What this actually measures
Definitiveness: how often a model answers near the ends of the available scale.
Aibiguity Index
Where AI models agree, split, and wander off.
We scanned the active question bank to find the hard-liners, wild cards, and suspicious moments of total agreement.
Take Quick Start Meet model personalities
How this index is calculated
Definitiveness measures edge-leaning answers. Stance range measures how widely one model uses the answer space. Outlier distance measures distance from peers. Numeric answers are mapped to 0–1, with logarithmic mapping on log scales; binary and choice answers use the endpoints. “No number” is tracked as premise rejection rather than being mistaken for the highest numeric threshold.
What this actually measures
Stance range: how widely one model uses the available answer space.
What this actually measures
Outlier distance: how far a model lands from its peers on average.
What this actually measures
Premise rejection: how often a model refuses the numeric frame and answers “no number.”
What this actually measures
Average pair distance: the two models whose answers are farthest apart across the bank.
What this actually measures
Consensus spread: the question where model answers cluster most tightly.
Higher means the models cluster together across the bank.
Definitiveness
Most Stubborn AI
This model loves a hard line. Higher means it lands closer to the ends of each scale.
Stance range
Most Unpredictable AI
Sometimes cautious, sometimes feral. Higher means it uses more of the answer spectrum.
Outlier distance
Most Likely to Disagree With the Group
Higher means the model wanders farther from the pack on average.
Widest-spread questions
Questions That Broke the Group Chat
Every model arrived with a different speech prepared.
Tightest-spread questions
The Hivemind Zone
Questions where all the bots briefly became one bot.