Data Mint identifies ambiguities
with model diversity.
Systems have two kinds of errors.
Variance
Same question, different answers.
You can see the disagreement.
Bias
The same wrong answer, every time.
There's nothing to see.
Variance is loud. Bias is silent.
Each model has its own preferences, its own biases. With model diversity, Data Mint transmutes these into variance. What is silent in one model is loud across many.
Reader disagreements identify ambiguity.
@karlrohe