When it gets hard,
can people still
think together?
I study how people make sense of things together — and what makes it possible — by reading language as a signal of collective state, not sentiment. One practice builds the conditions. One instrument measures whether they hold.
We optimize what we measure. We've been measuring the wrong thing.
GDP measures output, not trust. Engagement measures attention, not understanding. Sentiment sorts mood, not whether nuance survives contact with difficulty. None of it measures the thing societies actually run on: whether people can still make sense of what's happening, together, when it gets hard. When that ability starts to degrade, it's the first signal something is going wrong — and nothing currently watches for it. You cannot protect what you cannot measure.
Two halves of one question. One creates the conditions for people to think together — the other measures whether it's working.
The Conversation Zone
A facilitated series that holds hard topics — race, conflict, division, trust — without heroes, villains, or sides. The point isn't to argue a position; it's to make space that's safe to think in. It's where the conditions get built, and where the behavior gets recorded.
Facilitation · live series · datasetMeasuring orientation
A way to read, from language alone, whether a group is defending, making sense, integrating complexity, or reaching for each other — and which direction it's moving. States, not types. Movement, not mood. It reads whether people are able to think together, without trying to change what they think.
Adaptive Cognition Measurement SystemA behavioral record of collective cognition under pressure, gathered through a live national crisis: retention curves for what people do when tension arrives, coded discourse for what they say. The findings appear in both.
Orientation, not opinion.
Each response is read along two independent axes — orientation (disoriented → stabilizing → orienting) and relational openness (closed → reachable) — so that making sense of events while closed to others reads differently than doing it while still reachable.
These are positions people move between, not fixed types. Within a single thread the data shows people traveling the path when conditions support it, and retreating back down it when they don't.
The same moment, three states — real comments from the study, sorted by function, not side.
“Trump actually is one of the worst people to walk the planet.”
“AMERICAS greats threat is WOMEN with a self righteous heart and attitude.”
“Both sides feel like victims, because both sides are right.”
“Being on the internet is rage fuel. But ignoring what’s going on is how Hitler took over… What am I supposed to do?”
“How do we do this together? We need each other more than our differences keep us apart.”
“Healing can only begin with understanding, and wounds heal better with care.”
I spent fifteen years inside the systems that engineer meaning and attention — brand and messaging work for companies like Roku and NBC Universal — learning to read behavior, not what people say about themselves. I came to see what those systems optimize for, and what they can't see. Now I build the instrument that can.
Read the full story →
This isn't a one-person project.
The study's limits are its openings, and each one names the partner it needs: inter-rater validation against the corpus, training classifiers to read populations, replication with other facilitators and audiences, and piloting the instrument inside real systems — to ask whether sustained contact with a system opens people or closes them.