Owen Biesel

Last month, this New York Times headline caught my eye for three reasons:

Netherlands Was 10 Percent Liable in Srebrenica Deaths, Top Dutch Court Finds

Reason #1: Having lived in the Netherlands, headlines about Dutch affairs usually stick out to me.

Reason #2: I’d just been reading The Themis Files, a fiction series in which the real-life slaughter of thousands near the town of Srebrenica makes a minor appearance, so (uncharacteristically for me) I knew some of the context already.

Reason #3: What the heck does “10 percent liable” even mean? I’ve heard people solemnly say “I take full responsibility for what happened” but never “I take ten percent responsibility”!

So I investigated. Here’s what I found:

Tied into that 10% figure are two interpretations: how the number was arrived at, and how it will be used. How it will be used is straightforward: the Dutch state is only required to compensate the victims’ surviving relatives for 10% of the total damages (however those are calculated). In that sense, any percentage liability is at least meaningful. But how that 10% figure was arrived at is a little more complicated.

The reason the Dutch state bears any responsibility at all is that an offshoot of the Dutch military, called Dutchbat, was in charge of protecting the Bozniak refugees as they were evacuated. Dutchbat was present on orders of the United Nations, but their decisions after the evacuation was ordered are legally attributable to the Netherlands. That’s why the Dutch State is on the hook for these two conclusions that their supreme court reached:

• The way that Dutchbat evacuated the refugees made it very easy for the Serbs to separate out the male refugees, and ultimately kill them.
• If Dutchbat had allowed the 350 male refugees who hadn’t already been noticed by the Serbs to stay at the compound, there was a 10% chance they could have escaped discovery and lived.

That 10% chance is where the Netherlands’ 10% liability comes from.

The simplest general principle I can come up to summarize the supreme court’s ruling is this:

“If there’s something you could do to lower the probability of something bad happening by X%, and you don’t do it, your liability for the bad thing is X%.”

The slaughter of the male refugees was (by the way the court was talking about it) 100% predictable given the choice to proceed with the evacuation, whereas if the 350 men had been allowed to stay behind, then their deaths would only be 90% likely, and 100% – 90% = 10% liability. I’ll call this method the Dutch liability method, although it may be in use elsewhere.

The Dutch liability method seems pretty straightforward, but let’s back up for a moment and ask ourselves what we’d want from a general method for distributing liability. It’s not clear how exactly to do so, but here are a couple of sensible conditions we’d want such a method to follow:

1. Once we assign percentage liabilities to the various parties, they should never add up to more than 100%!
2. On the other hand, if an outcome is completely determined by the actions of the parties, then the percentage liabilities should add up to exactly 100%.

These conditions sound plausible, but the Dutch liability method fails them both! Sometimes it assigns too much liability and fails condition 1, and sometimes it assigns not enough liability and fails condition 2. Here are some examples:

The sad shepherd’s pie

When we lived in the Netherlands, Clara and I did all of our baking in a tiny combination microwave and convection oven, a “combi.” If we wanted to bake the equivalent of a 9×13 shepherd’s pie, we had to stack two smaller baking dishes on a complicated rack assembly that caused the upper dish to bake very quickly, and meanwhile trapped the lower dish in place until the racks could be removed. We would usually pull out the top shepherd’s pie when it was ready, and leave the bottom one to bake a little longer while the combi cooled down. Unfortunately, that meant we’d occasionally forget about that second pie until the next morning, when it was irredeemably lost to the vagaries of time. It was very sad.

So which of us is responsible for the sad shepherd’s pie? Let’s say that each of us had just one moment where we remembered to take out the pie, but had a 60% chance of “leaving it to cool some more.” Then when it’s time to consider the consequences of my actions, if I didn’t put the pie away myself, there’d be a 60% chance Clara would also leave it out and it would be spoiled. If I did put it away, the chance would go down to 0%. That means the Dutch liability method calculates my liability as 60% – 0% = 60%.

However, the same reasoning also applies to Clara’s actions, so by the Dutch liability method she’s 60% liable too! And 60% + 60% is more than 100%, violating condition 1 for sensible liability assignment. Here’s a different scenario where condition 2 is violated:

The case of the mistaken Chipotle

Suppose you’re meeting a friend in a town unfamiliar to both of you, and you have agreed to meet up at the Chipotle (assuming there is one). However, when you made those plans, neither of you realized that there are not one, but two Chipotles, one on the east side of town and one one the west side, equidistant from both the train station and bus station where you and your friend will be arriving. You have no way of reaching your friend in time to make a new plan, and as far as you know, they have a 50/50 chance of showing up at either Chipotle. You pick one at random and hope for the best. If you don’t manage to meet up, what are your percentage liabilities?

In this case, the choices you and your friend make about which Chipotle to head to completely determine whether you meet up or not, so the second condition says your liabilities should add up to 100%. (This was actually true about the sad shepherd’s pie, too.) So let’s see what the Dutch liability method tells us: if you choose the east Chipotle, you have a 50% chance of missing your friend. However, if you had chosen the west Chipotle, the chance of missing your friend would go down to… 50%. That’s a 0% difference! So the Dutch liability method says you and your friend are each 0% liable, which doesn’t add up to even nearly 100%.

Conclusions

So the Dutch liability method fails on a couple of basic counts: sometimes ascribing too much liability, and sometimes too little. And in neither situation did it pick up on on the symmetry of the situation, since presumably the two people involved should “go Dutch” and split the 100% liability equally between them. So, really, what does the 10% probability in the Srenbrenica case have to do with the Dutch state’s percentage liability?

On the other hand, it’s not clear what a better system would be. It’s even not clear whether it always makes sense to try to divide up liability between everyone involved, instead of saying “the group as a whole is responsible.” When I was getting Clara’s permission to tell the sad shepherd’s pie story, she even said it felt fine because, “it wasn’t my [Clara’s] mistake, it was our mistake.” Could it be that some group liability just isn’t reducible to individual liability?

And maybe in both cases, focusing on the in-the-moment decisions misses the bigger picture of what other options there are. The two friends could have made a less ambiguous meetup plan, and Clara and I no longer rely on just remembering to put away things we’ve left out to cool. (Instead, when we make the decision to leave something to cool, one of us puts a post-it note or oven mitt or something on our pillows to remind us just before we go to sleep.)

What are your thoughts? Are there other ways of interpreting the Dutch supreme court’s ruling that generalize more sensibly? Any other thoughts on doling out blame? I’d love to hear about it in the comments!