The Archaeology of Belief

How a 50-Year-Old's Politics Were Decided by Age 12

Brian Demsey | Published in The Information | 2025

← Back to Articles

Seven voices. Ten variables. One uncomfortable truth.

Thesis: The information ecosystem we inhabit is not a marketplace of ideas—it is a mirror. By age 12, family and social environment have encoded our political operating system. Everything after is confirmation bias with better vocabulary.

Subject: A person born in 1975, now age 50. Formative years: 1980-1993.

The Ten Variables

#VariableSpectrum
1Economic ModelFree Market ↔ Regulated/Redistributive
2Government RoleMinimal ↔ Expansive
3Social HierarchyTraditional/Ordered ↔ Egalitarian/Fluid
4National IdentityBlood/Soil ↔ Civic/Ideological
5Outsider TrustSuspicious ↔ Welcoming
6Authority SourceReligious/Traditional ↔ Secular/Empirical
7Risk ToleranceSecurity First ↔ Liberty First
8Time OrientationRestore Past ↔ Build Future
9Moral FoundationPurity/Loyalty ↔ Care/Fairness
10Change VelocitySlow/Cautious ↔ Rapid/Disruptive

The Seven Voices

Voice 1: The Republican

White, Suburban Ohio, Methodist | Father: small business owner | Mother: homemaker

I was twelve when Reagan said "Mr. Gorbachev, tear down this wall." My father cried. First time I saw that. He'd built his plumbing supply company from nothing—no loans, no handouts. Government was the thing that audited him, taxed him, regulated him.

My variables were set at the kitchen table:

Economic Model:Free Market — "Dad's business = family survival"
Government Role:Minimal — "IRS audit when I was 9"
Social Hierarchy:Traditional — "Church every Sunday, gender roles clear"
Outsider Trust:Suspicious — "Lock your doors, this neighborhood's changing"
Time Orientation:Restore Past — "Things were better before..."
Everything I've read since age 12 has confirmed my father was right. Or I've only read things that confirm it. I cannot tell the difference anymore.

Voice 2: The Democrat

Black, Urban Detroit, Baptist | Father: UAW autoworker | Mother: school nurse

I was twelve when they closed the Chrysler plant. My father had twenty-two years in. They gave him a watch. We kept the house because my mother's job had benefits.

My variables were set on the picket line:

Economic Model:Regulated — "Union saved us; corporation abandoned us"
Government Role:Expansive — "Food stamps bridged the gap, no shame"
Social Hierarchy:Egalitarian — "Skin color had determined everything for generations"
Time Orientation:Build Future — "Your generation will see better"
Change Velocity:Rapid — "Patience had been weaponized against us"
The internet feeds me exactly what I already knew. The algorithm is my grandmother with better data.

Voice 3: The Independent

White, Rural Montana, Agnostic | Father: ranch hand | Mother: waitress, later paralegal

I was twelve when the bank took the Hendersons' place next door. My parents didn't own—we rented—so we just watched. Nobody came from Washington or anywhere else. The Hendersons drove away and we never saw them again.

My variables were set in the silence:

Economic Model:Skeptical of Both — "Banks are predators; government is absent"
Government Role:Irrelevant — "It never showed up when needed"
National Identity:Local — "Montana first, America second"
Risk Tolerance:Self-Reliance — "Neither security nor liberty—just survival"
Change Velocity:Indifferent — "Change happens to us, not by us"
I don't trust either party because neither party knows I exist. The internet offers me conspiracy theories and I understand why people believe them. Abandoned people build their own explanations.

Voice 4: La Voz Hispana

Mexican-American, Los Angeles, Catholic | Padre: construction foreman | Madre: hotel housekeeper

I was twelve when my uncle got deported. He'd been here nineteen years. His kids were American. My mother didn't sleep for a month, terrified my father was next. He had papers. She didn't, until 1986.

Mis variables fueron establecidas en el miedo:

Economic Model:Labor-Protective — "Watched parents exploited, underpaid"
Government Role:Ambivalent — "It can legalize you or deport you"
National Identity:Hyphenated — "Too Mexican for Americans, too American for Mexico"
Risk Tolerance:Security First — "Papers = safety"
Moral Foundation:Family/Loyalty — "La familia lo es todo"
The algorithm serves me in two languages. I code-switch between worlds the internet thinks are separate.

