Slashdot Mirror
Computer Characters Tortured for Science
Rob Carr writes "Considered unethical to ever perform again with humans, researcher Mel Slater recreated the Milgram experiment in a immersive virtual environment. Subjects (some of whom could see and hear the computerized woman, others who were only able to read text messages from her) were told that they were interacting with a computer character and told to give increasingly powerful electric shocks when wrong answers were given or the 'woman' took too long to respond. The computer program would correspondingly complain and beg as the 'shocks' were ramped up, falling apparently unconscious before the last shock. The skin conductance and electrocardiograms of the subjects were monitored. Even though the subjects knew they were only 'shocking' a computer program, their bodies reacted with increased stress responses. Several of the ones who could see and hear the woman stopped before reaching the 'lethal' voltage, and about half considered stopping the study. The full results of the experimental report can be read online at PLoS One. Already, some (like William Dutton of the Oxford Internet Institute) are asking whether even this sanitized experiment is ethical."
So when does this come out for the Wii?
There are no loopholes. It's either legal or it's not.
Take two groups: One has first gone through this "virtual torturing", the other is the control group. After this, each group will actually torture a volunteer in the same manner. Would the first group have less of an emotional response than the control group? I am sure there are many wrinkles to work out in the methodology, but this would be interesting to see the result of media on human response. It should pretty effectively answer who is right (or how right each side is) in this debate.
See my journal for slashdot ID's by year. Mine created in 2005. http://slashdot.org/journal/289875/slashdot-ids-by-year
I don't see anything in the study that says that they made any attempt to find out whether or not the subjects had ever heard about the original Milgram experiment.
The Milgram subjects almost certainly had no knowledge of whether the situation was real or what the purpose of the experiment was, and probably believed that they were "supposed" to follow orders.
Today's subjects may well have heard something. Even if they couldn't have named "Milgram" as the investigator, they may have had more than an inkling that the purpose of the experiment was to see whether they were virtual sadists, and may have suspected that, despite their instructions, the "approved" behavior was to not to follow orders.
"How to Do Nothing," kids activities, back in print!
Seriously! Who hasn't played Edgar Allen Poe with their Sims character, walling them into a room to see how long it would take them to die. ...or was that just me?
Test subject Andrew "Ender" Wiggin was reported to say, "It took a while to master this VR, but I'm getting better. The simulated victim spills the beans 70% of the time now, but I want to try for 75%."
As copyright owner of this comment, I authorize everyone to defeat any technological measure which limits access to it.
Yes, they were told beforehand that they were shocking a computer program. Even so, they felt increased stress levels.
Now, is it still unethical?
Don't thank God, thank a doctor!
John Dean (former aide to Nixon) treats this, and more, in his book "Conservatives Without Conscience", where he helps explain the reasons so many people blindly follow authority (and why some people so like to be blindly followed). Milgram's work was seminal in the study of authoritarian followers, and you do it no justice by blatantly misrepresenting it.
At any ate, the point of this study is that some people do not emotionally differentiate between virtual actions and real actions.
"Trolls they were, but filled with the evil will of their master: a fell race..." -- J.R.R. Tolkien on Olog-hai
There's a big difference: Since the participants were well aware that the subject was a computer character this experiment seems to be basically about psychological/physiological responses from the participants, while the original experiment was much more interesting as people really believed they were hurting human beings.
That's why the original experiment, IMHO, is so important: because it exposed the risks of "obedience-without-thinking".
But then again, I have little knowledge about the whole thing, so these are just my impressions.
Er Galvão Abbott - IT Consultant and Developer
Speaking as a cognitive scientist, I can tell you that dealing with IRBs (internal review boards) for getting experimental approval gives you a different appreciation of the term "ethics" than might be understood by someone trying to live a morally sound lifestyle (for which, btw, I applaud you).
Essentially, in order for an IRB to approve a study to be performed using human subjects, one of two requirements must be met: either there is ABSOLUTELY NO RISK involved on the part of the subject (i.e. simple psychophysical tests of perception and so forth), or the risks must be outweighed by the potential gain in scientific knowledge that the experiment offers (i.e. clinical trials of drugs that, while risky, hold the promise of curing serious illnesses).
All studies require that subjects sign an IRB-approved consent form that enumerates the risks inherent in the experiment (or lack thereof), and IRBs require a submission of experimental intent and aim so that they may weigh the potential risks and benefits of the experiment. Some people would argue that *any* experiment is ethical as long as you can find subjects willing to be a part of them and sign a liability waiver, but the reality of the situation is that before any subject even has the option to sign their life away an IRB must first approve that such a situation would be ethically sound. Without IRB approval, you won't get published, and without being published, you won't get funding. That's the cycle.
