One can safely say that the assumption of rationality is central to economics. One idea behind it that economic agents must be trying there best. But what if it were optimal to behave in an irrational way? Surely, one could build some twisted example where irrationality would be preferable, same some game theoretic environment where being unpredictable is superior. But this is still a rational choice, called mixing strategy. Would there be some truly irrational behavior that could at least weakly improve outcomes?
James Feigenbaum, Frank Caliendo and Emin Gahramanov come up with such an example. They use an overlapping generation economy where one can improve on the permanent income rule. Now, we know OLG models can quirky and lead to strange results, but this is still an interesting example. They key here is that there is a publicly shared rule of thumb, say, "save more now" that can alter aggregate outcome, in this example increase labor income and thus wealth. Households are not just price takers, but they consider that if everybody deviates, it will have consequences. All you need it to have some trigger that lets everyone converge on this new focal point. In short, animal spirits at work.