Choice Architect
Treat your criteria as polytomous test items, your options as respondents. Recover a calibrated latent fit-to-need score for each option, with a 95% credible band — the same Rasch + Bayesian mathematics that powers the Emotion Tensor Forge, applied to choice.
Criteria
Weight: 1 = standard · >1 = critical · <1 = minor. Maps to IRT b-difficulty.
Options
Group lets you compare cohorts (e.g. "incumbent" vs "new"). Each option becomes an IRT respondent.
House note
The quick-rank below updates instantly as you score. Run IRT for calibrated θ, then Bayesian for credible bands — both compute on GCP Cloud Run.
Score matrix
| Criterion | 🎯 Fit | 💰 Cost | ⚠️ Risk | ⚡ Speed | ↩️ Reversibility | 🚀 Upside |
|---|---|---|---|---|---|---|
| Strategic alignmentw=1.5 | ||||||
| Implementation costw=1 | ||||||
| Time to valuew=1.2 |
0 = absent · 2 = neutral · 4 = strong. Each cell is one ordinal IRT response.
Ranked options
Score combines axis polarity, criterion weight, and ordinal response. Run IRT for calibrated θ.