Milgram's original experiments were deemed unethical because of the psychological trauma experienced by the subjects being ordered to up the voltage. They were put in the emotionally distressing situation of having to choose between following the experimenter's (i.e. authority figure's) orders and their own moral code, and this situation has since been deemed unacceptable. The reason for this is that the experiment's potential insights into the frailty of human morality in the face of authority simply weren't interesting or essential enough for the advancement of science to justify the risks of seriously traumatizing the subjects.
As far as I can tell, the reason this experiment is more experimentally justifiable is simply because the "victim" is explicitly virtual--a fact of which subjects are aware--so the situation, as it doesn't involve hurting actual people, isn't as emotionally traumatizing.
I left my wallet in El Sigundo!
Also, "ouch" and "touché."
But it's not mentioned in the "methodology" section, and I think the paragraph you mention does cast some doubt on the validity of the results:
"For those 12 in the VC who wanted to stop before the end, 5 claimed to be well-acquainted with the original Milgram study, and therefore we cannot rule out the possibility that this influenced their behaviour. However, if we treat 'wanting to stop' as a binary response variable in order to test for differences between the proportions (using binary logistic regression) then the VC was significantly different from the HC (?2 = 6.691 on 1 d.f., P = 0.0097) whereas knowledge of Milgram did not have a significant impact (?2 = 1.525 on 1 d.f., P = 0.22) and there was no interaction effect between group and knowledge of Milgram."
In the first place, this seems a little bit like throwing in a statistical fudge factor, since it does not say in their methodology that they planned to ask about knowledge of Milgram after the experiment, and they seem to have applied this statistical test a posteriori, whereas statistical tests are only valid if the test to be performed is stated in advance.
In the second place, it's all very well to say that five of the subjects "claimed to be well-acquainted" with the Milgram experiment, but that does not take into account the number of subjects that, while not well-acquainted with it, might nevertheless have had some vague or even subconscious knowledge of it. The Milgram study has been around a long time and is practically in the folkways.
There are probably millions of people who would say they knew nothing about John B. Watson's experiments with rats, who nevertheless would be extremely familiar with the idea of running rats through a maze.
"How to Do Nothing," kids activities, back in print!
Expanding on what you said, do the experiment with eight groups.
The first and second groups act as they did in this study.
The second and third groups act as the first and second, but with a man.
The fourth and fifth groups act as the first and second, but with a man, but of a different race (black subjects get a white victim, etc).
The sixth and seventh groups act as the first and second, but with a man they are told is an enemy combatant.
There are a lot of variations of this, and I doubt any of them are very ethical. But being unethical doesn't make the results uninteresting or invalid, but without a sufficiently large group, any results would be generally untrusted (but still interesting!).
And instead of becoming accustomed to the virtual person and ceasing to empathise, many volunteers became more anxious as the study continued. Measures of stress, such as heart rate and sweatiness of palms, increased. These measures are nearly impossible to fake, and confirmed for Slater that the volunteers were actually feeling uncomfortable, rather than performing as they thought the experimenter would expect.
I've got to wonder what the participants' exposure to video games or other "virtual environments" would have on their responses. To a gamer, I'm not sure rapid heart rate, and sweaty palms indicate increased anxiety. They might have just been "getting into the game."
I remember when Half-Life first came out my friend and I spent a lot of time running around beating the innocent bystanders with our crowbars and watching them beg for forgiveness. We weren't doing it because we were sadists, just curious gamers. We'd never seen NPC's react in such a realistic way before, and thought it was "cool". My girlfriend came into the room while we were doing this and was horrified, got really upset and asked us to stop. Not being as avid a gamer, I don't think she was used to dissociating her emotions from video game characters.
I don't think video game violence numbs players to real world violence, but it sure numbs them to video game violence. Seems to me like prior experience would play a major role in your reaction to this experiment.
Whether your "volunteer" has actually harmed someone or not, the psychological trauma is very real. That's the part where they describe the very real stress indicators. For those that don't know, the Nazi's kept free liquor flowing to the guards in the concentration camps. Why did they need liquor? Because of the emotional trauma associated with performing such vile acts on another human being.
It makes me wonder if the human subjects of this experiment truly trusted the statements of those in authority that they were NOT shocking real humans. Was something clicking in the backs of their heads warning them that they may be torturing real humans instead of electronic simulations?
