Because that interpretation is so reasonable, I was surprised by the results of a forthcoming research paper in Political Psychology. The author, James Igoe Walsh, conducted two survey experiments that bear on the story above and on our general understanding of the costs of using force in the modern era.
Experiment 1: Which costs matter most?
Walsh first constructs a baseline level of support for the use of force by asking respondents about their approval of a military operation against a terrorist training camp that will likely be successful, cause no U.S. casualties, and cause no collateral civilian deaths. He then varies the prompt, relaxing each of these positive outcomes in order to study which pitfall (or combination thereof) creates the greatest drop in average approval among respondents.
Students of international politics know instinctively that troop losses are politically costly, which is why we accept the logic of "tripwires" without much thought. Yet, in the most surprising finding, Walsh's respondents disapprove more strongly of an attack which risks a positive number of civilian casualties than they do of an attack which will likely cause 25 U.S. military deaths:
More interesting is that the difference from the baseline treatment [for the civilian casualty treatment] is larger than for the treatments introducing mission failure and American military casualties, and that these differences are statistically significant (p=.02 for the difference between military and civilian casualties, and p<.01 for the difference between mission failure and civilian casualties). This means that, in this experiment, civilian casualties from a drone strike creates the largest decline in the willingness to support the use of force.
So should we stop worrying that drones will allow for more careless uses of force abroad because the American public actually cares about foreign casualties? Aside from the author's responsible caveats (see p. 18 and conclusion), I will explain my own hesitation. First, if the public never learns how many civilian casualties are associated with a given operation, or that an operation even occurs, the executive may not be constrained by this kind of sensitivity. Still, to the extent that the executive believes that the press and NGOs will publicize or even inflate civilian casualties after the fact, a lack of transparency may not be sufficient to overcome this constraint.
Second, perhaps the use of U.S. troops abroad has historically been constrained not by a public averse to troop losses, but rather by a loss-averse inner circle of advisers. Many in that circle are in uniform themselves or are civilian heads of military bureaucracies (think of Bob Gates's memoir, in which the former Secretary of Defense wrote openly of his anguish regarding American injuries and losses). If the executive pays political costs primarily because he or she alienates other elites, a finding that the public is worried about foreign civilians is less reassuring because relative concern for U.S. troops may be overrepresented in the upper echelons of government.
Walsh's finding is still important and these issues were outside the scope of his project. I mention them only because the finding is so attractive to drone optimists that I don't want to advertise it uncritically.
Experiment 2: Weasel words or unintended constraints?
Walsh included another experiment to help us understand the implications of framing certain weapons platforms as more precise than others. Our first instinct is to believe that this framing could allow governments to launch more strikes and paper over the casualties that do occur while evading public disapproval. But what if a "precision" framing also strengthens the constraint on force uncovered in the prior experiment?
Respondents were read one of several attack plans for an impending operation. Roughly speaking, there were three varieties of attack plan, one of which was assigned to each person: high precision ("the chance of killing civilians is very low" with some technological justifications), moderate precision, and low precision. The respondents were told about the historical civilian casualty rate for the type of plan they were read, and asked about their support level for that plan. Next, regardless of the plan each respondent was given, each respondent was told that the attack was carried out, resulting in the deaths of both militants and civilians. So, to summarize, all respondents received the same outcome story even though they each received one of the attack plans differentiated by precision level.
It turns out that those in the "high precision" treatment group preferred higher levels of compensation to victims' families, by a statistically significant but not huge amount. Less surprisingly in my view, the drop in satisfaction upon learning of civilian deaths was largest for the high precision group as well (this might be explained by the greater initial satisfaction upon hearing a precision attack plan as opposed to a less precise one).
Interestingly, though, two variants of the high precision attack plan--one manned and one unmanned--yielded divergent results. Those who were read a drone attack plan expressed more regret about civilian casualties than those read a high-precision manned aircraft attack plan, judging by the extent to which respondents wanted to make costly compensation payments versus cheaper forms of apology.
So, an attempt to gain public approval for operations by declaring them to be precise can backfire if the attacks turn out to kill civilians, especially if the attacks are unmanned. As in the first experiment, drawing a line between this finding and a real-world constraint on the use of force depends on the assumption that the public can, in practice, learn about civilian casualties after strikes occur.
Walsh concludes his paper by noting that his results "are consistent with the proposition that the use of precision weapons makes individuals more sensitive to civilian casualties." I find the argument convincing but I encourage those interested to read the article for more detail.