FEDERALIST NO. 10, REFRACTED

THE FULL COLOR PAPERS · PAPER NO. 01

What Happens to Faction When Reality Itself Becomes Contested?

On Factual Truth, Political Power, Trust, Synthetic Representation, and the Conditions of Human Judgment

Kathleen C. Thompson, J.D.

Independent Researcher and Attorney · Louisville, Kentucky

Thinking in Full Color

---

Introduction

To the People in the State of Judgment, Attention, and Uncertainty

Among the promises of artificial intelligence—the expansion of human capacity, the acceleration of learning, and the possibility that one person may investigate questions once reserved for entire institutions—none deserves closer examination than its power to change what we understand by representation.

In November 1787, James Madison, writing as Publius, addressed the people of New York in support of an unratified Constitution.

His republic was an argument before it was an institution.

Neither its success nor its justice could be established by the eloquence of its advocates.

Madison asked how a people divided by interests and passions might govern themselves without surrendering liberty to an oppressive faction. He proposed not the removal of disagreement, but institutions capable of controlling its effects.

We face a related problem under profoundly different conditions.

Machines trained on vast quantities of human expression can help us formulate arguments, criticize them, summarize disagreement, and imagine alternatives. They can also produce the appearance of voices that have never spoken, agreement that no actual community has reached, and institutional recognition that no institution has conferred.

A room can look crowded before anyone arrives.

But there is a second difficulty.

The room may already have agreed upon something that never happened.

And the people who eventually arrive may be quite real.

Human beings have always used language to represent and remake their understanding of the world. We remember, interpret, persuade, accuse, testify, imagine, and predict. We tell stories about what happened and construct possibilities about what could happen next.

Sometimes those stories are true. Sometimes they are mistaken. Sometimes they are deliberate fabrications. Sometimes they concern futures that do not yet exist but might, through human choice and collective action, become real.

These are not equivalent acts.

A novelist may invent a republic. A political leader may misrepresent an election. A citizen may imagine a better institution. A court, exercising lawful authority, may enter a judgment that changes the legal rights of actual people.

All four acts employ language.

They do not possess the same relationship to factual truth, legitimacy, or power.

Yet all four may produce consequences.

Consider what happens when an actual political faction organizes around a false factual account.

Donald Trump's repeated assertions that the 2020 presidential election was stolen did not make those assertions true. Official results, recounts, audits, and other examinations established an outcome contrary to his claims.

But the people who believed him were real.

Their convictions, political participation, demands upon institutions, and resulting actions were real.

A false account had entered the world as a consequential political force without becoming a true account of the event it described.

Artificial intelligence was not required.

That distinction is essential.

The present danger did not begin when machines learned to generate convincing language. Human beings have long invented evidence, spread propaganda, fabricated consensus, and organized political movements around contested or false accounts of reality.

The causes of faction remain human.

What is changing are the conditions under which representations can be produced, circulated, personalized, evaluated, and mistaken for something they are not.

A single interested actor can generate thousands of apparent voices. A machine can simulate an argument among experts who have never met. Another machine can assess the simulated discussion. A third can summarize the supposed agreement.

An organization can receive an impressive record of consultation without consulting the people whose interests are supposedly represented.

A person can encounter repeated statements of agreement without recognizing that each originated from the same source.

And an authentic recording, document, or testimony may be dismissed as fabricated precisely because fabrication has become so plausible.

We face the possibility of believing what is false while becoming increasingly capable of disbelieving what is true.

The consequences for constitutional government are not confined to misinformation.

They reach the foundations of trust.

A republic cannot require every citizen to believe the same things. Indeed, its legitimacy depends in part upon the freedom to question authority, challenge evidence, and organize opposition.

Nor should distrust automatically be treated as a defect in the citizen. Institutions have earned skepticism through error, exclusion, secrecy, and sometimes misconduct. The American constitutional experiment itself began in a society that denied meaningful political participation to many of the people whose lives its institutions governed.

The solution cannot simply be to demand more trust.

The question is how trust can be warranted, challenged, lost, and earned again.

And trust requires attention.

