Regarding the “Repent” example: as conformists, human beings are more likely to make particular decisions (like wear a “Repent” sign) if they believe others would do the same. So instead of framing this study as showing that “sign-wearing volunteers overestimate the probability others would volunteer”, one could flip the implied causality and say “people who think others would volunteer are more likely to volunteer themselves”, a much more banal claim. One could test the effect by re-running the experiment on self-identified nonconformists, or using behaviors for which conformity is not believed to play a big role. I predict the False Consensus Effect discovered in those settings would be much weaker.
The blue/red ball analogy is good food for thought, but there are way too many differences between it and the “Repent” study for the numerical similarity to be considered anything more than a coincidence. Our approximations of other people’s behavior are much more elaborate than making a naive inference based on a sample of one.
I wonder how long-lasting this “quota” effect is. The study only looked at the immediate effects of moral behavior, not the more important long-term effects.
To make an analogy with physical exercise, maybe flexing your moral muscles exhausts your ability to be moral for the rest of the day, but when you wake up tomorrow your moral strength will be not only restored but actually strengthened. Most forms of exertion I can think of (e.g. learning, writing, working) work like this, so I wouldn’t be surprised if the same held for doing good deeds.