> ## Content Index
> Fetch the complete content index at: https://journal.shiroand.io/llms.txt
> Use this file to discover other available public pages before exploring further.

# The World Needs Imperfect Humans
- URL: https://journal.shiroand.io/the-world-needs-imperfect-humans/
- Published: 2026-10-13T20:22:49.000Z
- Updated: 2026-10-13T20:22:49.000Z
- Author: Kosuke Shirako
- Tags: Essays

*Why AI Ethics Cannot Replace Human Judgment*

**By Kosuke Shirako | SHIRO & Co.**

### The Problem With a Perfect World

Artificial intelligence is often discussed as a technology that will make the world more efficient, more predictable, and perhaps even more ethical.

We imagine systems that can eliminate human error, reduce bias, apply rules consistently, and make decisions without fatigue, anger, or personal prejudice.

It is an attractive vision.

But there is something unsettling about it.

What happens when we remove too much human imperfection from the world?

A society in which every action is evaluated, every decision optimized, and every deviation corrected may become remarkably efficient. It may even appear fairer than the society we inhabit today.

Yet it could also become a world in which human beings find it increasingly difficult to breathe.

The problem is not that artificial intelligence lacks intelligence.

The deeper problem is that intelligence and human existence are not the same thing.

And ethics, when reduced to a system of rules, may not be enough to bridge that distance.

### We Are Not Born Innocent

Recently, I have been thinking about a concept from Japanese Buddhism.

Shinran, the thirteenth-century Buddhist thinker associated with the development of Jōdo Shinshū, challenged conventional assumptions about goodness, morality, and salvation.

One of the most provocative ideas associated with his teaching appears in the *Tannishō*, a text compiled by a follower after Shinran's death.

It is commonly expressed through the paradox that if even a good person can attain birth in the Pure Land, then all the more so can an evil person.

Taken literally and without context, this may sound like a celebration of wrongdoing.

It is not.

The point is far more disturbing, and far more humane.

We are not autonomous moral agents who can simply choose to be good through perfect understanding and discipline.

Our actions emerge from conditions we did not create.

We are shaped by history, family, desire, fear, attachment, social structures, and circumstances beyond our control.

Buddhist thought gives us a language for this conditioned existence through the concept of *karma*, or *gō* in Japanese.

In this context, karma is not simply a system of rewards and punishments. It concerns the conditions through which our actions and their consequences arise.

We are beings who act without ever fully escaping the circumstances that shape our actions.

I find this view of humanity increasingly relevant to the age of artificial intelligence.

### The Limits of AI Ethics

Much of contemporary AI ethics is organized around principles such as fairness, accountability, transparency, safety, and human oversight.

These principles are necessary.

But they do not exhaust the problem of human judgment.

An AI system can be designed to apply consistent standards. It can document its reasoning, identify prohibited actions, and defer decisions when uncertainty exceeds a threshold.

These are meaningful achievements.

Yet even a well-governed system cannot fully represent what it means to be a human being making a decision.

Human decisions are rarely made from a position of moral clarity.

We decide while afraid.

We decide while grieving.

We decide while carrying memories that may be incomplete or contradictory.

We sometimes change our minds because the world has changed. Sometimes because we have changed. Sometimes for reasons we cannot entirely explain.

From the perspective of a formal decision system, this inconsistency may look like a defect.

From the perspective of human existence, it may be unavoidable.

And in some situations, it may be essential.

A person who forgives someone against every reasonable expectation, who revises an earlier judgment, or who refuses to abandon another person despite an unfavorable assessment is not necessarily acting irrationally.

They may be responding to dimensions of reality that the formal system was never designed to capture.

This does not mean that human decisions are always wiser, kinder, or more ethical than machine decisions.

Human beings can be cruel, prejudiced, irresponsible, and dangerously mistaken.

The point is not to idealize human imperfection.

It is to recognize that eliminating inconsistency is not the same as achieving wisdom.

### A World Too Clean to Breathe

There is a particular kind of uneasiness I associate with overly perfect environments.

In contemporary visual culture, the term *liminal space* often describes transitional environments that feel strangely empty or disconnected from ordinary life.

An abandoned shopping mall. A silent corridor. A room that seems familiar but no longer belongs to anyone.

Such places can feel unsettling precisely because the traces of everyday human activity have disappeared.

I sometimes imagine a society governed by perfectly optimized systems as a similar kind of space.

Everything functions.

Nothing is out of place.

Every process is measurable.

Every decision has a justification.

And yet something is missing.

Not efficiency.

Not intelligence.

Human presence.

Human communities are not made meaningful only by the absence of error.

They are also formed through hesitation, misunderstanding, repair, forgiveness, memory, and the possibility of beginning again.

A society that attempts to eliminate every ambiguity may also eliminate some of the conditions through which people recognize one another as human.

A perfectly administered world is not necessarily a livable world.

### Decision Is Not Authority

In my work at SHIRO & Co., I have been developing systems that explore the boundaries of AI-assisted decisions.

One principle has become especially important:

**Decision is not authority.**

An AI system may evaluate evidence, classify a situation, recommend a course of action, or identify an apparently optimal outcome.