Citizens must have some meaningful opportunity to examine the claims made in their name. They must be able to distinguish evidence from allegation, independent agreement from repetition, actual participation from simulation, and institutional authority from its polished imitation.

Attention, however, is finite.

The pace of representation is not.

When language can be produced and revised at machine speed, the burden of examining it may grow faster than the capacity of people and institutions to do so.

The question is not merely whether a statement can be verified.

It is whether anyone has the time, means, independence, and authority to challenge it before consequential action follows.

This problem has two sides.

Artificial intelligence can make inquiry more accessible. It can help citizens articulate objections, explore unfamiliar evidence, imagine institutional alternatives, and investigate questions previously beyond their practical reach.

It can also make unverified assertions appear established, simulated constituencies appear authentic, and repeated representations appear independently corroborated.

We must preserve the first possibility without surrendering judgment to the second.

That requires distinctions.

A belief is not proof.

A prediction is not an outcome.

A simulation is not participation.

A generated consensus is not consent.

An imagined institution is not an authorized institution.

And a machine's confident assessment of an idea is not evidence that the idea has survived independent examination.

These distinctions are not abstractions reserved for philosophers.

They determine what enters public records, what counts as evidence, which people are heard, which claims receive institutional recognition, and which imagined futures acquire the power to govern actual lives.

They also apply to the present inquiry.

In the course of developing the Full Color Futures research program, I used artificial intelligence to generate analyses, criticism, research questions, imagined conversations, and professional documents describing the project itself.

One such document presented a proposed research architecture in the language of an established, reproducible framework. Later generated assessments praised its significance with confidence exceeding the evidence available.

The documents were real artifacts of human–machine interaction.

The institutional validation their language appeared to imply was not established.

The machine had generated a representation of recognition.

Recognition would require someone beyond the representation.

I offer that record not as proof of a universal mechanism, but as a specimen worthy of examination—and as a reminder that this argument cannot exempt its own author from the standards it proposes.

I am a trial lawyer. I am also a person who has never seen color. I have spent much of my life relying upon language to understand aspects of a world I cannot directly perceive.

I know that more light does not always mean more sight.

The conditions under which information reaches us matter.

The same principle deserves investigation in an age when human beings can encounter more representations than ever before, often without clearer knowledge of their origins, limitations, or proper use.

Madison understood that suppressing liberty to eliminate faction would destroy the very object republican government was meant to protect.

We should approach artificial intelligence with similar caution.

The answer cannot be to abolish imagination, prohibit dissent, or declare that only unaided human language is authentic. Nor can it be to let every sufficiently convincing representation acquire the standing of evidence, public opinion, or institutional judgment.

The task is to develop and test better ways of distinguishing what happened, what is believed, what is merely possible, who has actually participated, and what actions are justified.

I propose examining a candidate practice that I call the Refraction Point: the moment when continued generation is no longer necessarily the competent next step, and the responsible response may instead be to verify, observe, ask, challenge, route, hold, act, or stop.

I also propose examining epistemic due process as an organizing idea for the conditions under which consequential representations may be examined, challenged, and corrected. It is not offered here as a newly established legal doctrine or a validated technological remedy.

Both proposals require independent evaluation. Simpler methods may work better. Existing institutions may already possess safeguards that deserve improvement rather than replacement.

The purpose of this paper is not to announce that a solution has been found.

It is to identify a constitutional and epistemic problem that demands more precise investigation.

Madison asked how a republic could preserve liberty while controlling the effects of faction.

We must now ask:

How can a constitutional republic preserve legitimate disagreement when the factual evidence, apparent participation, and supposed agreement upon which public judgment depends can be manufactured, amplified, or credibly disputed?

How do we protect the right to imagine a different world without confusing imagination with its existence?

How do we permit citizens to question institutions without destroying the possibility of establishing facts?

How do we prevent the appearance of participation from substituting for the people themselves?

And how might artificial intelligence help us examine more possibilities while preserving the conditions necessary to judge among them?

We need institutions that can imagine faster without believing faster.

We need procedures through which a representation may be challenged before it becomes consequential—and through which justified objections can change what happens next.

A republic can look unanimous before anyone has been heard.

A room can look crowded before anyone arrives.