Voice 5: The Chinese-American Voice

San Francisco, Buddhist/Secular | Father: engineer (immigrated 1969) | Mother: pharmacist

I was twelve when Vincent Chin was beaten to death with a baseball bat by autoworkers who thought he was Japanese. They served no jail time. My father, who had built missile systems for Lockheed, stopped wearing his company badge outside work.

My variables were set in the gap between achievement and acceptance:

Economic Model:Meritocratic Ideal — "Work twice as hard, complain half as much"
Social Hierarchy:Contradictory — "Model minority myth—praised but not trusted"
Outsider Trust:Conditional — "Welcome until geopolitics shift"
Authority Source:Secular/Empirical — "Education is the only unambiguous asset"
Moral Foundation:Achievement/Harmony — "Don't make waves; make progress"
My children face questions I thought we'd answered. The algorithm cannot decide if I am American or threat.

Voice 6: The Muslim Voice

Palestinian-American, Chicago, Sunni | Father: doctor (immigrated 1967) | Mother: professor of literature

I was twelve during the Achille Lauro hijacking. A boy in my class asked if my father was a terrorist. My father had treated that boy's grandmother in the ER the week before.

My variables were set in the space between belonging and suspicion:

Government Role:Wary — "Post-9/11 surveillance, PATRIOT Act"
National Identity:Contested — "American—until foreign policy intrudes"
Risk Tolerance:Liberty First — "Security theater targets us specifically"
Time Orientation:Long Memory — "Palestine is not abstract; it is family"
Change Velocity:Urgent — "Status quo is not neutral"
The algorithm shows me two internets: one where I am neighbor, one where I am suspect. Same country. Same person. Different feeds.

Voice 7: The Jewish Voice

Reform, New York, Second-Generation | Father: attorney | Mother: social worker | Grandparents: Holocaust survivors

I was twelve when they caught Jonathan Pollard. The questions at school shifted. Dual loyalty. "Are you American or Jewish?" As if my grandmother's number tattooed on her arm wasn't answer enough about what happens when a country decides you don't belong.

My variables were set in inherited memory:

Government Role:Protective — "Strong institutions prevent pogroms"
National Identity:Dual Consciousness — "American Jews are safe until we're not"
Authority Source:Textual/Questioning — "Talmudic tradition: argue, interpret, argue again"
Time Orientation:Cyclical — "History rhymes; always watch for signs"
Moral Foundation:Justice/Survival — "Tikkun olam—repair the world—but protect your own"
The algorithm cannot reconcile my positions. I am progressive on domestic policy, hawkish on Israel, suspicious of both left and right antisemitism. The internet wants me to choose a tribe. My tribe has already been chosen for me.

The Uncomfortable Truth

Each of these voices arrived at age 50 having spent 38 years confirming what they knew at 12.

The variables were set by:

The internet did not change this. The internet weaponized it.

Every search is a mirror. Every feed is an echo. Every algorithm is grandmother's wisdom at scale—optimized for engagement, not truth.

The Smallest Set of Words

If we reduce political identity to its minimum:

Safety
Whose? From what? Guaranteed by whom?
Fairness
Defined how? For whom? Measured how?
Belonging
Who is "us"? What must "they" do to qualify?

These three words explain 90% of political conflict. Everything else is commentary.

Brian's Challenge

"I make no apology. I spend your tokens with reckless abandon and wait for the day that one of you will be a hero."

The hero's task is not to eliminate bias. That is impossible.

The hero's task is to make the bias visible.

H-LLM doesn't judge which voice is correct. H-LLM shows where eight models agree (probable fact). H-LLM shows where eight models diverge (probable bias, probable projection).

The algorithm doesn't cure prejudice. The algorithm diagnoses it.

That's the beginning. Not the end.

Don't lament. Engage.

Brian Demsey is the founder and CEO of Hallucinations.cloud LLC, an AI safety company focused on multi-model truth verification. He has over fifty years of experience in enterprise technology.