Conflict simulation, peacebuilding, and development

Daily Archives: 11/01/2019

Lin-Greenberg: Drones, escalation, and experimental wargames


WoTRdrones.pngAt War on the Rocks, Erik Lin-Greenberg discusses what a series of experimental wargames reveal about drones and escalation risk. The finding: the loss of unmanned platforms presents less risk of escalation.

I developed an innovative approach to explore these dynamics: the experimental wargame. The method allows observers to compare nearly identical, simultaneous wargames — a set of control games, in which a factor of interest does not appear, and a set of treatment games, in which it does. In my experiment, all participants are exposed to the same aircraft shootdown scenario, but participants in treatment games are told the downed aircraft is a drone while those in control games are told it is manned. This allows policymakers to examine whether drones affect decision-making.

The experimental wargames revealed that the deployment of drones can actually contribute to lowerlevels of escalation and greater crisis stability than the deployment of manned assets. These findings help explain how drones affect stability by shedding light on escalation dynamics after an initial drone deployment, something that few existing studies on drones have addressed.

My findings build upon existing research on the low barrier to drone deployment by suggesting that, once conflict has begun, states may find drones useful for limiting escalation. Indeed, states can take action using or against drones without risking significant escalation. The results should ease concerns of drone pessimists and offer valuable insights to policymakers about drones’ effects on conflict dynamics. More broadly, experimental wargaming offers a novel approach to generating insights about national security decision-making that can be used to inform military planning and policy development.

You will find a longer and more detailed account of the study here.

This is a good example of using multiple wargames as an experimental method. Above and beyond this, it also shows how that wargames can generate questions worthy of further investigation.

More specifically, while the loss of a drone is less escalatory, an actor might be more likely to introduce a drone for this reason—possibly deploying one in a situation where they would not have risked a manned platform. If this is true, however, drones may still prove more escalatory overall. In other words, if the wargame is expanded to include the prior decision to deploy assets in the first place, the actual outcome might have been something like this:

  • Blue scenario 1: Deploy manned platform?
    • No, too risky.
    • No platform deployed.
    • Nothing shot down.
    • Result: No escalation.
  • Blue scenario 2: Deploy drone?
    • Yes, because no pilot at risk.
    • Drone shot down.
    • Result: Minor escalation.

Or, with regard to another situation—perhaps local air defences would have been reluctant to engage a manned aircraft because of the evident risk of escalation, but would happily shoot down a drone. In this case the experimental findings might have been:

  • Red scenario 1: Shoot down aircraft?
    • No, too risky.
    • Nothing shot down.
    • Result: No escalation.
  • Red scenario 2: Shoot down drone?
    • Yes, because no pilot at risk.
    • Drone shot down.
    • Result: Minor escalation.

In fact, if you read the full paper you will see this is exactly what occurred in a scenario involving a  shoot-down decision: participants were much more likely to use force against an unmanned drone.

In other words, while the study suggests that drones might reduce the chance of escalation, it also suggests that we also need to investigate whether the lower perceived risk of drone-related escalation might cause Blue to undertake more provocative overflights, or might lead Red to undertake more potentially escalatory shoot-downs.

Figure 1 below shows the main experiment: aircraft shoot-downs lead to major escalations, drone shoot-downs to minor escalation.


Figure 1: Experimental results suggest shoot-down of manned aircraft results in greater escalation.

Given the risk of escalation, however, decision-makers might decide against overflight in the first place.

Figure 2 examines a situation where no drones are available. It incorporates the possibility that decision-makers simply refrain from overflight because of the escalation risk, and assigns a (plausible but entirely made-up) probability to this. Moreover, knowing that a shoot-down of a manned aircraft is likely to cause escalation—a tendency noted by Lin-Greenberg’s other experiment—perhaps Red won’t actually open fire. Again, I have assigned a (plausible) probability to this. These numbers are just for the purposes of illustration, but here we note that with manned overflight as the only option there is a 16% chance of escalation.


Figure 2: Considering other decision points. Should Blue even send an aircraft, given risk of escalation? Should Red engage it, given the risks?

In this fuller model, now let us introduce drones (Figure 3). Given that they are less likely to cause escalation, let us assume that (1) Blue is likely to prefer them over a manned ISR platform, (as per earlier findings) (2) Red is more likely to shoot them down, and that (3) shooting down a drone causes minor rather than major escalation. Once again, I’ve assigned some plausible probabilities for the purposes of illustration.


Figure: Adding drones to the mix.

When we add drones into the mix, the risk of major escalation drops from 16% to 4%, but, the risk of some form of escalation actually increases to 60%.  Does this mean that drones have actually limited the risk of escalation, or increased it? Moreover, it is possible that tit-for-tat minor escalation over drone shoot-downs could grow over time to major escalation. If that were the case, it is possible that drones—rather than limiting conflict—are a sort of easy-to-use “gateway drug” to more serious problems.

Remember that I’ve essentially invented all of my probabilities to make a methodological point (although I have tried to make them plausible). My point here is not in any way to criticize Lin-Greenberg’s experimental findings—I suspect he is right. It is to say that the two sets of wargame experiments he undertook are useful not only for their immediate findings, but also to the extent that they generate additional questions to be investigated.



%d bloggers like this: