The sunk cost fallacy has always seriously bothered me.
For example, in her Checklist of Rationality Habits, Anna Salamon writes:
When facing a difficult decision, I check which considerations are consequentialist - which considerations are actually about future consequences. (Recent example from Eliezer: I bought a $1400 mattress in my quest for sleep, over the Internet hence much cheaper than the mattress I tried in the store, but non-returnable. When the new mattress didn't seem to work too well once I actually tried sleeping nights on it, this was making me reluctant to spend even more money trying another mattress. I reminded myself that the $1400 was a sunk cost rather than a future consequence, and didn't change the importance and scope of future better sleep at stake (occurring once per day and a large effect size each day).
This makes logical sense, but if I were the one who had bought the mattress, and Anna Salamon was in the same room as me trying to convince me to buy a new one on these grounds, I would start freaking out and gibbering (actually, this describes my interactions with Anna so far pretty well). This is despite usually being pretty good at internalizing this kind of reasoning, so I started wondering whether there was something flawed about this particular line of thinking.
The first obvious flaw, one that everyone always brings up, is that it only works if you have so much spending money that the loss of $1400 doesn't make a big difference to your overall financial situation either way. That is, if you start with $2800 in savings, you might be willing to spend $1400 on a mattress, but after you've bought one mattress and it's not very good, you only have $1400 left, and you might want to keep that money so that you have some savings left for staples and emergencies and so on. This is a good argument, but it doesn't apply to things that cost much less, and I have the same instinctive reaction to those too.
So here's a second argument I've never heard before: buying one mattress and hating it provides important information both on average mattress quality and on your ability to choose good mattresses. When you bought your first mattress, you might have naively thought a mattress was a mattress and couldn't possibly fail at mattressing. Or you might have thought that you know what you want and you're smart enough to choose the best mattress in the catalog. Now you have a data point showing that's not so. Apparently choosing the right mattress is pretty hard, and whatever your previous mattress-choosing criteria were aren't good enough.
So you realize that you're not just purchasing a new mattress for $1400. You're purchasing an X% chance at having a new mattress, and a (1-X)% chance of getting it, realizing it's useless, and throwing it away. You started off believing X was nearly 100%. Now you know better. Maybe now you think X is more like 50%. That means in order to purchase the second $1400 mattress, you would have to value a good mattress at $2800, not $1400. This means it may be rational not to get the second mattress.
And here's a third argument, that probably says more about my own thought processes than it says about rationality: not buying the second mattress is a way of punishing yourself for screwing up the first one.
I am both lazy and cursed with a strong tendency to self-punish, but these two flaws cancel each other out to a degree. If I knew that every time I bought a bad product, I would just shrug and spend even more money to buy a second version and see if that one worked, then that confidence would make me careless when ordering something: shopping is no fun and if I buy the first thing I see I can move on to something more interesting. But when I know that I'm sticking with what I get, I am more careful; probably right around the right amount of careful.
This might be a just-so story explaining something that really comes from my desire to signal to my parents and my friends that I am financially responsible and not the sort of person who will throw away $1400 on a mattress, then decide I don't like it, then throw away another $1400 on another mattress. But I think it also has some truth to it.