But none of these capabilities automatically grants the authority to act.

This distinction is not simply a technical safeguard.

It is a philosophical boundary.

A decision-support system can help structure uncertainty. It can make assumptions visible, distinguish observations from inferences, and identify situations that require additional review.

But the ability to produce a recommendation should not be confused with legitimate permission to execute it.

Nor should human oversight become a ceremonial approval of conclusions that can no longer be meaningfully questioned.

For human judgment to matter, people must retain the practical ability to disagree, delay, challenge, revise, and sometimes refuse.

This is one reason I have become increasingly interested in the idea of reversibility.

What happens after a decision is made?

Can its consequences be reviewed?

Can an action be stopped?

Can a person return to a previous condition?

Can an institution acknowledge that its original judgment was wrong?

These questions concern more than the reliability of technology.

They concern the kind of society that technology makes possible.

### The Moral Importance of Being Able to Change

The modern technological imagination often associates progress with the elimination of friction.

Faster decisions.

Fewer exceptions.

More automation.

Less uncertainty.

But friction is not always a failure.

Sometimes, friction is the space in which judgment occurs.

A delay may prevent an irreversible mistake.

A disagreement may reveal an assumption that nobody thought to question.

An exception may expose the limitations of a rule.

And a person who changes their mind may be demonstrating an ability to respond to the world rather than merely follow a predefined procedure.

This is particularly important when AI systems become deeply embedded in institutions.

The danger is not necessarily that machines will become malicious.

It is that institutions may become so confident in machine-supported processes that they gradually lose the capacity to question them.

A recommendation becomes a standard.

A standard becomes a requirement.

A requirement becomes an action.

And eventually, no one can clearly identify the moment at which advice became authority.

This is not merely an alignment problem.

It is a problem of institutional responsibility.

### Beyond the Good Human and the Bad Human

Shinran's thought does not offer a ready-made framework for AI governance.

It would be a mistake to pretend otherwise.

His concern was religious and existential, rooted in questions of human limitation, reliance, and liberation rather than technological decision-making.

But his understanding of human beings challenges an assumption that sometimes appears beneath modern efforts to engineer ethical systems.

The assumption is that a sufficiently refined framework of knowledge and rules could allow us to identify the good person, the correct action, and the ideal outcome with increasing certainty.

Shinran's perspective invites a different question.

What if human beings cannot fully stand outside their own conditions to judge themselves or others from a position of purity?

What if every human judgment remains entangled with the world that produces it?

This does not mean abandoning moral distinctions.

It does not mean treating harmful and compassionate actions as equivalent.

It means recognizing that human beings cannot be reduced to fixed moral classifications.

A person is more than the worst action they have taken.

A person is also more than the favorable score a system assigns them.

The possibility of transformation matters.

So does the possibility of being misunderstood.

And so does the possibility that our categories are incomplete.

### Why We Still Need Human Judgment

I do not believe the future of AI should be built around the assumption that human beings are morally superior to machines.

That would be too simple.

Human judgment is inconsistent, vulnerable, and often deeply flawed.

But precisely because human beings inhabit a world of relationships, histories, responsibilities, and consequences, their role cannot be understood merely as a source of error to be eliminated.

We need institutions that can make defensible decisions.

We need AI systems that are reliable, accountable, and safe.

And we need human beings who can intervene when established procedures no longer adequately represent the situation before them.

Human judgment is not valuable because it is always correct.

It is valuable, in part, because human beings can be held responsible, can recognize their limitations, and can participate in the difficult process of revising what they once believed.

Of course, humans do not always exercise these capacities.

That is why human involvement must be substantive rather than symbolic, and why institutions need mechanisms for accountability, appeal, and correction.

The challenge is not to preserve arbitrary human discretion.

It is to preserve meaningful human responsibility.

### The World Needs Imperfect Humans

Perhaps the central question of AI governance is not how to create a world in which every decision is correct.

Perhaps it is how to create a world in which decisions can remain open to human responsibility, correction, and change.

A world in which a person can question an automated conclusion.

A world in which uncertainty is not automatically treated as failure.

A world in which disagreement does not always need to be resolved immediately.

A world in which people are not permanently defined by an assessment, a classification, or a past mistake.

This may seem less efficient than a fully optimized society.

It probably is.

But efficiency is not the only measure of a civilization.

We are not merely decision-making units.

We are conditioned, contradictory, unfinished beings who live among other conditioned, contradictory, unfinished beings.

And perhaps that is precisely why the world needs us.

Not because we are perfect.

But because we are not.

---

*Kosuke Shirako is the founder of SHIRO & Co., an independent research and development company exploring AI decision boundaries, technological governance, and the relationship between human judgment and intelligent systems.*

*Explore:* [*SHIRO & Co.*](https://www.shiroand.io/?ref=journal.shiroand.io) *·* [*Decision Boundary*](https://proposal.shiroand.io/decision-boundary?ref=journal.shiroand.io) *·* [*LIFE*](https://book.shiroand.io/life?ref=journal.shiroand.io)