Method

Revealed preference

A person cannot describe themselves accurately. Social desirability, the limits of self-knowledge, the share of language. But in the moment of choosing, they reveal what they want. ChoiceCodex is built on this principle: observation, not description.

Bayesian update

Ten Origins begin at equal probability. Two paths in each scene; the choice is mapped to twenty-two tags; a tag-affinity matrix updates the distribution. After the twenty-first, one or two patterns come forward.

Six axes × six modifiers

The ten Origins are defined along six behavioural axes: activation, resource priority, mode of analysis, rule orientation, threat response, scope. Each axis is a spectrum, not a category. Origins land in different places along these spectrums.

Uncertainty report

The posterior is a distribution, not a verdict. ChoiceCodex does not conceal it. The top three Origins are shown. A confidence band is reported. Readings below threshold are marked uncertain.

What we avoid

• No 'you are X.'
• No score. No ranking. Only a distribution.
• No right answer. Only tag-affinity.
• No life prediction. Only the pattern of this voyage.

Replicability

The same decisions yield the same posterior. The scenarios, the tag-affinity matrix, and the Origin definitions are published as open documentation.