Can “Terminators” Actually be our Salvation?
Just war theory defines the principles underlying most of the international laws regulating warfare, including the Geneva and Hague Conventions. Walzer’s classic book Just and Unjust Wars was a standard text at the West Point Military Academy for many years, although it was recently removed from the required reading list.
Asaro asserts that robotic technology, like all military force, could be just or unjust, depending on the situation.
h+: We’re using semi-autonomous robots now in Iraq and, of course, we’ve been using smart bombs for some time now. What is the tipping point – at what point does a war become a “robot war”?
PETER ASARO: There are many kinds of technologies being used already by the U.S. military, and I think it is quite easy to see the U.S. military as being a technological system. I wouldn’t call it robotic yet, though, as I think there is something important about having a “human-in-the-loop,” even if the military is trying to train soldiers to behave “robotically” and follow orders without question.
I think there is always a chance that a soldier will question a bad order, even if they are trained not to, and there is a lot of pressure on them to obey.
Ron Arkin is a roboticist at Georgia Tech who has designed an architecture for lethal robots that allows them to question their orders. He thinks we can actually make robots super-moral, and thereby reduce civilian casualties and war crimes.
I think Ron has made a good start on the kinds of technological design that might make this possible. The real technical and practical challenges are in properly identifying soldiers and civilians.
The criteria for doing this are obscure, and humans often make mistakes because information is ambiguous, incomplete, and uncertain. A robot and its computer might be able to do what is optimal in such a situation, but that might not be much better than what humans can do.
More importantly, human soldiers have the capacity to understand complex social situations, even if they often make mistakes because of a lack of cultural understanding.
I think we are a long way from achieving this with a computer, which at best will be using simplified models and making numerous potentially hazardous assumptions about the people they are deciding whether or not to kill.
Also, while it would surely be better if no soldiers were killed, having the technological ability to fight a war without casualties would certainly make it easier to wage unjust and imperial wars. This is not the only constraint, but it is probably the strongest one in domestic U.S. politics of the past 40 years or so.
By the way, I see robots primarily as a way to reduce the number of soldiers needed to fight a war. I don’t see them improving the capabilities of the military, but rather just automating them. The military hold an ideal vision of itself as operating like a well-oiled machine, so it seems that it can be rationalized and automated and roboticized. The reality is that the [human] military is a complex socio-technical system, and the social structure does a lot of hidden work in regulating the system and making it work well. Eliminating it altogether holds a lot of hidden dangers.
h+: Does robotic warfare heighten the possibility of accidental war, or might it guard against it?
PA: There was a news item March 2008 about a unit of the Swiss Army, about 170 infantry soldiers, entering into Liechtenstein at night by way of a dark forest. This turned out to be an accident –- they were lost during a training exercise –- so there wound up being no international incident. If there had been tensions between the countries, there could have been a just cause for Liechtenstein to declare war on Switzerland on the basis of an aggression.
Of course, Liechtenstein does not even have an army. But something similar happened in 2002 when a platoon of British Royal marines accidently invaded a Spanish beach, instead of Gibraltar.
I think the same is true of machines. They could inadvertently start a war, though this depends both on the technology malfunctioning and on the human political leadership desiring a war. Many wars have been started on false pretenses, or misconstrued or inadvertent acts: consider the sinking of the Maine in Havana or the Gulf of Tonkin incident.
h+: You talk about the notion that robots could have moral agency – – even superior moral agency –- to human soldiers. What military would build such a soldier? Wouldn’t such a solider be likely to start overruling the military commanders on policy decisions?
PA: I think there are varying degrees of moral agency, ranging from amoral agents to fully autonomous moral agents. Our current robots are between these extremes, though they definitely have the potential to improve.
I think we are now starting to see robots that are capable of taking morally significant actions, and we’re beginning to see the design of systems that choose these actions based on moral reasoning. In this sense, they are moral, but not really autonomous because they are not coming up with the morality themselves… or for themselves.
They are a long way from being Kantian moral agents –- like some humans –- who are asserting and engaging their moral autonomy through their moral deliberations and choices. [Philosopher Immanuel Kant’s “categorical imperative” is the standard of rationality from which moral requirements are derived.]
We might be able to design robotic soldiers that could be more ethical than human soldiers.
Robots might be better at distinguishing civilians from combatants; or at choosing targets with lower risk of collateral damage, or understanding the implications of their actions. Or they might even be programmed with cultural or linguistic knowledge that is impractical to train every human soldier to understand.
Ron Arkin thinks we can design machines like this. He also thinks that because robots can be programmed to be more inclined to self-sacrifice, they will also be able to avoid making overly hasty decisions without enough information. Ron also designed architecture for robots to override their orders when they see them as being in conflict with humanitarian laws or the rules of engagement. I think this is possible in principle, but only if we really invest time and effort into ensuring that robots really do act this way. So the question is how to get the military to do this.
It does seem like a hard sell to convince the military to build robots that might disobey orders. But they actually do tell soldiers to disobey illegal orders. The problem is that there are usually strong social and psychological pressures on soldiers to obey their commanders, so they usually carry them out anyway. The laws of war generally only hold commanders responsible for war crimes for this reason. For a killing in war to truly be just, then the one doing the killing must actually be on the just side in the war. In other words, the combatants do not have equal liability to be killed in war. For a robot to be really sure that any act of killing is just, it would first have to be sure that it was fighting for a just cause. It would have to question the nature of the war it is fighting in and it would need to understand international politics and so forth.
The robots would need to be more knowledgeable than most of the high school graduates who currently get recruited into the military. As long as the war is just and the orders are legal, then the robot would obey, otherwise it wouldn’t. I don’t think we are likely to see this capability in robots any time soon.
I do think that human soldiers are very concerned about morality and ethics, as they bear most of the moral burdens of war. They are worried about the public reaction as well, and want to be sure that there are systems in place to prevent tragic events that will outrage the public. It’s not impossible to try to control robot soldiers in this way. What we need is both the political will, and the technological design innovation to come together and shape a new set of international arms control agreements that ensures that all lethal robots will be required to have these types of ethical control systems.
Of course, there are also issues of proliferation, verification and enforcement for any such arms control strategy. There is also the problem of generating the political will for these controls. I think that robotic armies probably have the potential to change the geo-political balance of power in ways far more dramatic than nuclear arms.
We will have to come up with some very innovative strategies to contain and control them. I believe that it is very important that we are not naive about what the implications of developing robotic soldiers will mean for civil society.