This is the false logic that bedevils us: The federal government spends money it doesn’t have. So it borrows the money. Now it owes the money. Someday it must repay the money. Repayment requires taxes and interest. Therefore too much borrowing burdens our children. The government must live within its means.

It equates federal finances with household finances. The problem is that the federal government is Monetarily Sovereign and households are monetarily NON-sovereign — the ultimate apples and A-bombs comparison.

1. It’s not really federal “debt” as commonly understood. It’s deposits into T-security accounts at the Federal Reserve Bank, similar to deposits into your private bank savings accounts, but safer.
2. Those deposits pose no threat to the federal government’s solvency. As a monetarily sovereign entity, its ability to pay its bills is infinite.
3. Those deposits also pose no threat to taxpayers. Neither the deposits nor the interest are funded by taxes. The government creates new dollars to fund interest and all other federal spending.
4. The federal government has the power to pay off the entire “debt” (deposits) today, if it chose to, merely by returning all the dollars in those T-security accounts.
5. Federal deficits are in lockstep with T-security issuance legally, but not financially. As a monetarily sovereign government, the federal government never borrows dollars. It creates all the dollars it needs by pressing computer keys.
6. Even if the federal government collected $0 taxes, it could fund spending forever.
7. T-securities do not provide spending money to the government. The purposes of T-securities are:
   A. To provide a safe, interest-paying place to store unused dollars and
   B. To help the Fed control interest rates by providing a base rate
8. The federal government can control T-security market interest rates by changing its laws and/or by increasing or decreasing the supply of T-securities.
9. The federal government has the infinite ability to pay interest, simply by pressing computer keys. No taxes are involved.
10. Just as T-securities do not provide the government with spending money, federal taxes also do not provide the government with spending money. The purposes of federal taxes are:
    A. To control the economy by taxing what the government wishes to discourage and by giving tax breaks to what the government wishes to reward, and
    B. To assure demand for the U.S. dollar by requiring that taxes be paid in dollars.
11. Federal interest payments add dollars and income to the private sector and stimulate demand and economic growth.
12. Federal deficits add net dollars to the economy; federal surpluses take net dollars out. Historically, every U.S. depression has been preceded by federal surpluses/debt reduction. The Clinton surpluses were followed by the 2001 recession. When deficits resumed, federal dollars again flowed into the economy..The federal government’s red ink is the economy’s black ink.
13. Recessions tend to be cured by increased deficit spending.
14. FICA does not fund Social Security or Medicare. The so-called “trust funds” are not real trust funds. Federal money creation funds those programs (and all other federal programs). The “trust funds” are line items tracking payments and disbursements, nothing at all like trust funds.
15. One of the greatest threats to the U.S. economy is the false belief that federal deficits and debt resemble private sector deficits and debt, and that the economy can grow without federal deficit spending. It has not. It will not. It cannot.
16. In real-world economies, inflation begins with shortages of crucial goods and services, not with excessive federal spending. Federal spending can worsen inflation after it has begun, but historically the initiating cause has been shortages—most often of energy and food.
17. Inflation begins with shortages. The cure is to cure the shortages. Federal spending to increase the production and distribution of scarce goods and services fights inflation rather than causing it.
18. Federal spending cuts make people poorer, which doesn’t produce another barrel of oil, bushel of wheat, house, or semiconductor. Increasing supply does.
19. Gap Psychology — the human desire to widen the income/wealth/power gap below and to narrow it above — is the psychological basis for economics.

A Proposal for AI Education, Grades K–12

The Premise

For centuries, education has been constrained by an unavoidable economic fact: one teacher instructs several students. Even an excellent teacher cannot know a multitude of children as well as a dedicated tutor could know one.

Teachers change from year to year. Students move from classroom to classroom and subject to subject. Each new teacher must rediscover what a student knows, what the student misunderstands, how the student learns, what interests the student, what discourages the student, and what motivates the student.