Too bad Philip K. Dick is dead.
We have always been at war with Eurasia!
I agree.
This made me think of a personal antecdote. I don't know how many people played any games from the dungeon keeper series, but I used to play the first one a lot. One of the hallmarks of the game is that you're an evil character. However, as much as the ability to play a different persona appeals to me, every time I set out to play an evil character in any game I end up feeling remorse for killing innocents, even though they aren't real.
Anyway, in dungeon keeper (Real time strategy) you start out with a group of loyal imps. They are weak, small and do all of the mining and grunt work in your dungeon. They are unique in the game in that they can be created, and will always serve you no matter how poorly you treat them. The game allows you to slap creatures to make them work harder. There is little downside in doing this with imps since they won't flee the dungeon in anger and since they are poor fighters their health level isn't really important. Logically, all imps should be regularly slapped for maximum dungeon efficancy. And in fact, the computer controlled rival keepers do just that.
But I couldn't really do it as a matter of course. I actually felt bad, knowing full well that they weren't real. They made noises like they were in pain but of course thats just the computer. It was only when I was in a dire spot (doing a fast gem seam grab at the start of the map and then fortifying the walls to hold off an attack) that I would slap them, and even then I felt kind of bad.
So I can sort of understand how the results are similar to the original experiment. Its evoking an emotional response, and playing it again logic.
It's not the damage to the "virtual" victim that's considered the side effect, the Milgram experiment never damaged the "victim" virtual or otherwise.
It's the damage to the administrator of the virtual torture that's considered unethical. As long as that person has any common sense, they'll eventually discover that they are torturing the subject, and if they persist, they will be led to beileve that they have killed the subject. This should bring the administrator in conflict with their sense of morals, and the experiment is designed to break the virtues of compassion and human decency in a high-pressure situation.
We have all felt the pains of high pressure sales tactics, imagine the mistakes you could make if you were initially told that the test was harmless, then put under high pressure to continue it, eventually leading to the death (real in your point-of-view) of the subject? That's the Milgram experiment in a nut shell, and it relies on abusing the trust the administrator has in the researcher. The researcher deliberately lies to the administrator to achieve the desired results.
The fact that you're later debriefed won't console you because the experimenter lied to you, so they could lie to you about "it's all fake, after all". It also prevents resolution of the pain, because complaining about the stresses incurred while virtually torturing someone will either ostracize you from peers or be dismissed as non-real stress. If you actually tortured someone you could atone (if you wish to) for your actions by helping those you hurt, seeking forgiveness within your faith, or by deeds. But nobody understands redressing virtual wrongs; there's no avenue for repentance.
As a good Christian (in the best meaning of the phrase) can you condone the treatment of the real subjects, the ones administring the virtual shocks?
There's also other scientific grounds for dismissal. If the researcher deliberately manipulates the administrator to achieve the desired results, then is it science? In other "hard" science fields, manipulation of the experiment with a desire to achieve certain results would be a serious infraction of the scientific process, but in the near-voodoo corners of Psychology, it's considered a technique.
The news is that they've reproduced it. This one isn't nearly as reproducable as it claims to be, and the effect doesn't support what every Psychology student is told; that "You would do the same thing in the same situation." which (fortunately) isn't true according to the less fantastic failures to reproduce the same outcome.
People for the Ethical Treatment of Avatars.
When they gave the test to Donald Rumesfeld it took three techs to pry the button out of his hand. They said it was the giggling that was really creepy.
I consider myself as having more ethics than the average. I am a Christian (yeah, hold your slams, that's not the point). I try to live consistent with what Christianity teaches.
Funny. So were most of the original test subjects in Milgram's 1963 experiments. This stands an an irrelevant comment except to basically brag about how you feel morally superior to most people -- and then you have the sheer, unmitigated gall to ask that people "hold [their] slams." That's Pride. We are all sinners; remember that, and you'll do far better as a Christian than to parade around like a Pharisee waving your religiousity around like it's a badge, proclaming that you have "more ethics than average."
I don't see what's morally or ethically wrong with the experiment, even with a real human subject. I mean, the "victim" isn't actually being shocked, whether the "victim" is human or virtual.
The victims are the test subjects -- the people being pressured into harming other people in spite of their normal moral inclination to avoid such a thing. They are being put under stress and are being led to sincerely attempt to cause mortal harm to another to avoid the displeasure of an authority figure. They are caught between their conscience and the pressure to conform. In the end they are harmed in two ways: (a) they are put under immense stress, (b) they are led to commit a deeply wrong act that they would've never considered.
