Sanctity Be the Voice
Phase Shift Tournament begins: close calls only

Before we start

A few quick filters so the profiles you see match what you are actually choosing between.

These filters only shape the profiles you will be shown in this session. None of your answers are stored beyond this game session unless you opt in at the end.

One moment before your results

Tell us about yourself so we can show you how your worldview compares to others: once enough people have played. Skip any field. None of this is required.

Stored locally on your device only. When peer-comparison launches at the sample-size threshold, your tags will be hashed and aggregated: never shown individually.
A research preview by Sanctity AI · v4

BasketSort

Forced-choice preference elicitation. Pick the better basket repeatedly, and the math behind it discovers what you actually value.

Sessions on this device 0
How deep?

Active games 8 Basket bundles and 1-vs-1 decisions

Under development 4 Not in this demo build

in session Round 1 / 30
Progress
exploring · priors

-

Option A -
VS
Option B -
Admin · Weight trace

Your archetype

-

-

Picks
0
In-sample fit
-
Top
-
Spread
-

Your ranking

95% credible intervals

Confidence tiers i Adjacent ranks with overlapping CIs get grouped. The model is confident about which tier each item belongs to, less confident about the order within a tier.

grouped by overlapping CIs

Convergence over rounds

how your worldview took shape

Pairwise matchup matrix i For any two entries, predicted probability the row beats the column 1-on-1.

row beats column

Pivotal round i The single round whose pick shifted your inferred ranking the most.

biggest shift

Your profile: in words

auto-generated narrative

Why is one ranked above the other? i Pick any two entries to see direct evidence (rounds they appeared opposite each other), counter-evidence, and the weight gap.

unpacks the model's reasoning
vs

Decision quality i Decisive (>80% predicted), borderline (40-60%), contradicted (<20%: you went against your own pattern).

how predictable your picks were

Information per round i Bits of evidence each round contributed. Tall bars = close calls that resolved uncertainty. Short bars = obvious picks.

evidence per round

Calibration i For groups of rounds the model predicted with similar confidence, what fraction did go that way. Diagonal = perfect calibration.

predicted vs actual

Time travel: replay your worldview round by round i Slide through every round to see your inferred ranking at that exact point.

drag the slider
R1 R30

Robustness i How many picks would need to flip before each adjacent pair swaps. Larger = more robust ranking.

picks-to-flip per gap

Logical inconsistencies Beta i 3-cycles (A>B>C>A) in your beats graph. Logically inconsistent picks, common in real human judgment.

cycles in your beats-graph

Bundle simulator i Toggle items into Side A and Side B to see which combination your model predicts would win, with confidence.

try any matchup

Tournament champion Beta i Monte Carlo 10,000 single-elimination tournaments using your inferred weights. Most-frequent champion.

10,000 bracket simulations

Bootstrap rank stability Beta i Resample your picks with replacement 200 times, refit the model each time, and record the rank position each item lands at. Tells you which positions are confidently held vs interchangeable.

200 bootstrap refits

Cross-session drift Beta i If you've played this game before, this compares your current ranking to your last session via Spearman rank correlation.

how your worldview shifts

Play more