Machine learning systems trained on datasets collected from biased samples can exhibit these biases when they are used; this is called

Select one:

A$.$ an algorithmic bias

B$.$ data shift

C$.$ a false positive

D$.$ underspecification