The task is not to close the room.

It is to preserve the conditions under which someone may enter, contradict the record, and change the result.

---

Table of Contents

PART I — THE HUMAN ORIGINS OF FACTION AND REALITY-MAKING

Before the machine, there was the human capacity to imagine, believe, persuade, and organize.

I. Of Faction, and of the People Who Are There — Madison's problem and the incomplete republic

• The original Federalist No. 10: liberty, interest, passion, and the danger of faction.
• The extended republic, authentic diversity, and the promise and failure of representative filtering.
• The founding republic's exclusions: whose interests counted and whose voices were missing.
• Why visible multiplicity is not the same as independent participation.
• The new problem: a thousand apparent speakers may originate from one interested actor.

II. Of the Worlds Human Language Makes — Imagination, memory, persuasion, and the possibility of changing reality

• Human imagination and episodic future simulation.
• Language as description, interpretation, persuasion, and action.
• Fiction, propaganda, political myth, and collective possibility.
• Speech acts, institutional facts, and the creation of legitimate social arrangements.
• Why the capacity to imagine something before it exists is essential to political freedom.

III. Of What Is Real, Believed, Possible, and Authorized — Four distinctions upon which the argument depends

• What happened? The status of factual claims and their supporting evidence.
• What is believed? Individual judgment, collective conviction, uncertainty, and error.
• What might happen? Hypotheses, predictions, simulations, and possible futures.
• Who may act? Participation, consent, institutional authority, and implementation.
• How one kind of representation becomes improperly promoted into another.

PART II — THE POLITICS OF FACTUAL TRUTH AND DISTRUST

The republic encounters the consequences of falsehood before artificial intelligence enters the case.

IV. Of Falsehood and Political Power — Donald Trump, the 2020 election, and the factual premises of political action

• The documented election result and the evidentiary record.
• The repetition of false electoral claims despite contrary evidence.
• Distinguishing factual falsity, knowledge, intent, persuasion, and political responsibility.
• How allegiance can influence which sources citizens accept as credible.
• Why neither popular belief nor political power changes an established electoral result.

V. Of the Consequences of a False Account — When fabricated narratives produce real-world consequences

• The Springfield, Ohio, allegations concerning Haitian immigrants in 2024.
• The spread of unsupported factual claims through political and media channels.
• Real reactions, institutional responses, public fear, and the limits of causal inference.
• Why the falsity of an account does not make its consequences imaginary.
• The reverse danger: authentic evidence dismissed as fabrication—the liar's dividend.

PART III — THE REPRESENTATION MACHINE

What changes when language, apparent participation, and evaluation operate at machine scale?

VI. The Representation Machine — The Five Powers of consequential representation

• Framing Power: who determines which questions are asked.
• Generative Power: who can produce arguments, alternatives, identities, and accounts.
• Selection Power: which representations survive and which disappear.
• Evaluative Power: who determines credibility, importance, and sufficiency.
• Implementation Power: what enters records, policies, institutions, and lives.
• Why control over the entire process may matter more than the number of generated perspectives.

VII. The Three Letters and the Three Headlines — Two tests of authenticity: who spoke, and what happened?

• An authentic constituent using AI to communicate.
• An authorized institution employing AI-assisted advocacy.
• A fabricated constituent presented as an actual participant.
• Authentic reporting, misleading context, and entirely fabricated events.
• Why machine-assisted language is not inherently counterfeit participation.
• Why content authenticity and speaker authenticity must be evaluated separately.

VIII. The Generated We — Synthetic plurality and the counterfeit constituency

• The FCC net-neutrality comment scandal as a pre-generative-AI precedent.
• AI-generated constituent correspondence and legislative attention.
• Synthetic surveys and their limitations as substitutes for actual public opinion.
• Simulated stakeholder consultations, manufactured agreement, and concentrated framing power.
• The difference between predicting a community's views and giving that community a voice.
• The constitutional problem of apparent representation without independent participation.