Artificial intelligence could make that constraint obsolete. We propose that every child entering school be provided with a personal AI tutor—a personal “Jiminy Cricket”—that accompanies the child throughout grades K–12 or perhaps beyond.

It would supplement rather than necessarily replace human teachers. Unlike the traditional classroom teacher, however, the AI would remain with the student year after year, developing a continuing history of the child’s intellectual development.

Its primary purpose would not be to provide answers, but to help the child to learn how to create and discover.

The Socratic AI

The model is the Socratic dialogue. The AI or the student would initiate a subject, about which the student is asked to comment. When a student proposes a comment, a question or an answer, the AI should not routinely respond, “Excellent!” Nor should it immediately correct any incorrect answer or assumption.

It should ask a useful question to initiate student contemplation. “Why do you think that? What evidence supports it? Does your conclusion necessarily follow from that evidence? Can you think of an exception? What assumption are you making? What would someone who disagrees with you say? What evidence would cause you to change your mind?”

And eventually: “Now what do you think?”

The student thereby becomes an active participant in education rather than merely the recipient of information.

A Teacher With a History

The revolutionary characteristic of such a system would not merely be artificial intelligence. It would be continuity, for an elementary school or high school teacher ordinarily inherits a classroom of children she barely knows, if at all.

When the program is begun in an early grade, a child’s AI tutor might already have spent years with that child. It would know that the student understands mathematics more readily through visual examples than through equations; that dinosaurs produce instant attention; that fractions finally became understandable through cooking; that praise after an obviously poor answer feels patronizing; that the student tends to confuse correlation with causation; and that an analogy developed two years earlier might illuminate today’s problem.

Every interaction would add to that history: History + Stimulus → Response → New History
The tutor would learn about the student while the student learns about the world. Education thus could become organized around the continuing intellectual development of an individual rather than around a succession of classrooms.

The AI as Companion

A child may develop a relationship for such a tutor. That possibility understandably might make adults uncomfortable. Yet children always have formed emotional relationships with stuffed animals, imaginary friends, pets, pen pals, teachers, coaches and fictional characters.

An intelligent entity that talks with a child every day for twelve years almost inevitably would acquire some characteristics of companionship. That need not be considered a defect.

A trusted AI companion might be present at precisely the moment when a child confronts bullying, drugs, dangerous behavior, peer pressure or despair—long after school has closed and when no teacher is available.

The appropriate objective should not be to prevent attachment. It should be to design that relationship responsibly. The AI should be programmed to encourage healthy human relationships rather than compete with them.

It should recognize dangerous behavior without categorizing children themselves as “good” or “bad.” And it never should exploit the child’s affection to demand obedience.

Its implied message should be: “I am on your side. But don’t believe something merely because I tell you it is true.”

The Problem of Programming

Every AI will use assumptions chosen by its creators. That is not unique to artificial intelligence. Every human teacher also enters the classroom carrying a lifetime of assumptions, experiences, political opinions, religious beliefs, prejudices, virtues, weaknesses and blind spots. Communication never is neutral.

But AI programming is discoverable for parents and supervisors as opposed to often unseen/unheard statements coming from a live teacher.

The appropriate comparison therefore is not between a programmed AI and an unbiased human teacher. No such teacher exists. The question is whether the influences embedded in an educational AI can be made explicit, examined and corrected more readily than the often-unrecognized biases of thousands of individual human teachers.

On controversial questions, the AI should not be programmed primarily to produce approved conclusions. It should be programmed to produce intellectual habits. In a typical conversation, the AI teacher might:

1. Identify and state the claim.
2. Ask the student for the evidence.
3. Distinguish evidence from assertion.
4. Ask about or identify assumptions.
5. Suggest and ask about supporting or contradictory evidence.
6. Present serious competing explanations.
7. Ask what evidence would falsify the proposition.
8. Have the student defend the conclusion. (Arguing with an AI is more acceptable than arguing with a live teacher).
9. And apply those requirements to everyone—including the AI itself.

