The fact that the kitchen bin is full does not establish that taking it out now is optimal. It establishes that we are approaching a capacity constraint. The distinction matters because disposal is irreversible, while waiting preserves the possibility of fitting something else in.
My girlfriend's preferred stopping rule is that the lid should close. This is easy to observe, but its relationship to the underlying quantity is weak. A bag containing a rigid salad container may prevent the lid from closing while retaining enough usable volume for several tissues, a banana peel, or the foil from a yogurt. These are ordinary forms of household waste. A model that rounds their disposal needs down to zero is systematically biased toward taking out the trash.
The remaining capacity need not be large. It need only be worth more than the marginal cost of keeping it available. In simplified form:
ΔU(wait) = p·C + B − O − L
Here p is the probability that the next discarded object fits, C is the value of accommodating it without opening another bag, and B is the benefit of combining disposal with a later trip downstairs. O captures additional odor exposure and L the expected cost of leakage. My girlfriend has discussed the last two terms at considerable length. Neither makes the first two disappear.
She points out that a new bag costs approximately nineteen cents. I accept this estimate. Small efficiency gains remain efficiency gains, and the principle scales across repeated decisions. It is particularly strange to hear people endorse resource conservation in the abstract and then become impatient when someone actually declines to throw away unused storage capacity in their kitchen.
The odor term deserves care. The relevant cost is the additional smell produced by waiting, relative to the smell already present. Much of the unpleasantness she attributes to my policy existed when she first asked me to take the bag out. Charging the entire smell to each subsequent hour would substantially overstate the cost of delay.
Leakage is a genuine tail risk. However, the bag has remained intact through three previous requests for removal. This is evidence about its load tolerance. I would update sharply on an actual failure; I see little justification for treating her increasing certainty that it will split as three independent observations.
I am therefore retaining the bag. If she values immediate removal more highly, she is free to perform it herself. That would satisfy her preference, but it would also destroy the remaining capacity before its value could be observed. I would not regard the subsequent absence of an overflow as evidence that her timing was correct.