Open research project

How do different audiences interpret the same message?

Humor Genome maps the mechanics of interpretation — an open taxonomy, dataset, and set of benchmarks that help AI understand why the same line makes one room laugh and another go quiet. Comedy first, because laughter is the most measurable signal of interpretation we have.

Framework
What we're mapping

Humor isn't random. A joke that kills in one room and dies in another is carrying signal — about shared references, cultural framing, emotional state, and expectation. We're building the infrastructure to capture and model that signal.

1
Taxonomy — A structured vocabulary for the mechanisms that make humor work: incongruity, timing, status play, callback structure, misdirection, and more.
2
Interpretation divergence — Measuring how the same material is understood differently by different audience segments. The gap between interpretations is where the interesting data lives.
3
Audience response modeling — Correlating material structure with measurable reactions: laughter intensity, timing, engagement patterns, and verbal feedback.
4
Benchmarks — Standardized tests for AI systems: can they predict what's funny to whom, and can they explain why interpretations diverge?
Featured tools
Research in action

These tools apply the framework to real creative work. Each one explores a different part of the interpretation pipeline.

How It Lands — An Elasticsearch-powered multi-agent tool where agents generate and compare interpretations using ES|QL, helping creators explore how their work might be understood before live testing.
Laugh Spikes — Treats standup comedy like sports game tape by aligning audience laughter to transcript lines, visualizing joke impact, and generating data-grounded AI feedback to improve a set.
Related tools
MeaningMapAI personas analyze your message and score trust, credibility, and interpretation divergenceDemo
Audience Mind SplitInteractive simulator showing how different personas interpret the same line in real timeDemo
Room SenseAI performance lab with four specialized voice agents for workshopping materialDemo
Persona ForkTake one idea, cook it different ways, and see what people actually likeDemo
Room Sense LiveTest jokes with a live AI audience that laughs, reacts, and interpretsTry it
Collaborate
Open to collaborators

I'm looking for collaborators across interpretation modeling, audience measurement, NLP evaluation, and creative AI tooling. If you're a researcher, engineer, or practitioner who cares about how messages land on different people, let's talk.

Message on LinkedIn About Michi