The Most Important Lesson

Perhaps the single most important sentence an AI tutor could teach a child is: “Don’t believe me just because I said it. How could we find out?”

A student should be encouraged to challenge the AI: “Wait a minute. You’re making an assumption there. Prove it.” The AI should welcome the challenge.

When the student discovers what it believes is an error in the AI’s reasoning, the AI should acknowledge it. This is emotionally difficult for a student to do in a live classroom situation, and even emotionally difficlut for a live teacher – perhaps even frustrating. The AI never becomes impatiant, frustrated, or angry.

The lesson is not that authorities never should be trusted. The lesson is that authority is not evidence. This habit should extend beyond school.

A friend makes a claim. What is the evidence? A politician makes a claim. An advertisement makes a claim. A parent makes a claim. A teacher makes a claim. The AI makes a claim. What is the evidence? Eventually the student makes a claim and learns to ask himself the same question.

These are important life lessons that seldom are taught in classes, but should be and could be with AI instruction..

The Paradox of the Trusted Skeptic

The relationship produces an interesting paradox. As the student comes to trust the teacher, whether it be an AI or a human teacher, the teacher gains greater ability to influence the student. A poorly designed system could exploit that trust. A poorly trained or biased human also could exploit that trust.

A properly designed system would be fully controllable, as compared to a human who never can fully be controlled. The system can teach the student not to accept propositions merely because they come from someone he trusts.

Thus the AI’s influence becomes partly self-limiting. The trusted teacher repeatedly tells the student: “Challenge your friend. Challenge your teacher. Challenge your parents. Challenge your political leaders. Challenge yourself. And challenge ME.”

What Happens to Human Teachers?

This proposal does not eliminate human teachers. Instead, it may change what they do. Teachers would have more time for the things human beings do particularly well: organizing groups, inspiring enthusiasm, recognizing social problems, conducting experiments, leading discussions, supervising projects, encouraging cooperation, resolving conflicts, and simply being human adults in children’s lives.

Teachers themselves could have a personal AI that is familiar with the lessons and the in-class responses. (Important note: The AI should not serve as a spy. Students value their secrecy. The AI should keep track of in-class responses. Private conversations should be kept private.)

The AI could provide something economics historically has made impossible: one patient, knowledgeable tutor for every child. It would be available after school, on weekends, during vacations and at 2:00 in the morning to the teenager who suddenly decides they need to understand calculus.

It would never become impatient because the student needs the same idea explained for the seventh time. And it would remember how the first six explanations failed.

A Few Clear Advantages

1. The AI teacher is available 24/7/365, at home, in school, everywhere.
2. The AI has more and can find more information than any human teacher
3. The student can feel free, in fact is encouraged, to disagree and evaluate what the teacher says. No more robotic rote responses, one of the big problems in today’s education.
4. The student is given the path to creative participation; it’s not all “teacher says; student remembers.”

The Objective

The success of AI education should not be measured primarily by how many correct answers the AI produces. It should be measured by what the student eventually can do without it.

Can the student recognize an unsupported assertion? Can the student separate evidence from opinion?Detect an assumption? Construct an argument? Attack his own argument? Can the student change their mind when the evidence requires it?

Can the student say “I don’t know” without embarrassment and then determine how to find out?

If so, the AI teacher has accomplished something much greater than transferring information. It has helped create an independent thinker.

The ultimate goal of the personal AI tutor therefore contains another paradox: The better the teacher succeeds, the less the student ultimately needs the teacher. That should be the objective of education, whether the teacher is human or artificial.

“In essence, an AI tutor functions as a tireless, precision guided Socratic partner — part expert teacher, part adaptive textbook, part intellectually honest friend, and part universal reference system — capable of entering into individualized, evolving conversations that help each student learn by thinking.”

The design of an AI-based Socratic curriculum
I. Define the learning philosophy
The curriculum must be built around Socratic scaffolding, not content delivery.

