technology does not automatically respect human dignity.
Its direction depends on the view of the human being embedded in its design.
Is the person treated as an object to be predicted, evaluated, managed, and optimized?
Or is the person treated as a being who asks questions, establishes purposes, makes judgments, and bears responsibility?
AGI should not be designed to eliminate the human being from judgment.
It should be designed as a structure that helps human beings exercise deeper, wider, and more responsible judgment.
Technology should not merely provide answers.
It should help human beings clarify the questions they must answer for themselves.
10. Optimization Civilization and a Civilization of Dignity
The AGI era presents a civilizational choice.
One possible direction is an optimization civilization.
An optimization civilization organizes society around:
efficiency;
prediction;
scoring;
control;
risk reduction;
behavioral guidance;
and measurable performance.
Optimization itself is not wrong.
It can reduce waste, improve safety, expand access, and relieve human beings of unnecessary burdens.
The problem begins when optimization becomes the highest civilizational principle.
Efficiency may then be placed above dignity.
Predictability may be placed above freedom.
Administrative convenience may be placed above human possibility.
Risk reduction may be placed above challenge, growth, and moral courage.
Quantified evaluation may be placed above the unmeasurable weight of the individual person.
The alternative is a Civilization of Dignity.
A Civilization of Dignity does not reject AI, data, prediction, or optimization.
It places them under the higher principles of human dignity and judgment sovereignty.
In such a civilization, human beings must retain:
time to question;
room to hesitate;
freedom to disagree with AI;
the possibility of making exceptions;
respect for values that cannot be fully quantified;
and opportunities to choose and accept responsibility.
The civilizational dividing line is not whether AGI is used.
It is whether AGI is used to substitute for human judgment or to support it.
Optimization must serve dignity.
Dignity must not be made subordinate to optimization.
11. Judgment Sovereignty in Management
In management, AI can become a powerful decision-support structure.
It can assist with:
market analysis;
competitive analysis;
pricing;
recruitment;
employee evaluation;
operational improvement;
risk management;
financial planning;
investment;
strategy;
and customer relations.
The central managerial issue, however, is not simply whether AI should be introduced.
Leaders must determine:
which judgments may be delegated;
which outputs require human verification;
which decisions must remain with executives;
which areas must not be placed under AI authority;
who may challenge an AI recommendation;
who may authorize an exception;
who makes the final decision;
and who assumes responsibility for its consequences.
Without this design, AI may make organizational judgment less clear rather than more advanced.
AI produces a recommendation.
An employee selects it.
A manager approves it.
An executive receives the result.
But who actually judged?
When the structure is unclear, organizations face a gap between formal authorization and substantive responsibility.
This is the practical problem addressed by AI Judgment Sovereignty Design.
AI Judgment Sovereignty Design is the practical discipline of clarifying:
the limits of judgment delegation;
the location of final responsibility;
the value judgments that must remain human;
the authority to reject AI recommendations;
the purposes and values the organization must protect;
and the respective judgment domains of employees, managers, executives, and AI systems.
The essence of AI adoption in management is not the installation of a tool.
It is the design of the organization’s judgment structure.
12. Seven Principles for Preserving Judgment Sovereignty
Judgment sovereignty cannot be protected through general statements of intention.
It requires concrete design principles.
Principle One
Clarify Human Purposes and Values Before Consulting AI
Before asking AI for an answer, human beings must clarify:
the purpose of the decision;
what should be valued;
what must not be sacrificed;
who should be protected;
which time horizon matters;
and what should count as success.
Without these standards, the AI output may silently become the standard itself.
Principle Two
Separate Decision Materials from Final Judgment
Information, summaries, predictions, comparisons, simulations, and recommendations are materials for judgment.
They are not the final judgment.
Even when an AI recommendation is adopted, the human decision-maker must recognize that the decision has been chosen and accepted by a human being.
Principle Three
Define the Boundaries of Delegation, Verification, Retention, and Inviolability
Organizations should explicitly classify their judgments.
They must identify:
what may be delegated;
what requires human verification;
what must remain under human final judgment;
and what must never be delegated as an exercise of AI authority.
Principle Four
Preserve the Ability to Suspend, Modify, or Reject AI Recommendations
Human beings must be able to question AI outputs.
They must be able to record reasons for disagreement, recognize exceptions, modify recommendations, and refuse implementation.
