In the same way as it is impossible to not communicate it is impossible
not to be emotional. This entails that for modeling communication with
robots we need to be aware of the emotional aspects of the interaction.
However, how are roboticists supposed to model emotional interactions in
robots? On the one hand, emotions are argued to simply display the
internal state of a system and thus to be universal in the way how they
are displayed. Yet, it has been observed that the emotional displays
depends on whether or not observers are present. This means that the
emotional display is subject to -- probably cultural -- rules. For a robot
to interact in a meaningful and understandable way it therefore needs to
integrate both perspectives. Emotions have also been shown to have a
significant effect on learning and memory. Again, this effect can be
distinguished between an internal effect -- motivation -- and an external
effect -- interaction. Cognitive accounts of emotion argue that emotional
states are related to the different states of achieving a goal.
Interactive accounts of emotion, in contrast, would argue that through
giving emotional feedback to a learning system one not only affects the
learner's motivation but also his/her understanding of the to be learned
action. By relating action states with emotional displays the learner
learns which situations are desirable and which should be avoided. In my
talk I will present results of our own research on emotional human-robot
interaction as well as insights in tutoring robots in order to address
the question how emotions can be used to facilitate learning.