To ask the obvious question: how do they verify the videos are genuine, unique, recent, taken by this unit, show what they claim to show?
Since the drones are centrally produced, they could easily implement digital watermarks for provenance
Perhaps. Of course there will be high tech ways to bypass this. And some of the footage isn't taken on centrally produced drones, but on random personal smartphones.
But if your watermark well enough, then maybe you can get genuine videos of cardboard tanks blowing up.
Remember, decoys are often used in war. And they have to fool the people on the ground. So you get people thinking "well it's obviously a decoy tank. I saw a gust of wind blow it about. But we could probably make it look real if we shoot it at the right angle. "
Uniqueness of video would be easy to do by digital checks. Uniqueness of event is harder. But if you have accurate time and at least a rough idea of location, then seeing 2 tanks blown up in the same second should raise alarm bells. Uniqueness of tank is even harder. A tank can be shot at, duct taped up a bit (by either side) and then shot at again.
I think current Russia-Ukraine war is a perfect place to implement such system. It's an attrition war, there is not many goals which are not reduced to "kill and destroy as many as you can". There is a strategic aspect: Russia pays exorbitant compensations to families of killed soldiers, decades of income for poor regions. So, when Russian soldier dies, two things can happen:
I can easily see how such system could fail terribly in, say, Afghanistan (if you paid for every killed terrorist, there is an easy loophole "kill civilian, say they are terrorist"). It's fine for current stage of Ukraine war.
Also I don't see how kill markets contribute to ability of military to coup. Payments are made in purely virtual points, soldiers can't spend them on something else.
I agree, also I think the concerns of OP don’t seem well justified. This seems like a straightforwardly good system, modulo some obvious but probably avoidable goodharting risks
The incentives are worse than that. They aren't rewarding kills, they are rewarding videos-of-kills.
So, if the video cameras aren't rolling, the incentive might be to not kill until you can get the camera set up.
Also, drone footage cuts out when the drone hits the target.
Decoys exist. So does smoke. And equipment has a pretty continuous scale from fully functional to fully broken. This rule incentivizes wasting ammo blowing up a tank that's already basically scrap.
And if it's videos that are being used as proof. Well AI videos aren't quite there yet, but probably will be soon. I wouldn't expect kill markets to work as easily in a world where AI can easily generate fake kill footage.
Also, drone footage cuts out when the drone hits the target.
Yes, for small, explosive drones. But the Vampire drones mentioned in the article are much bigger and are usually used to drop an explosive payload that detaches from the drone and drops onto the enemy.
(Otherwise I agree with your point)
This works probably better for the drone units than the infantry (for instance).
Specifically, the policy of sending drones to the units that confirm the most kills in a way that is really hard to fake (the video, and the fact that lying here would result in punishment (obviously)) is a regular logistics policy.
This is just doing 3 things. The first is making the requisition game more legible to individual soldiers (for NATO style militaries this is very good), because this policy of supply priority and flexibility focusing on the most successful units is not a new system (it is built into the actual practice of professional militaries). Secondly it probably results in better data, because now they have a data pipeline for it. Third is that it affects morale, because all military communication also does that.
A Vampire drone, for example, costs 43 points.
I think you could do better with some kind of auction, which would let you identify the most effective weapons you should increase production of.
Prices would adjust to match supply and demand as well as acting as both supply cost and demand value signals. If no one buys the vampire drone, supply side stops production and starts dropping price to liquidate inventory, possibly with a liquidation auction.
Badly done dynamic pricing and auctions feel awful to market participants and can result in issues seen in Ebay auctions like sniping.
Combining kill markets with mis-optimized (incompletely specified, etc) RL agents is the stuff of nightmare.
Logistics push generally works better than logistics pull. Also, offensives generally involve seizing territory and forcing retreats, where digging in as you seize space may be more important than destruction.
Also, it suggests a few other bad things.
One possibility is that they don't have the integrated command necessary to identify things like areas that are under heavy attack (and therefore need to spend more ammunition defending them).
Another is that their training programs are ineffective or non-existent, if one drone operator is substantially better than another, it means that operators are not interchangeable. This is undesirable, because it complicates deployment. Most specialists in most military occupations are intended to be interchangeable with each other. If they're not, either the skill they have was acquired outside of training, or the training pipeline produces an irregular product.
Also, this is probably not great for morale if units are directly competing with one another for resources. Sabotaging other units' performance is a way to get yourself better supplies.
