I believe the answer is: Known-groups paradigm
A study is considered to have Known-groups paradigm if the researchers understand that the measure that they take will discriminate a certain groujp of subjects because of their known behavior.
From the case above Dr. Sheffield create a test to determine whether the subject unable to resist impulses to gamble.
The subjects that he use is a group of ex-gamblers who acknowledge they have a gambling problem (gamblers anonymous) and a group of ex-alcoholic who acknowledge they have drinking problem (AA). We can say with high certainty that dr. Sheffield knew really well that people on GA will have higher scores. He just want to how big is the difference compared to the scores from AA group.