The Machine Can Draft the Law. Can Humans Complete the Judgment?

What if the most revealing question about artificial intelligence is not whether it can think, but what human beings refuse to think about themselves?

A machine helped write a law in Brazil, and the immediate objection was not necessarily that the law was wrong. It was that its origin might make the law unacceptable. The response was pragmatic: it would be unfair to expose the population to the risk that a project would fail simply because artificial intelligence had written it.

That small episode opens a much larger question. When we judge an idea, do we judge its quality, its source, or the parts of ourselves involved in producing and accepting it?

The question becomes sharper when placed beside a psychological model that divides human consciousness into four basic functions: thinking, feeling, sensation, and intuition. A complete understanding of ourselves and the world requires all four. Yet individuals usually develop some functions more readily than others. The neglected function remains close to the unconscious, often appearing as the place where we repeatedly suffer, act foolishly, or “knock our head against the wall.”

This offers a powerful way to understand our relationship with artificial intelligence. We are not merely adopting a new tool. We are delegating certain mental functions to an external system while discovering that the functions we delegate most easily are not necessarily the functions we can afford to lose.

The real problem is not machine intelligence, but human incompleteness

Imagine a person whose dominant habit is analytical thinking. They are skilled at classification, consistency, abstraction, and logical comparison. Their strength is real, but it can become narrow. They may struggle to recognize emotional consequences, physical realities, or possibilities that cannot yet be proven.

Such a person does not become whole by intensifying analysis indefinitely. They become more capable by developing complementary functions. Feeling can reveal value and human significance. Sensation can anchor thought in concrete facts. Intuition can detect patterns and possibilities before they are fully articulated.

The same structure appears in institutions. A legal system is not only a machine for producing internally consistent text. It must also account for lived experience, measurable conditions, moral priorities, and future consequences. A law can be logically elegant and still be cruel. It can be compassionate in intention and still fail in practice. It can respond accurately to present facts while creating dangers that no one anticipated.

Artificial intelligence is extraordinarily attractive because it amplifies some functions with remarkable speed. It can compare large bodies of text, detect patterns, generate alternatives, summarize arguments, and produce language that sounds coherent. In the psychological vocabulary, it is especially useful in domains resembling analysis, memory, formal structure, and pattern recognition.

But usefulness in one function can create blindness in another. When a system produces a polished legal proposal, the visible artifact may conceal the invisible labor still required. Someone must determine whether the proposal describes reality accurately. Someone must ask whom it benefits, whom it burdens, and what values it encodes. Someone must imagine second order effects. Someone must decide whether the language is not merely defensible, but worthy of obedience.

A coherent document is not yet a legitimate decision. Legitimacy begins when a community takes responsibility for what the document will do in the world.

This is why the controversy around machine written law is psychologically revealing. The suspicion may appear to concern authorship, but underneath it lies a valid concern about missing functions. People sense that text generation is not the same as judgment. They worry that a document can be technically competent while lacking contact with human consequence.

The danger, however, is not solved by treating artificial intelligence as inherently illegitimate. Rejecting an idea solely because a machine helped produce it is itself a failure of judgment. It confuses origin with merit. A proposal should not be approved merely because a human wrote it, any more than it should be rejected merely because artificial intelligence assisted in writing it.

The better question is: Which functions were present in the process, and which were absent?

Four functions for testing any machine assisted decision

A practical framework emerges by treating thinking, feeling, sensation, and intuition as four tests that every important decision must pass.

Thinking asks whether it makes sense. Are the definitions clear? Do the provisions contradict one another? Does the reasoning follow from the evidence? Can the proposal withstand scrutiny from people who disagree with it?

Artificial intelligence can be a powerful partner here. It can identify inconsistencies, propose counterarguments, compare versions, and expose gaps that a tired human drafter might miss. In a legal context, this can increase quality rather than diminish it.

Feeling asks what matters to people. Who experiences the policy as protection, humiliation, opportunity, or threat? What forms of dignity are involved? Are certain groups being treated as abstractions rather than citizens with particular vulnerabilities?

This does not mean abandoning reason for sentiment. Feeling is not the same as emotional impulsiveness. It is the capacity to register value. A policy that ignores fear, trust, belonging, resentment, or hope may be rational on paper and unstable in practice.

Sensation asks what is actually happening. What do the relevant facts look like on the ground? Can the institutions responsible for enforcement carry out the proposal? What happens in a rural clinic, a crowded courtroom, a small business, or a household with limited internet access? Does the law work under real conditions, not merely in the imagined environment of its designers?

Sensation is the antidote to abstraction. It reminds us that every policy eventually becomes a queue, a form, a fee, a conversation with an official, or a decision made under pressure.

Intuition asks what the decision may become. What patterns are beginning to form? What unintended uses might emerge? Could a temporary exception become a permanent power? Could a system introduced for convenience alter the public’s expectations of privacy, accountability, or authority?

Intuition is not prophecy. It is disciplined sensitivity to possibility. It looks beyond the immediate output toward trajectories and consequences that are not yet fully visible.

A machine assisted law should therefore be evaluated through a four part process:

1. Test the reasoning and internal consistency.
2. Examine the human values and emotional consequences.
3. Verify the proposal against concrete conditions.
4. Explore plausible future effects and unintended uses.

