Most of this article is behind a paywall but the idea is interesting. Do we not like precision and perfection as much as we might think? Should imperfection be built into our robotics?
Here’s another article that discusses the same concept. It is not behind a paywall.
Quote: “However, the subjects who found the error-prone robot more likable didn’t see it as more intelligent or more human-like than the robot that made fewer mistakes, the researchers found.”
The quote from the article suggests that there is a balance in robot design between perceived intelligence and likability. Social robots can be more well-accepted if they make mistakes. But, robots that must be both likable and perceived as intelligent for the tasks they are performing, need to have the right balance of imperfection to perform their function optimally.
So to answer your question, I think that the amount of imperfection that should be built into our robots is a function of the task that the robot will perform, and thus, said amount will be unique for each robot that exists.
And just for fun, here’s another independent study that comes to the same conclusion that robots that make more mistakes are more likeable.
However, unlike the study from the other article, “the faulty robots … weren’t rated as significantly less intelligent or anthropomorphic.” So until further studies can be performed, nothing definitive can be said about whether there is or is not a trade-off between mistakes and perceived robot intelligence.