In earlier work it has been shown how culture can be used as a parameter influencing human robot interaction in general. While this is a good starting point, in our work with concrete application fields we encounter that
culture in its usual definition as national culture is too general a concept to be useful in these concrete applications. Thus, we shifted our focus instead to a concept of local cultural practices, which is derived from situated practices as in Wenger's communities of practice and grounded loosely in Sperber's idea of an epidemiology of representations, i.e. culture or rather cultural practices as an emergent phenomenon from learning processes in a given group. Developing this new kind of culture-aware robots can then not start from a general
definition of culture like Hofstede, Schwartz and Sagiv, etc. but has to take the actual group of users (and stakeholders) into account. We exemplify this approach with our work in a residency for citizens with acquired brain damage.