High precision considers false positives as costly whereas a false negative is not as costly. For instance, it will be an issue if a facial recognition system falsely classifies me to be a member on the system and gives me access so we would increase the threshold on our classifier to reduce the chances of this happening. Therefore, if we have high precision then we will have low false positives.

On the other hand, if we have high recall then false negatives are considered more costly and false positives are not as costly. I.e. predicting fraud when it’s not fraud is much better than saying no fraud when there actually is fraud. Therefore, false negatives will be lower.