The first test is often the easiest to automate. The other three demand participation from citizens, practitioners, affected communities, and institutions with different forms of knowledge.

Artificial intelligence may be our collective inferior function

The idea of an inferior psychological function adds another layer. The inferior function is not simply a weakness. It is a neglected capacity, one that can disrupt a person precisely because it has not been consciously developed. It is where recurring mistakes, disproportionate reactions, and humiliating blind spots often appear.

At a collective level, artificial intelligence may function as a mirror of this neglected area. Modern institutions have become highly skilled at formal reasoning, measurement, optimization, and administrative scale. They can calculate more quickly than ever. Yet they often struggle with meaning, trust, embodied experience, and long term responsibility.

Artificial intelligence arrives as an amplifier of the abilities institutions already prize. It can make bureaucracies faster, companies more efficient, and governments more productive. But speed can expose what efficiency has been hiding. If a flawed process becomes ten times faster, the flaw does not disappear. It spreads.

This is the paradox: the more competent artificial intelligence becomes at producing answers, the more urgently humans must develop the capacities that cannot be reduced to answer production.

A government might use artificial intelligence to draft a bill in minutes. But the real institutional work includes listening to people who will live under the bill, inspecting the places where it will be enforced, considering its moral symbolism, and imagining how future officials might misuse its ambiguities. If those activities were neglected before automation, automation can make the neglect harder to notice because the final document looks finished.

The recurring mistake is then predictable: people treat completion of the text as completion of the thought.

The psychological model suggests a better response. Before attempting to strengthen the inferior function directly, one develops auxiliary and complementary functions. In practical terms, this means that institutions should not respond to artificial intelligence by demanding only more machine output or by asking humans to exercise judgment in a vague, heroic way. They should build structured forms of complementarity.

A legal drafting team, for example, could include:

• A technical analyst who tests consistency and evidence.
• A community representative who identifies lived consequences.
• A frontline practitioner who understands implementation constraints.
• A strategic thinker who explores long term and unintended effects.
• A public decision maker who accepts responsibility for the final choice.

Artificial intelligence can participate in each stage, but it cannot substitute for the social distribution of responsibility. The point is not to preserve a ritual in which a human signs a machine produced document. The point is to ensure that the whole field of judgment is present.

The origin test is inferior to the responsibility test

There is a temptation to respond to machine involvement with a purity rule: human written good, machine written bad. But this rule fails for the same reason that judging a person solely by their dominant psychological habit fails. It mistakes a familiar source for a complete mind.

Human authorship has never guaranteed wisdom. Human beings produce propaganda, careless regulations, corrupt contracts, and brilliant reforms. Artificial intelligence does not erase human responsibility, but neither does human authorship prove that responsibility was exercised.

A more useful standard is the responsibility test. Ask four questions:

1. Was the proposal examined for accuracy and coherence?
2. Were the people affected by it given meaningful consideration?
3. Was it tested against practical reality?
4. Is there a clearly accountable person or institution willing to defend and revise it?

This standard changes the debate. Instead of asking whether a machine touched the text, we ask whether the decision making process was whole enough to deserve trust.

The distinction matters beyond law. Consider a hospital using artificial intelligence to prioritize patients, a school using it to recommend disciplinary action, or a company using it to screen job applicants. In each case, the system may improve pattern recognition while creating new risks of exclusion or dehumanization. The relevant question is not simply whether the model is accurate. It is whether the institution has developed the complementary functions required to interpret accuracy responsibly.

A screening model may correctly predict that an applicant resembles a group historically hired at lower rates. That prediction can be statistically accurate and socially unacceptable. Thinking has detected a pattern. Feeling must ask what dignity and fairness require. Sensation must examine the applicant’s actual circumstances. Intuition must consider how repeated use of the system will reshape the labor market.

No single function can carry the entire burden.

Key Takeaways

• Evaluate processes, not origins. Do not accept or reject an idea solely because it was produced by a person or assisted by artificial intelligence. Examine the quality of reasoning, evidence, values, implementation, and accountability.

• Use the four function test. For any important decision, ask: Does it make sense? What does it mean for people? Will it work under real conditions? What could it become over time?

• Find your personal blind spot. Notice where you repeatedly make avoidable mistakes or feel exposed and foolish. That pattern may reveal a neglected capacity that technology cannot develop for you.

• Design complementary teams. Pair analytical tools with people who bring moral awareness, practical experience, affected community knowledge, and long range imagination.

• Never confuse polished language with completed judgment. A finished document is only an object. Responsibility begins when someone considers its consequences and remains answerable for them.

The deepest lesson is not that artificial intelligence should replace human judgment, nor that humans should defend themselves by excluding machines. It is that powerful tools make partial minds more dangerous. When one function is amplified without the others, confidence grows faster than wisdom.

The law assisted by artificial intelligence therefore presents a challenge more serious than authorship. It asks whether a society can distinguish intelligence from judgment, efficiency from legitimacy, and textual completion from collective understanding.

Perhaps the future will not be decided by whether machines become more human. It will be decided by whether humans become more complete. The machine may draft the sentence, detect the pattern, or propose the structure. But only a sufficiently whole community can decide what deserves to govern a human life.