A common objection to shitting my pants is that there is a toilet eleven meters away. I agree that the toilet exists. My narrower claim is that its existence does not settle whether walking to it is the best use of the next four minutes.
The usual analysis compares two end states, clean pants versus soiled pants, and stops there. It ignores the path. Using the toilet means standing up, pausing whatever I am doing, walking, possibly waiting, performing the act, washing, walking back, and then rebuilding the mental context I had before I left. Programmers call that last cost a context switch. It is the largest term, and it is the one nobody counts.
Let the expected utility of shitting my pants be:
EU = T + F + Q − L − O − R
T = transit time saved
F = preserved flow state
Q = avoided queueing (we have one bathroom)
L = laundry and garment depreciation
O = olfactory externality
R = reputational loss
The neglected term is F. On Tuesday at 2:40 PM I was forty minutes into refactoring a permissions module for an HVAC-distribution dashboard, holding roughly nine variables in working memory. Research on interrupted knowledge work suggests it takes more than twenty minutes to get back to full depth after a break. That makes a four-minute bathroom trip a twenty-five-minute productivity event. My pants, by contrast, required no context switch at all.
Scott Alexander has observed that trivial inconveniences have large effects on behavior.
People are right to treat the walk to the bathroom as trivial. They are wrong to conclude that it is free. A cost that is small, certain, and paid several times a day compounds. Over a career it adds up to a meaningful fraction of my productive life spent in a hallway.
There is also a queueing problem. Our apartment has one bathroom and two adults. My girlfriend's morning routine occupies it for a period she has described as "none of your business." When demand arrives stochastically and capacity is fixed, waiting times get long.
Every additional trip I take raises her expected wait as well. Seen this way, shitting my pants is partly a congestion-reduction measure. I have not been thanked for it.
None of this is to deny L, O or R. The laundry cost is real, although it is mostly a machine cost, and the machine was going to run anyway. The olfactory term is harder to dismiss. My girlfriend has described the kitchen, where I work, as "a crime scene." But she arrived home three hours after the event, and it is worth separating the smell caused by my decision from the smell that would have existed anyway had I simply been a person who lives there. Charging the entire atmosphere of the apartment to one afternoon overstates the marginal harm.
The reputational term R depends on observability. I work remotely. My camera was on, but the relevant region was below frame. My colleagues learned nothing, which is the definition of zero reputational cost. People who insist that I "would have known" are describing guilt. Guilt is a feeling, and I am happy to add it to the model once someone gives me its magnitude.
I think much of the resistance here is Chesterton's fence applied too generously. The toilet exists, so we assume it must be used every time. But the fence was built for an average worker in an average building doing average work. It does not follow that every agent at every moment should route through it. The norm is a useful default. It is not a law of nature.
I am not claiming that everyone should shit their pants. My claim is only that "you're an adult" is a conclusion-shaped premise, and that once flow is priced honestly, the case is at least decision-theoretically open.

I tried to check the "meaningful fraction of my productive life spent in a hallway" claim. Say three trips a day at your twenty-five minutes each, 230 workdays a year, a forty-year career. That's 75 minutes a day, about 287 hours a year, 11,500 hours over a career. At 2,000 hours per working year that is nearly six years. Six years seems wrong to me but I can't find the error. All of it comes from the twenty-five-minute context switch. Using the bare four minutes it's under a year, which is still more than I would have guessed.
The error is applying the full resumption cost to every trip. A lot of trips happen when you're already between tasks, at a commit or after a meeting, and those cost close to nothing.
Few of mine do. That's partly why I wrote the post.
If the congestion argument is right then the socially optimal policy is for both of you to do it, at which point the bathroom sits idle and you've stranded the capital. I'd want to see the post deal with that. The logical endpoint is subletting the bathroom.
I'm not proposing a household policy. She's free to use it at whatever rate she likes, and as noted she does.
The twenty-minute number in the research is for being interrupted by somebody else. Here you choose the timing. You could have gone at the end of the function.
The urge chooses the timing. I don't.
You've got Chesterton's fence backwards. The principle is that you shouldn't take down a fence until you know why it was put up. You know exactly why this one was put up.
I know why it was put up for the average worker in the average building. That was the paragraph.
I solved this a different way two years ago. I moved my desk into the bathroom. Transit time is zero, there's no context switch, and nothing goes in my pants. My partner uses the downstairs one.
How's the wifi in there
Fine once I put a mesh node in the hallway. Video calls are a different problem.
You asked for a magnitude for guilt, so here's a proxy. What would you pay for nobody to ever find out? That's the reputational term internalized, and you can just put it in R.
About forty dollars. I've added it. The result still favors the pants by a wide margin at my hourly rate.
On R. Did you stand up at any point during the call? Asking because "below frame" assumes you stayed seated for the rest of it.
No. That was the point.
Has anyone run this past a gastroenterologist? I'm not saying it's wrong, I'd just like to know what "holding it" does to the F term over a long session.