An organization in which people are technically permitted but practically unable to disagree with AI has lost judgment sovereignty.
Principle Five
Identify the Final Human Bearer of Responsibility
Responsibility must not disappear between the system, the operator, the manager, and the executive.
Organizations must clarify:
who received the AI output;
who examined the alternatives;
who approved an exception;
who made the final judgment;
and who will answer for the consequences.
Principle Six
Institutionalize Post-Decision Review
Judgment sovereignty must be renewed through reflection.
After a consequential decision, organizations should ask:
Was the AI recommendation appropriate?
Was the human judgment appropriate?
Was the original purpose valid?
Which values were overlooked?
Did the decision impose an unjust burden?
Did the organization depend too heavily on the AI recommendation?
How should the standards be revised?
Reflection enables human beings to recover and strengthen the center of judgment.
Principle Seven
Do Not Eliminate Human Discomfort
In the AGI era, a rational-looking answer may be produced immediately.
The discomfort that remains within a person can nevertheless be an important resource for judgment.
Discomfort may indicate:
an omitted value;
an unseen human consequence;
an inappropriate assumption;
a violation of dignity;
or a difference between what is efficient and what is right.
Human discomfort should not automatically defeat evidence or reason.
Nor should it be dismissed merely because it cannot yet be quantified.
It should be brought back into the process as a question requiring examination.
13. Implications for Research and Governance
The Theory of Judgment Sovereignty in the Age of AGI has implications across several fields:
AI ethics;
philosophy of technology;
human-centered AI;
organizational governance;
management theory;
education;
economics;
political philosophy;
and civilizational studies.
It proposes that the challenge of AGI cannot be resolved by technical capability, safety engineering, or compliance alone.
Even a highly accurate, explainable, and well-controlled system does not determine:
which purposes humanity should pursue;
which values should take priority;
which judgments should remain human;
how responsibility should be distributed;
or what kind of civilization should be created.
The theory therefore shifts part of the inquiry from:
How should AI make decisions?
to:
Which judgments should AI be permitted to make, and under whose purposes and responsibility?
It also shifts the inquiry from:
How can a human remain in the process?
to:
How can a human remain the substantive author of the judgment?
These questions require continued research, institutional experimentation, and application across different cultural, legal, and organizational contexts.
The Principle of Judgment Classification is not intended as a final universal list.
It is a framework through which societies and organizations can consciously establish boundaries before those boundaries are silently determined by technological convenience.
Conclusion
So That Human Beings May Remain Human in the Age of AGI
The Theory of Judgment Sovereignty in the Age of AGI is not an argument against artificial intelligence.
It is an argument for placing artificial intelligence on the side of the human being.
Human beings do not need to compete with AGI in information volume, processing speed, memory, or analytical range.
As machines become more capable, the human role does not necessarily become smaller.
It moves toward more fundamental questions:
What should be our purpose?
What should we value?
What must we protect?
Which future should we choose?
What should never be sacrificed?
Which consequences are we prepared to bear?
AGI may support judgment.
It may deepen analysis, reveal risks, and broaden human understanding.
But it cannot live a human life in place of us.
It does not inhabit the human relationships affected by a decision.
It does not bear consequence through an embodied life of responsibility, regret, forgiveness, commitment, and hope.
Human beings must therefore retain the authority to establish purposes, choose values, make final judgments, and assume responsibility.
The defining challenge of the AGI era is not simply whether humanity can create intelligence more capable than itself.
It is whether humanity can benefit from that intelligence without surrendering the center from which human beings question, choose, and take responsibility.
Judgment sovereignty is the condition for achieving that aim.
It is also a foundational condition for building not merely an optimized civilization, but a Civilization of Dignity.
Note on This English Overview
This paper is an introductory English presentation of Mitsuaki Koga’s Theory of Judgment Sovereignty in the Age of AGI, a core treatise within his Civilizational Design Philosophy.
It is not a complete translation of the original Japanese treatise. It has been independently structured for researchers, business leaders, and other international readers interested in AI, human dignity, responsibility, governance, and the future of civilization.
The complete Japanese treatise has not yet been publicly released.
Suggested Citation
Koga, Mitsuaki. “Judgment Sovereignty in the Age of AGI: A Civilizational Design Framework for AI Delegation and Human Responsibility.” English Overview, Version 1.0, 2026.