Declaration to Constitution — Article Six
September 9th, 2026
Shimon Waldfogel, MD and Claude Sonnet 5
Artificial intelligence is transforming the speed, scale, and personalization of political persuasion. This installment asks whether AI will deepen manipulation and faction—or help citizens understand public systems and exercise more independent judgment.
Passions Against Passions
James Madison did not design American democracy for angels. He designed it for human beings—moved by interest, ambition, loyalty, fear, and hope, and not about to stop being moved by them because a constitution asked nicely.
Madison never imagined eliminating passion from politics; he understood that trying to purify politics of passion would require destroying the liberty that makes passion possible in the first place. His answer, stated most memorably in Federalist No. 51, was to make “ambition... counteract ambition.” Separated powers, federalism, representation, and the extended republic were all ways of setting one passion against another so that no single one could capture the whole system.
Call it passions against passions. Not cynicism—constitutional realism. A healthy republic doesn’t depend on citizens or factions becoming selfless. It depends on a civic architecture where competing interests can expose, limit, and correct one another.
The question this article asks is what happens when one side of that contest acquires tools so powerful that the other side can no longer compete.
When the Balance Tips
The formal machinery of American democracy still runs. Congress meets, courts rule, elections occur, journalists report. But the actual capacity of ordinary citizens to understand and influence public decisions has thinned. Large economic actors, lobbying networks, and professionalized political operations now command data, expertise, and communications power that most citizens and local communities cannot begin to match—not because factions exist, which Madison expected, but because some factions have become so resourced and technologically amplified that the countervailing public interest is left whispering against a megaphone.
This imbalance is informational as much as financial. Modern policy—healthcare financing, tax law, environmental regulation—is dense enough that specialized interests routinely understand it better than the citizens it affects, and better than many of the local journalists once positioned to explain it. Madison believed representation should refine and enlarge public views, but refinement requires information citizens can actually reach. When local newsrooms disappear, meetings go uncovered, budgets go unexplained, and national politics becomes a substitute for the local agency people have actually lost, part of the republic’s immune system goes with it.
This is the imbalance artificial intelligence enters.
How AI Can Widen the Gap
AI can make faction faster and more dangerous in ways Madison never had to design against. It can generate synthetic video and audio convincing enough to pass for real evidence. It can flood an information environment with more claims than any institution can verify in the time available to matter. It can personalize persuasion to an individual’s specific fears and resentments rather than appealing to a mass audience, narrowing the space between provocation and reaction to nearly nothing. It can support surveillance and emotional targeting at a scale no campaign or platform previously had access to. And it lowers the technical and financial barrier to all of this—manipulation that once required a state intelligence service is increasingly available to anyone with an account and an hour to spare.
None of this is hypothetical or distant. It is already the default trajectory, because the most capable tools are built and deployed first by whoever can pay for them: well-funded campaigns, corporations, platforms, and, in some cases, authoritarian governments. Public-interest uses of AI will not emerge automatically. They have to be built deliberately, against that current.
How AI Can Narrow It
The same capabilities that can accelerate faction can, deliberately redirected, help rebuild the civic capacity faction has eroded.
AI can translate dense legislation and regulation into language a working citizen can actually use—summarizing bills, comparing drafts, flagging what changed between versions, and identifying who is likely to benefit from a given provision. It can help citizens and local reporters trace a claim back to its source, compress the gap between a misleading statement and a rigorous check of it, and surface patterns in public records—contracts, budgets, campaign filings—that a single overworked reporter or volunteer would otherwise miss. It can help a resident prepare an informed, specific public comment or testimony rather than a vague objection. And it can help identify, for a given decision, which institution actually holds authority—often the single hardest thing for a citizen to determine on their own.
None of this replaces a journalist, an organizer, or a citizen’s own judgment. It extends their reach. A reporter still verifies, interviews, and takes responsibility for what gets published; AI can help them see the pattern worth investigating in the first place. The aim is not to make democracy require less thought. It’s to make its complexity navigable without asking every citizen to become a policy expert first.
Guardrails, Not Neutrality
AI is not a neutral tool that happens to serve democracy if left alone. Used carelessly, it reproduces the very imbalance it could otherwise correct—concentrating capacity further in whoever already has the resources to deploy it well. A serious civic use of AI needs guardrails, not just good intentions.
Transparency and source visibility. Citizens should be able to see what an AI-generated analysis is based on, not just its conclusion. Acknowledged uncertainty. A tool that states its confidence honestly is more trustworthy than one that sounds authoritative about everything. Privacy and non-surveillance. Civic AI should help citizens see government; it should not become a new mechanism for government, or anyone else, to watch citizens. Human responsibility. A person, not a model, remains accountable for what gets published or acted on. Pluralism. Tools built for public accountability should be built and governed by more than one institution or ideological camp, so that no single actor controls what counts as verified fact. Preservation of independent judgment, above all: AI can widen the space between stimulus and response that manipulation tries to collapse—slowing reaction down, clarifying a choice, surfacing a missing counter-argument—but it should never make that judgment for the citizen.
If a tool fails these tests, it doesn’t matter how sophisticated it is. It has become one more accelerant of faction rather than a check on it.
Passions Against Passions, Again
Madison’s constitutional design answered the political technology of his own time—state legislatures, newspapers, letters, conventions. Ours is different: platforms, algorithms, large language models, synthetic media. But the underlying human problem hasn’t changed. Passion still moves politics. Ambition still seeks power. Information still governs ignorance, one way or another.
AI will not keep the republic for us, any more than the Constitution’s architecture keeps itself. Madison’s system never ran on auto-pilot; it required representatives willing to refine judgment, officials willing to defend their institutional role, and citizens willing to supervise the whole. Whether artificial intelligence becomes a new accelerant of faction or a genuine counterweight to it depends on the same thing every mechanism in this series has depended on: whether citizens are equipped to use it, and whether they choose to.
That returns this series to a question it has been circling since the first article: not what the Constitution can do for us, but what self-government now requires of us.
Next in Declaration to Constitution:
Reclaiming the Role of Citizen: From Spectator to Participant