So what we see is a military trying an innovative way of figuring out how to allocate scarce resources to units trying to defend fixed positions by having them directly compete with one another. This implies that they're not expecting to seize much territory, that their unity of command is falling apart, and that their logistics are breaking down. Plausibly, it could also mean that some units simply are not fighting and that isn't easily observable using sny other system.
Yikes.
The idea of kill markets are of course not so new. One prominent example was the practice of paying bounties for Indian scalps which was practiced by various US government and white civilian entities during frontier conflicts against natives. They were not the first to adopt this practice but they were the ones to spread it across the continent. This incentivized wholesale extermination of native nations, and predation on innocents, rather than mere military submission, a common outcome of those frontier wars.
I found your post fantastic and very thought provoking. Wrote an adjacent piece here. Just wanted to say thank you for the inspiration.
https://www.prue0.com/2025/05/05/algorithmic-kill-markets-and-reward-hacking/
Politico writes:
Now, this is clearly a repugnant market. Repugnant market is a market where some people would like to engage in it and other people think they shouldn’t. (Think market in human kidneys. Or prostitution. Or the market in abortions. Also, watch a lecture about repugnant markets by Al Roth, who introduced the concept of repugnance into economics.)
Having acknowledged that, let’s put the feeling of disgust aside and try to think about the consequences of this new approach to war clearly.
First, let’s contemplate a bit on the raw, brutal efficiency of the markets. Cowen’s and Tabarrok’s course in microeconomics begins with a story that shows how market incentives beat the widely cherished moral and religious principles hands down:
Survival rate went from 66% to 99%. In other words, if appeals on patriotic feelings and on civil duty fail to make the army fight well, market incentives will likely do much better.
One could even argue that their effect will be even more pronounced than it was with the prisoner transports. If a ship owner wanted to improve conditions for the prisoners being transported, all it took was an order for the crew to treat the prisoners better and a willingness to absorb the associated costs.
With military it is different. Armies are big, heavy organizations that the commanders are trying to steer. They exhibits what Clausewitz calls “friction”: You issue a command and nothing happens. Or, maybe it happens in a partial, half-hearted way. There’s an inherent incentive not to fight. Nobody wants to get killed after all.
Introducing a market system, on the other hand, allows the lower-level units to take calculated risks. Destroy that many enemy units and you can buy, say, an armored vehicle, that improves your safety. Friction gets greatly reduced.
All in all, if the Ukrainian experiment succeeds, there’s a non-trivial chance that the kill markets will become the most deadly military technology of our time.
And the appeal to Ukrainians is easy to see: This is an asymmetric technology, one that the Russian army would have difficulty adopting, for the very same reasons the Russian state can’t truly embrace free markets. Too much top-down control breaks the markets, make them serve the vested interests, allows for corruption and generally drives their efficiency down to the ground.
All of that being said, kill markets are far from a flawless instrument.
Once you start optimizing for kills, Goodhart’s law kicks in and it turns out that you are no longer optimizing for winning the war.
Read your Clausewitz again: The ultimate objective of any war is political, not military. You are not trying to kill as many enemy soldiers as possible. You are trying to gain territory, shield your citizens from enemy hostilities, or whatever your political goal is.
But you can’t build your political goals directly to the incentives for the individual fighting units! You have to introduce a proxy goal and if you get it wrong — which is all too easy — the machine can run amok, not only causing a lot of unnecessary suffering, but also not contributing to your war goals.
But there may be an even bigger problem at hand: How do you control the army?
Kamil Kazani has just published a great post on the Wagner rebellion, which drives the point home nicely:
And it’s not a problem specific to Russia. We are seeing a military out of control in Myanmar right now. We can also easily find many such examples in Africa and Latin America, either now or in the recent past.
So, if you make the top-down control of the army looser by introducing market mechanisms, if you give individual units more power over what they can do, you are also increasing the danger of the army (or a part thereof) becoming an independent political entity, with its own resources and its own goals, struggling for power, just like what happened back in 2023 with Wagner Group.
Finally, if we take the possibility that kill markers will become an extremely deadly military technology seriously, is there a way to ban it?
I am skeptical. Bans may sometimes work where the impact of a technology is fully symmetric. No side gets an advantage when both sides use it. Then there’s no point in using it in the first place.
But kill markets do not feel like that. The side that is able to execute better — and there will be huge differences, just consider how well are free markets managed in different countries — will get an upper hand.
Moreover, a working ban requires enforcement mechanisms, the ability to detect when your adversary is breaking the rules. But markets, unlike, say, chemical weapons, are not directly visible on the battlefield. Each side would suspect the other of using them despite the ban and might try to secretly use them as well.