If tempting people to hurt others and causing distress and emotional turmoil (and in one subject seizures) aren't unethical in your worldview, then I think you need to hit the Good Book a little harder and work some more on those superior ethics of yours.
If it's for-profit but free, you're not the customer -- you're the product (e.g., the Slashdot Beta's "audience").
Seems that most of humans in the test were in fact UNETHICAL.
Interesting. You fall into the same trap as many of the psychologists who decried the experiments when they were published -- that the subjects must all be inherently flawed and that normal, good, decent people would never find themselves failing in the same manner. You ignore that in 1963, most of the test subjects were probably church-going Christians themselves. Even the researchers performing the experiment were convinced beforehand that no more than 1.2% of subjects would go all the way to deadly voltage.
In fact, 65% did. None stopped before 300 volts, and the shocks started at 45 volts and increased 15 volts each time. That means that a minimum of 16 shocks were delivered by every single subject. The experiment was stopped when the subjects either refused to continue or objected more than 4 times. The experimenters expected that most subjects would stop at 150 volts (8 shocks) and that no more than 4% would make it to 300 volts.
Let me repeat that again. Every single subject shocked people up to 16 times, well into the "dangerous" range on the dial. None of the subjects knew what they were getting into at the time. They were being led to believe that they were helping perform a test on memory of the "electocuted" actor.
Milgram got similar results in the 19 other experiments exploring variations on the theme. He found that increased emotional distance from the subject increased the willingness to go all the way while seeing and hearing the victim scream and writhe and having the authority figure be more remote decreased it. However, all replications of the original experiment got deadly shock results in the 61-66% range.
Now, what are the odds that 65% of Americans in 1963 were all rotten murderers and 100% were vicious torturers? Pretty bad if you believe that the populace isn't inherently ready to perform these kinds of acts without realizing it. The odds of getting so many latent torturers is pretty bad unless the percentage of latent torturers in the population is really, really high.
Quite frankly, you've never been tested like they have. You've probably never been put in a similar situation, and you have the advantage of knowing ahead of time to watch out for this sort of thing. Many of the subject expressed gratitude for having been in the experiment because it made them face what lies within them. At least one wrote back that when he was drafted for the Vietnam War, he was far more aware of what being put under someone else's authority would mean for him and how he was ready now to put the foot down if ordered to do something horrible.
Who knows, maybe you're actually more ethical than most, but I want you to remember your words every SINGLE time the temptation arises to cut corners to meet a boss' impossible deadlines, every time the temptation comes to make an excuse to the customer to avoid getting reported, every time the temptation comes to tell a white lie to avoid any form of disapproval from anyone for ANY reason.
Because the potential for evil lies within you too. It lies within us all.
If it's for-profit but free, you're not the customer -- you're the product (e.g., the Slashdot Beta's "audience").
As soon as I saw an electronic victim I would have just started jamming on the shock button as fast as I could until it died...just as an experiment. I would observe the response from the "authority". I'd note what he'd attribute my seemingly violent and malicious outburst to.
When he asks why I did it. I'll tell him it's the only way I can orgasm. [/Ichi]
I bet that'd throw his results off.
Milgram ran at least 20 experiments along this theme. The end result of variations in the base experiment revealed the emotional distance between the "teacher" and the "learner" had a very strong effect on the likeliness to continue. The more dehumanized the learner was, the more readily the teacher went further and further. Conversely, the more empathy the teacher was encouraged to have (say by seeing or directly hearing the learner through an open door instead of a speaker), the less likely they continued.
By demonizing the subject as a criminal, you would definitely observe a higher incidence of going too far. Demonizing your enemies is a central tactic in all societies committing to war for a reason -- it makes it easier to kill the other guy when you don't see him as being the same as you.
If it's for-profit but free, you're not the customer -- you're the product (e.g., the Slashdot Beta's "audience").
Mirror neurons are neurons that fire in response to what we see someone else experience. The other person picks up a pot, part of our brain that would be needed to pick up the pot fires as well. Pick up a hot pot without protection, and the person screams -- and part of your brain feels the burn. We're hard-wired for sympathy. The computer program is triggering mirror neurons in the human observer. I'd bet the researchers that wrote the computer program would have exhibited some signs of distress at the computer NPC being tortured -- even though they, better than anyone, would know that no human was involved.
This sig seemed like a good idea at the time....