According to scientists, results confirm effect of an oversight (effect at which the appeal of the person increases when he makes a mistake). In research team note that deeper studying of mistakes at communication can be useful to further optimization of interaction between the person and the robot. So, the robot "understanding" that in interaction there is some problem (by means of reading of social signals of the person) could let the user know that he knows about her and to begin to apply strategy for her elimination.
Read continuation of article in a source
Actually I consider this moment disputable. I don't want that robots were mistaken in those cases when it concerns human life.