Model Uncertainty refers to the probability that we are fundamentally confused about the situation we are modeling, and an attempt to give such an error some weight in our model. In a traditional probabilistic calculation, say a coin flip, the probabilities of heads or tails are expected to add up to 1. But this admits 0 probability of ANY third outcome. We should not be aboslutely certain of such a thing no matter how simple the system we are modeling. The coin may land on its side, the coin may be snatched from the air and made off with, or any number of things that we haven't thought of. We can reduce model uncertainty by acknowledging our key assumptions that define the model.