The statistics behind A/B testing only works if things are uncorrelated and independent.
If criminals do know know whether they are A or B, then the existence of the A group can deter crime in the B group. Given the increased scrutiny on A, you're likely to catch more crimes there. That doesn't mean that B doesn't have uncaught crimes though.
Thus your test is fundamentally messed up both coming and going.
If criminals do know know whether they are A or B, then the existence of the A group can deter crime in the B group. Given the increased scrutiny on A, you're likely to catch more crimes there. That doesn't mean that B doesn't have uncaught crimes though.
Thus your test is fundamentally messed up both coming and going.