Can we justify the rise in military robots?

Tech­no­lo­gies for mil­it­ary robots have made sig­ni­fic­ant pro­gress over the last two dec­ades, rais­ing issues about using autonom­ous robot­ic sol­diers for act­ive engage­ment in battles. What are the eth­ic­al concerns?

Alan Wag­n­er. There are pros and cons. Firstly, since robot­ic sol­diers don’t get emo­tion­al, revenge­ful, or angry they would – in the­ory – fol­low the rules of war very closely. This could pre­vent some of the worst atro­cit­ies that have occurred in war­time. In that sense, robots could poten­tially be more eth­ic­al than human sol­diers. How­ever, the coun­ter­ar­gu­ment is that, cur­rently, robot­ic sys­tems are gen­er­ally not cap­able of dis­tin­guish­ing between civil­ians and sol­diers. As a res­ult, there is a risk that robots would acci­dent­ally tar­get civil­ians. That being said, these two argu­ments are not mutu­ally exclusive.

The pos­sib­il­ity of an acci­dent raises ques­tions around respons­ib­il­ity and liab­il­ity; this is the core of the eth­ic­al debate in its cur­rent ver­sion. One of our val­ues when it comes to mil­it­ary decision-mak­ing is that a human is respons­ible for a decision.

The pos­sib­il­ity of an acci­dent is at the core of the cur­rent eth­ic­al debate. 

But respons­ib­il­ity is an extremely dif­fi­cult notion when it comes to mil­it­ary robots. If a com­mand­er author­ises an autonom­ous sys­tem, is the com­mand­er still respons­ible for its course of action? If the sys­tem makes mis­takes, how long does the author­ity per­sist? Over a fixed peri­od? Or only regard­ing cer­tain actions? These ques­tions need to be con­sidered more strongly, but also codi­fied, to decide what the lim­it­a­tions of these sys­tems are and to determ­ine their bound­ar­ies with regard to ethics.

Defin­ing respons­ib­il­ity and author­ity is a leg­al point, one that could be dealt with based on a set of rules. But there is also a philo­soph­ic­al prob­lem: is the pro­spect of flesh and bone sol­diers facing blood­less machines acceptable?

It comes back to our val­ues and belief sys­tems. The ques­tion is not just about how unfair it would be for a sol­dier to face some Ter­min­at­or-like unstop­pable machine killer. And if the value of both your mil­it­ary and soci­ety is such that only a human can decide to take anoth­er human’s life in a mil­it­ary con­text, then that would pre­clude the use of autonom­ous sys­tems for most battles or oth­er mil­it­ary operations.

But debat­ing in such abso­lute terms is sim­pli­fy­ing the eth­ic­al ques­tion. You might have a value sys­tem which favors max­im­ising the safety of your sol­diers. In that case, you may require autonom­ous robots in your mil­it­ary. Val­ues are often in con­flict with one anoth­er and there might be a trade off. The prin­cip­al value for most coun­tries is to not lose a war, because con­sequences are high, not just on the bat­tle­field but for soci­ety on the whole. This leads to a dif­fi­cult chal­lenge: if one coun­try is going to devel­op autonom­ous sys­tems that have no eth­ic­al val­ues, but give them a stra­tegic advant­age, are you required to do so as well in order to not let them have that advantage?

Con­versely, there is also the ques­tion of legit­im­acy. If you win a battle thanks to robots, will your adversary accept your vic­tory? Will you be able to really make peace and put an end to the war? This is a key ques­tion, though it goes unnoticed in eth­ic­al debates over mil­it­ary robots. And unfor­tu­nately, we’re walk­ing right into it. Con­sider the United States use of drone war­fare in Iraq. Evid­ence shows that when sol­diers weren’t at risk, the num­ber of drone attacks by the United States went up; sug­gest­ing that when people are not at risk it could be easi­er to foster wars, with more battles. On the oth­er hand, in the recent Armeni­an-Azerbaijani war, the use of drones may have ended the war end more quickly.

In the past, mech­an­isa­tion of war­fare has made it more costly and blood­i­er, before a reversal. Could it be the case with robots?

It’s not clear wheth­er robots will make war­fare blood­i­er. They could make it less bloody if the autonom­ous sys­tems are well developed. Many years from now, autonom­ous sys­tems could become per­fect at tar­get­ing, com­pletely avoid­ing civil­ian cas­u­al­ties. Thou­sands of lives would be saved. There­fore, one has to be care­ful about the very notion of “killer robot.”

Thou­sands of lives could be saved, so one has to be care­ful about the very notion of killer robot.

We might not like pre­ci­sion guided mis­siles, but they are a replace­ment for car­pet bomb­ing. The same happened in civil­ian indus­tries such as agri­cul­ture, where after one cen­tury of mass using of fer­til­isers, we are switch­ing to a pre­ci­sion mod­el. “Sur­gic­al strikes,” an expres­sion used in the 1990s, was chal­lenged as just anoth­er pub­lic rela­tions motto. But the under­ly­ing trend, which is quite con­sist­ent with our value sys­tems, is that we kept devel­op­ing tech­no­lo­gies that would min­im­ise civil­ian cas­u­al­ties. The 90s were the begin­ning of pre­ci­sion war­fare, with mostly pre­ci­sion guided mis­siles. Things have advanced: we have pre­ci­sion recon­nais­sance and capa­cit­ies for pre­ci­sion assas­sin­a­tion, with long dis­tance guns able to kill just one per­son in a car.

It is a dif­fi­cult trade off to know wheth­er we should have these tech­no­lo­gies versus do we have the wars that might res­ult without them. The slip­pery slope argu­ment says it might become a battle of who con­trols these tech­no­lo­gies and the engin­eers able to devel­op them. But anoth­er argu­ment is that if heads of State and oth­er key decision makers can be tar­geted, pre­ci­sion war­fare can per­tain to the same dis­sua­sion logic as nuc­le­ar bombs, invit­ing all sides to show restraint.

Does the pro­spect of arti­fi­cial intel­li­gence alter these considerations?

The way arti­fi­cial intel­li­gence and robot­ics relate is that the robot is the machine, includ­ing the sensors, the actu­at­ors, and the phys­ic­al sys­tem. The arti­fi­cial intel­li­gence (AI) is the brain that makes the machine do things. They are highly con­nec­ted: the smarter the sys­tem, the more cap­able it is. But AI is a vast field, encom­passing everything from com­puter vis­ion and per­cep­tion to intel­li­gent decision-mak­ing to intel­li­gent move­ment. All these things could go into robot­ic sys­tems and be used to make them more cap­able and less prone to flaws and errors. Here AI is enabling pre­ci­sion, which rein­forces the argu­ments above.

Can AI replace people in this con­text? Again, the answer is not simple. AI may replace the per­son for some decisions, non-leth­al ones, or all decisions, or just some of the time, but not all of the time. We are back to the leg­al bound­ar­ies and liab­il­ity issues.

What it really changes is the stra­tegic para­met­ers of the decision. We are talk­ing of kin­et­ic war­fare here. When using, for example, drones to lead a charge, you risk much less than when you charge with sol­diers. Drones are just man­u­fac­tured items, easy to replace. You may nev­er lose the momentum if you can get it, which is a game changer stra­tegic­ally. You could ima­gine a battle where you drop a bunch of robots in to con­trol a bridge and they do it for years. They don’t fade or sleep. They just sit there, and nobody crosses the bridge unauthorised.