Back in the early 1960s Stanley Milgram conducted a famous experiment at Yale University. Volunteers were told that it was designed to test the effect of punishment on learning. Would a person learn a list of word pairs better if they were punished every time they got an answer wrong? The volunteer was instructed to deliver an electric shock to the learner every time one of his answers was wrong. The shocks increased in intensity for every wrong answer. Of course, the experiment wasn't actually about the effect of punishment on learning at all. It was really designed to see how long the volunteers would obey the authority of the researcher. Would they continue to give electric shocks to the learner even when it appeared that doing so would kill the learner? Over sixty percent of them went ahead and gave the shock. They were led to believe that they had killed or seriously injured the learner (who was actually just an actor).
Milgram's experiment is one of the most famous experiments of all time. But it provoked a lot of controversy about whether it was ethical. Often the volunteers were reduced to nervous wrecks as they struggled over whether to continue obeying the researcher, or to refuse to do so. No review board would ever approve such an experiment today.
Mel Slater, a Computer Science researcher at University College London, has announced a possible way around these ethical concerns. He replicated Milgram's experiment using a virtual learner. LiveScience reports
When the virtual woman gave an incorrect answer, the participants were told to give a virtual 'electric shock' that buzzed to her, increasing the voltage each time she gave an incorrect answer... Over time, she responded with increasing discomfort and protests, eventually demanding the experiment stop. Near the end, her head would slump forward and she became unresponsive... 17 gave all 20 shocks and three gave 19 shocks, 18, 16 and 9 shocks were given by one person each. When volunteers were asked whether they had considered aborting the experiment, nearly half of those who could see and hear the virtual woman indicated they had because of their troubled feelings about what was happening. In addition, their heart rates indicated that participants reacted as though the situation was real.
I don't know. I'm having a hard time buying that a virtual learner could ever substitute for a real, living, breathing learner. However you parse it, thinking you've killed a virtual character is not the same as thinking you've killed a real person. It's like saying Milgram could have used mannequins instead of real people. It just wouldn't have been the same.