Is it time to reconsider the fundamental attribution error (FAE)?

Pick up any introductory psychology text, turn to the chapter on social psychology, and look for a description of the fundamental attribution error (FAE) . (It's also called the actor-observer hypothesis).* I'd estimate a 100% chance of finding it. It's the error of assuming other people behave in a certain way because of their personality, but I behave in a certain way because of my situation. In other words, assume someone has left a quarter in a soda machine. If someone else picks it up, you think, "He picked up the quarter left in the soda machine because he's dishonest.:" But if you pick it up, you think "I picked up that quarter because someone is going to take it anyway." But a new meta-analysis by Bertram Malle suggests that this error isn't nearly as robust as social psychologists think it is. To quote Malle, the discrepancy is only found when:

[...] the actor was portrayed as highly idiosyncratic, when hypothetical events were explained, when actor and observer were intimates, or when free-response explanations were coded. In addition, the asymmetry held for negative events, but a reverse asymmetry held for positive events.

* To be precise, the actor-observer hypothesis is not the same as the fundamental attribution error, but the differences are unimportant for the purposes of this post. Specifically, the fundamental attribution error is a hypothesis about an error: it assumes the actor is correct, and the observer is making the error.