IX. Of Faction Without Common Facts — Trust, attention, epistemic authority, and the politics of disbelief

• Institutional distrust and the distinction between warranted skepticism and generalized suspicion.
• Why genuine disagreement need not mean that factual evidence is indeterminate.
• Deepfakes, misinformation allegations, and the liar's dividend.
• Repetition, social belonging, perceived consensus, and competing accounts of reality.
• Attention as a limited human and institutional resource.
• What happens when accepting or rejecting evidence becomes a test of factional loyalty?
• The need to earn trust through accountable procedures rather than demand unquestioning belief.

X. The Record That Appears to Corroborate Itself — When language evaluates language without encountering independent evidence

• Repetition-induced belief and perceived consensus.
• Apparent source diversity that traces back to a single unsupported assertion.
• AI-generated analysis, automated criticism, synthetic agreement, and recursive evaluation.
• How earlier generated outputs can influence the next human question.
• Distinguishing a useful hypothesis from a verified finding.
• Why more reviews do not necessarily mean more evidence.
• The point at which confidence increases without a corresponding increase in independent support.

XI. The Machine's White Paper — An exhibit from the inquiry itself

• An AI-generated research document describing a proposed architecture in institutional language.
• The promotion of candidate methods into apparently established accomplishments.
• Subsequent generated praise and the appearance of independent recognition.
• What the artifact demonstrates—and what it cannot establish.
• Model sycophancy, evaluation bias, prompt provenance, and unearned confidence.
• Applying the paper's own evidentiary standards to the documents that helped create it.

PART IV — THE CONDITIONS OF ACCOUNTABLE JUDGMENT

The purpose is not to eliminate imagination or disagreement, but to preserve the possibility of correction.

XII. A Remedy That May Itself Be Questioned — Epistemic due process and the proposed Refraction Point

• Existing legal and administrative safeguards: provenance, procedural fairness, participation, and review.
• Four Refraction Point questions: Status, Foundation, Purpose, and Disposition.
• Distinguishing verification from consent, institutional authority, and public legitimacy.
• Designing comparative tests against ordinary AI assistance and simpler checklists.
• Measuring calibration, source independence, justified revision, accessibility, time, and burden.
• Preventing safeguards from becoming censorship, procedural theater, or barriers to authentic anonymous expression.
• Why a proposed remedy must permit independent criticism and the possibility of failure.

XIII. Of the Republic That Does Not Yet Exist — The constructive power of imagination and the conditions for bringing possibilities into the world

• The republic as an imagined arrangement before institutional realization.
• The constructive role of AI in expanding inquiry, foresight, and available alternatives.
• Possibility generation versus judgment about feasibility, legitimacy, and consequences.
• The proposed Possibility Ledger and the provisional concept of judgment debt.
• Participatory futures: the people affected must have opportunities to influence the future being proposed.
• What must happen before a generated possibility becomes an authorized public commitment.
• Building institutions that can imagine faster without believing faster.

XIV. The World's Right to Answer Back — Returning judgment to evidence, actual people, and consequential reality

• Madison's problem reconsidered under conditions of synthetic representation.
• The distinction between a generated constituency and an actual public.
• The independent witness, the critic, the missing person, and the objection that changes the decision.
• World Return: actual sources, tests, participants, observed consequences, corrections, and refusals.
• The author's responsibility to submit her own claims to independent evaluation.
• The incompleteness of every proposed remedy.
• Preserving the right of the world to contradict the representation.

---

The Empty Chair

The loop is not the enemy. The closed loop is.

Enter an appearance.

---

Source Apparatus and Research Record

Notes and Authorities

Madison and constitutional history; political truth and philosophy of language; cognitive science; social trust; election records; synthetic representation; survey research; administrative law; human–AI interaction; and automated evaluation.

Research Method and Status Disclosure

Distinguishing established scholarship, documented cases, conceptual proposals, individual interaction specimens, and claims requiring independent testing.

Illustrative Plates and Exhibits

The Three Letters; The Five Powers; The Three Headlines; The Generated We; The Record That Cites Itself; The Machine's White Paper; The Refraction Point; The Empty Chair.

Invitation to Enter an Appearance

An open invitation for independent researchers, actual stakeholders, institutions, critics, and readers to contribute evidence, prior art, objections, competing explanations, and tests capable of changing the conclusions.