I’m just back from the AAG Conference in Tampa, and there’s a lot to catch up on.

First, an art installation in Pakistan called #NotABlugSplat that reverses the paramilitary gaze and ‘targets Predator drone operators sitting thousands of miles away who refer to kills as BugSplats.’  Now they’ll see on their screens the face of a child who lost her parents and two young siblings in a drone strike.

It’s a collaboration between Pakistani and American artists, working with Reprieve and the Foundation for Fundamental Rights, who also designed it ‘to be captured by satellites in order to make it a permanent part of the landscape on online mapping sites.’

It’s an arresting project – but if you scroll through the comments that followed gizmodo‘s report you’ll see that ‘hope’ and ’empathy’ remain dismally distant for many people.

Perhaps some of them would benefit from a new report for the Center of Naval Analyses (CNA) by Larry Lewis, Drone strikes in Pakistan: reasons to assess civilian casualties.  I’ve noted his (largely classified) work on Drone strikes and civilian casualties in Afghanistan before, and in this – unclassified – report he leverages what we know about US military drone strikes in Afghanistan to address the cross-border attacks directed by the CIA.  Lewis makes two key points.

First, he notes that the US government’s claims about civilian casualties for its supposedly covert operations in Pakistan are significantly lower than ‘nearly every other estimate available’.  (En route, he draws attention to something that is usually overlooked in these calculations: under International Humanitarian Law, ‘the burden of proof is to determine whether a casualty is a combatant’, and where in doubt the casualty must be regarded and recorded as a civilian).  Based on his previous work in Afghanistan, Lewis suggests three overlapping reasons – apart from a disinclination to tell the truth – that ‘complicate the estimation process’:

• An irregular enemy –  it is exceptionally difficult to distinguish combatants from civilians in irregular warfare, and this is exacerbated by combatants ‘co-locating with the local population’;
• Misidentifications – ‘US forces mistakenly believe civilians to be enemy combatants’; I’m not sure how this is different from the first, but Lewis provides two examples that suggest he has in mind specific rather than general characteristics: mistaking men digging drainage ditches for militants burying an IED, for example, or assuming all those in close proximity to an engagement were involved (‘guilt by association’);
• Battle Damage Assessments (BDA) based on aerial surveillance – determining the consequences of an air strike without ‘boots on the ground’ is likely to be defective

Of all of these, Lewis suggests that it is misidentification that is likely to be ‘the basis for the majority of civilian casualty incidents’ and cites the case that I discussed in detail in Tampa: the strike carried out by two attack helicopters following persistent surveillance from a Predator of a ‘convoy’ of three vehicles in Uruzgan province in Afghanistan in February 2010.  I’ll post my version of events shortly, since I think it is a mistake to collapse this episode into a monotonic ‘Predator vision’; there were other eyes in the sky [see the image below], and – still more significantly – military vision is not a uniquely technical process (which is why the concept of visuality is so important) and in this case involved different interpretations of the Full Motion Video Feed from the drone by different people at different locations. In short, there was a de-centralized, distributed and dispersed geography of militarized vision that was never resolved into a plenary (still less totalizing) frame.

That said, Lewis’s second point is about process not platform.  He has no truck with claims like Avery Plaw’s – ‘Where civilian casualties cannot be avoided they must be minimized.  This is what drone strikes do’ – because they mistake ‘platform precision for a comprehensive process that minimizes civilian casualties’ and are in fact ‘contradicted by operational data’.  He cites his earlier analysis of 2010-2011 data from Afghanistan, which ‘showed that several forms of attack, including engagements by manned air platforms, were less likely to cause civilian casualties than drone strikes’ (my emphasis; see my earlier discussion here). In his view, then, ‘minimizing civilian casualties is less a matter of platform or ordnance selection as it is using an approach that considers factors that lead to civilian casualties and then effectively takes them into account.’

The point is sharpened by Mark Gubrud in a response to a report from Charli Carpenter at the Duck of Minerva:

‘…drones use the same targeting pods and precision-guided weapons as the manned platforms they replace; in fact, the quality of imagery from the drones is degraded by the limited bandwidth and frequent interruptions of satellite links, as well as the transmission delay which can frustrate last-moment aborts. On top of the “soda straw” vision as compared with low-flying aircraft in close air support, these factors mean that, if anything, the drones are actually inherently less discriminate.’

Again, all of these factors were in play in the Uruzgan attack: degradations and interruptions of both video and audio transmissions were of critical importance – again, see the image from my Tampa presentation above – but even more significant was the way in which the military field of view expanded, contracted and even closed at different locations as the episode unfolded.  More to come.