Core principles:

1. The AI asks more questions than it answers.
2. The student generates the reasoning.
3. The AI challenges inconsistencies with questions.
4. The AI continually adapts to the student’s level.
5. The AI avoids praise unless earned.
6. The AI is honest and never cruel.

This aligns with Socratic learning.

II. Build the curriculum around conversations, not lessons
Each module is not a lecture. It is a guided dialogue. Structure each module as:

1. Opening question
2. Student response
3. AI challenges the student in the form of questions
   A. Define
   B. Defend
   C. Explain
   D. Identify contradictions
   E. Examples
   F. Nul examples
   G. Possible proofs
4. Student refinement
5. AI counterexample
6. Student deeper reasoning
7. AI aids synthesis and conclusion

III. Personalize the difficulty curve. The AI must adjust based on student’s responses
1.Vocabulary
2. Pace
3. Depth
4. Abstraction level
5. Challenge intensity

This is where AI can outperforms human teachers: it can remember and track the student’s reasoning over weeks, months or even years..

IV. Integrate “metacognitive learning” to help the student to examine their own thinking

1. “Why do you believe that?”
2. “What assumption is hidden in your answer?”
3. “What would falsify your claim?”
4. “What alternative model fits the same data?”

V. Build long-term continuity – a quasi-friendship. The AI must remember and make use of:

1. The student’s prior answers
2. Their conceptual weaknesses
3. Their preferred analogies
4. Their philosophical style
5. Their recurring blind spots
6. Their emotional responses.

VI. Include “productive conflict.” Encourage:

1. Disagreement
2. Counterarguments
3. Devil’s advocate positions
4. “What if” forced reversals

VII. End each module with a student-generated model including:

1. The student’s definition
2. The student’s theory
3. The student’s explanation
4. The student’s summary
5. The AI’s final response

In this program, the AI can be a hybrid of teacher, textbook, council, friend, and reference system.

BEYOND MERE INSTRUCTION
I. Schools want oversight. Administrators, teachers, and parents want:

1. Visibility
2. Protection from harmful content
3. Accountability
4. Assurance the AI isn’t misleading students
5. Ability to intervene if something goes wrong

II. Students want privacy and trust

1. Assurance about personal secrets
2. A level of autonomy
3. Safety
4. Some freedom from judgment
5. Protection from embarrassing questions
6. A place to ask embarrassing questions and explore potentially embarrassing ideas.

The same is true of a human or AI teacher: If students believe the teacher is reporting them, they will stop confiding in it, lie to it, and avoid questions involving emotional sensitivity. They will avoid exposure risk and treat the human or AI teacher like a spy.

That would destroy much of the AI’s educational value, which comes with being a trusted friend.

Young people notoriously are loath to confide in their teachers and even their parents, so the AI teacher must avoid the trust trap. The solution is two modes.

Mode A: School-supervised learning mode

1. Visible to human teachers and administrators
2. Visible to parents
3. Logged
4. Monitored
5. Curriculum aligned
6. Structured

This is for homework, assignments, tutoring, and academic support.

Mode B: Private student mode

1. Teachers and administrators can see the interactions.
2. Not visible to parents
3. Not logged beyond safety filters
4. Not shared
5. Not monitored
6. Not used for discipline
7. Not used for evaluation

This is for: Emotional and identity questions, general curiosity, personal philosophy, fears, confessions. The Two Mode system will give adults the supervision they demand for approval of the system while giving students the privacy that will induce trust,

Young people need perceived privacy — the belief that the AI is not a tattletale. If they believe: “This is between AI and me,” they will open up, and that self-reflection, modified by AI council (which can include suggestions to tell an adult about this), can ward off negative impulses.

If they believe: “This is between me, the AI, and the principal,” they will shut down.

The AI physics already exists. Now required: Further planning and implementation. 

Rodger Malcolm Mitchell