Consider a credit assessment model that provides a calibrated score indicating an individual's creditworthiness, with a certain margin of error.
For instance, if the model predicts a score of 83%, the true score is likely between 81% and 85%.
If we set 83% as the threshold for determining creditworthiness and classify individuals with scores above this as creditworthy, are we likely to be overestimating or underestimating the actual creditworthiness of the population?