Most people never see what happens between answering an assessment’s questions and receiving a result. That gap is where trust is either earned or lost, so it is worth explaining, in plain terms, what actually happens inside Lux between those two moments.
What “19 layers” means in practice
Lux’s scoring is built as a 19-layer engine spanning 301 fields. Each layer has one job: turning raw responses into progressively more meaningful structure, from item-level scoring through to the narrative language a person eventually reads in their report. Splitting the engine this way means every step is small enough to test on its own, and scoring changes can be checked through automated regression tests and selected reference fixtures.
That matters because psychological scoring is easy to get subtly wrong in ways that are hard to notice later. A transposed weight, a miscalculated norm, a rounding error carried through several steps: these are the kinds of mistakes that don’t announce themselves.
Testing layers against committed fixtures, and versioning the whole pathway, helps us detect unintended computational changes before release.
What the reliability figures actually say
The complete ten-item Big Five item banks used as source/reference material carry documented internal consistency of α .800 to .898 in a 603,322-response source/reference dataset. Internal consistency describes how closely responses to items within each bank relate in that dataset.
That is source-bank internal-consistency evidence. It does not establish that a bank measures one underlying construct, that Lux’s proprietary fields are valid, that Lux itself is reliable, or that Lux is fit for a particular use. The 603,322 responses are not Lux-participant responses. We state what the figures support and stop there.
Why we lead with this instead of hiding it
Publishing the mechanics, available evidence and limits gives readers a clearer basis for judging the work. Versioning and testing make the scoring system more inspectable, but they do not replace psychological validation.
Lux is moving through an active, staged evidence-development programme. Test–retest work will assess score stability as repeat-response data becomes available. Later stages will examine convergent and discriminant validity for proprietary Lux fields, representative norms, intended-population fairness and clinical utility using suitable data and pre-defined study designs. This work is active, but it has not yet produced completed Lux-specific validation results.