The real problem is that the test itself might have affected the results. The presence of scientists might have created into the subjects a sense of necessity to fulfill the testers desires. (Which is pretty common in Sociological and behavioural scieneces and a HUGE problem) So, making the subjects play a good or evil character could have created a subconscious need of fulfilling expectations into the participants, who would affect the results in a pretty big way.
Devastation3000's comments