Hoplight Intelligence
"What happens if you message to someone's psychology instead of to their demographics?"
The groups progressives count on are leaving. The headlines say it all.
The standard progressive diagnosis assumes the problem is information. It's identity.
Abandon your issues.
Talk like a 1990s centrist.
Hope the base shows up anyway.
Double down on values.
Educate voters.
Risk winnable races.
Both sides assume you can only win with one message. We tested a third option.
Psychographics let us see within demographics, where the persuasion opportunity actually lives.
Who someone is.
Age, race, income, education.
What is this person's race?
How someone makes sense of the world.
Values, beliefs, perceived threats, emotional triggers.
What does this person perceive as a threat?
Cut across race, class, and party: a Latino voter could be in any of these five groups.
Believe in institutions, collective action, solidarity.
Left-leaning but deeply skeptical of power. Anti-corporate, anti-establishment.
Seeking safety, order, protection from chaos.
Value self-reliance, opportunity, meritocracy.
Grounded in faith, tradition, moral structure.
We tested whether different psychological frames could activate different values segments.
Issue tested: "Many security officers are poorly paid and resourced."
| Progressive Frame* | Personal Safety | Pride / Valor | |
|---|---|---|---|
| Hero | Officers as community caregivers | Officers as guardians of your family, office, school | Officers as frontline defenders, "the uniform" |
| Villain | Management interference, low standards | Big companies cutting corners | CEOs profiting while officers are under-equipped |
| Emotional Anchors | "Coming together across race to raise standards" | "Your safety depends on officers being properly equipped" | "Officers put their lives on the line. They deserve the same respect." |
* Human-generated by progressive communicators. Safety and Valor frames generated by the engine.
Click to read the exact text each group received.
Security Officer Messaging Study · n=3,006 · voter-file matched · August 2025
On union support, the staff-written progressive frame was the only message that lost ground, and it lost the most with the people who decide elections.
The progressive frame performs well with the base. It fails outside of it.
Points of agreement over the placebo group, officer pay. n.s. = the vendor did not flag this cell, so its movement cannot be told apart from no change.
The staff-written frame does best with the progressive segments. It fails outside them.
Different psychologies need different frames for the same goal.
When you match the frame to the listener's psychology, the backlash disappears.
With liberals and Democrats, the AI-generated frames performed about the same as the staff-written frame. With the groups Democrats lost in 2024, they recovered ground the staff-written frame lost.
Points against the placebo group. Significance as flagged by the vendor: 90% confidence versus the placebo. Respondents were general-population adults; the trial measured attitudes, not votes.
Deep canvassing proves that meeting people inside their worldview moves them. The problem is doing it at scale.
A canvasser who listens, adapts, speaks in the listener's values: that works. But these tactics are prohibitively expensive to scale.
Research shows large language models can persuade. But if the model changes, the message changes. No audit trail. No replicability.
Psychographic targeting with auditable infrastructure. The persuasion spec controls the output, the LLM writes to spec. Model-agnostic.
We invest millions in research and message development. But implementing those learnings at scale is riddled with drift.
The current process:
With the engine:
Built on 22 psychometric measurement traditions and 110 subscales. The four layers are the public map. The proprietary pipeline spans eight stages.
Measures who the listener is psychologically. Takes signals in and outputs a factor vector.
Projects the profile into felt worldview: archetype, antagonists, fears, and internal story.
Maps worldview directly to frame, narrative, metaphor, register, and appeal type.
LLM writes prose to spec. The moat is the spec it receives from Layers 1-3.
Build a message in the morning.
Field it in the afternoon.
Read results, tune the tool, repeat tomorrow.
Move the people other programs leave out.