My roommate says I keep stealing his groceries. I think this is a good example of how ordinary language can smuggle a surprising amount of moral content into what presents itself as a factual description.
I am not contesting the physical sequence of events. Last Tuesday, he bought a package of Greek yogurt, put it in the refrigerator, and I later ate it. The same basic pattern has occurred with some rotisserie chicken, several cans of LaCroix, half a container of raspberries, and, most controversially, a package of prosciutto that I did not realize had been purchased for a date. I am happy to stipulate all of this. What I object to is the immediate promotion of these events into the category “stealing,” because once you allow that word into the premises, most of the interesting part of the disagreement has already been decided.
When rationalists talk about tabooing words, the point is not that ordinary language is useless. The point is that certain words compress together a large bundle of observations and conclusions, and sometimes the compression prevents you from noticing that you are relying on the conclusion to justify itself. “Stealing” is almost maximally bad in this respect. It simultaneously means that one person possessed something, another person took it, the first person did not consent, the second person behaved wrongly, and some form of restitution or condemnation is appropriate. If I accept “you stole my yogurt” as the neutral description, there is not much left for us to analyze.
So I suggested that we temporarily avoid the words steal, theft, mine, yours, property, permission, and asshole, and instead describe the situation at the object level. My roommate claimed that removing the word asshole made the exercise “obviously rigged,” which I thought was revealing.
At the object level, the yogurt case looked like this. At approximately 11:45 p.m., I was hungry. There was food in a refrigerator that both of us use. I consumed one unit of that food. My hunger decreased. The available inventory in the refrigerator decreased. My roommate later learned that the unit was gone and became angry because he had expected to consume it at some future time. This is still a genuine conflict. What disappears is the assumption that one particular resolution of the conflict is encoded into the nouns.
My roommate's view is basically that because he paid for the yogurt, the decision rights over it belonged exclusively to him until he explicitly transferred them. This is certainly a common convention, and I understand why it exists. But the fact that a convention is common does not mean that every application of it is welfare-maximizing. There is no physical feature of the yogurt that changes when he scans his credit card at Trader Joe's. It does not acquire a detectable “Jonathan-ness.” If he leaves town for three weeks, the yogurt does not become less edible to me. What has changed is that we have socially assigned him a claim over the object.
Again, I am not saying claims are meaningless. I am saying they are a coordination technology, and coordination technologies can be evaluated.
The most obvious thing missing from his account is spoilage. My roommate buys food aspirationally. He frequently purchases groceries for a healthier, more organized version of himself who seems not to survive the drive home. There is usually spinach. There are often berries. There was once kefir for almost a month. At least some fraction of the food he regards as categorically unavailable to me will eventually be thrown away by him. I estimate that, conditional on an item remaining untouched for four days, the probability of eventual disposal is somewhere around 15 to 25 percent, depending on the item.
He says I invented those numbers. I did, in the sense that I estimated them rather than deriving them from a published dataset. I also offered to start keeping records, which he described as “somehow worse.”
But the underlying point survives considerable uncertainty. If I consume food that otherwise had some probability of being discarded, then the relevant counterfactual is not always “Jonathan eats yogurt, versus I eat yogurt.” Sometimes it is “I eat yogurt, versus yogurt enters the trash eight days later while Jonathan says he forgot it was there.” The second outcome does not become socially optimal merely because nobody violated a property convention on the way to producing it.
There is also a revealed-preference issue. In the yogurt case, the container had been in the refrigerator for four days while my roommate chose to eat two breakfast burritos, leftover Thai food, cereal, and a frozen pizza. He now says he had been saving the yogurt for Wednesday morning. That is possible, but this is exactly the sort of preference report that becomes difficult to evaluate after the resource has been removed. Ex ante behavior placed the yogurt below several competing breakfast and dinner options. Ex post testimony suddenly placed it near the top of the utility function.
I do not think he is lying. People routinely have future preferences they have not yet acted on. I merely think the strength of his claim should be discounted relative to the certainty with which he presents it.
The standard proposed solution is that I should ask before eating something. This sounds obviously reasonable until you actually inspect the mechanism. Suppose it is midnight and I text, “Can I have this yogurt?” He now has to notice the message, identify the yogurt, recall whether he has plans for it, decide whether those plans outweigh my current hunger, and then either say yes or incur the mild social cost of saying no. Asking is not free. It externalizes the allocation decision onto the person least likely to have immediate information about my hunger.
In many cases, unilateral allocation is simply cheaper.
My roommate asked why, under this theory, he cannot unilaterally allocate my beer to himself. This is a fair question, but the apparent symmetry disappears under closer inspection. I buy fewer groceries than he does and tend to buy them close to the time I expect to consume them. My beer therefore has both lower spoilage probability and higher demonstrated near-term demand than his average refrigerated item. The fact that two actions can both be summarized as “one roommate consumed something purchased by the other” does not mean they are equivalent at the object level.
He accused me of creating an economic theory in which every difference between us happens to favor me. I agree this would be suspicious if the differences were invented for that purpose. But it would be equally suspicious to ignore real asymmetries merely because they are convenient to one party.
The prosciutto case is harder. I concede that. It was relatively expensive, unopened, and apparently connected to a specific plan: he was going to cook dinner for a woman he had been seeing. I did not know this when I ate it. From my epistemic position, it was just unusually good ham that had remained untouched for several days. Had I known about the date, I probably would not have consumed it, because earmarking increases expected replacement cost and sharply lowers the probability that the item is surplus.
Unfortunately, after discovering the prosciutto was gone, he decided he “couldn't deal with going back out” and canceled the dinner. He later implied that this may have contributed to the relationship ending. I think attributing the full counterfactual value of a six-week romantic relationship to a package of cured meat is clearly excessive, although I accept that the incident had greater downstream costs than I originally modeled.
This is one reason I have proposed replacing our current informal system with something more explicit. Food could have a short exclusivity period after purchase, with longer protection for unusually expensive or precommitted items. Perishables that remain untouched past a specified window could then enter a common pool unless marked otherwise. This would preserve most of the benefits of personal ownership while reducing waste and eliminating ambiguity about items that have effectively been abandoned.
My roommate rejected this proposal and instead purchased a small refrigerator for his bedroom.
He has also installed a lock.
I regard this as a less elegant mechanism, but I concede that it has dramatically improved allocation clarity.