Everyone knows "women and children first." The question is why it reached some and not others. Watch the vocabulary find what no chart could.
Dataset891 passengers · 12 columns
StagesA through E
Key moveThe policy existed. The access did not.
Stage AThe Raw Data1 / 9
📊
891 rows. 12 columns. Every answer is in here somewhere. But right now this data says nothing — because it has no vocabulary.
Press Next to begin
The dataset had the answer all along. pclass, sex, age, fare — those columns were real. But "language barrier" is not a column. It is a concept a human who knows this history brought into the room. The AI confirmed it belonged. The data showed it was green on the death board and red on the survival board.
That is the method. Not better statistics. A different starting point — one where the vocabulary is built before the analysis begins.
"A rejected word is not a failure. It reveals an assumption the group was making without knowing it."
How the colours work
The tile colours are not assigned manually. Centrality — green, yellow, red — is estimated by the LLM using semantic density analysis across the passenger metadata and its training corpus. A concept that clusters densely with survival-relevant variables across the domain scores green. A concept present but not decisive scores yellow.
"Informed early" and "unaware" were proposed by the human team as plausible concepts. The AI analysis elevated them — finding high semantic density in maritime disaster literature linking information access to survival outcomes. The ★ marks tiles where AI centrality analysis confirmed weight the human team had underestimated.
"3rd class" scores yellow on both boards — present but not dominant on either side. "Language barrier," "below deck," and "far from exit" score green on the death board together. They are a cluster — not independent causes. Being a 3rd class emigrant meant being all three simultaneously. A regression finds that cluster and reports it as "class." The Databoard names what is inside it. Whether communication or physical distance was the primary driver is the next question the board opens — not one it forecloses.