Autoren: Wilkens, Cost Reyes, Treude, Kluge
There is increasing attention to the notion of a human-centered artificial intelligence (AI) but no common ground of what this exactly means. Different disciplines provide a range of interpretations which tend to go in hand with their implicit basic beliefs in human behavior at work. In order to develop work systems where people and AI interact and contribute to mutual task performance there is a need for a theoretical foundation and empirical validation of the criteria for the human-centricity of AI. It is the aim of this paper to make a substantial contribution to this specification of criteria on the basis of a transdisciplinary literature review. The intended outcome is a classification describing different levels of human-centered AI in the workplace. The methodology in use is a systematic transdisciplinary literature review including combinations of the keywords human-centered or people-centered with artificial intelligence or similar combinations – both in English and German language. The search includes journal paper, books and book chapters from certain disciplines. The evaluation process aims at an identification of clusters, e.g. whether contributions face the deficits or strength of human beings etc. The data evaluation process can lead to a list of criteria for the human-centricity of AI and a contextualization when a criteria is typically in use. The authors will develop a reference model for human-centered AI which makes distinctions according to most relevant criteria. The first review reveals that engineering and information studies are the core disciplines facing human-centered AI in general while psychology and work science try to elaborate on criteria which allow a classification. These classifications address collaborative work settings and either give emphasis to the support function of AI, face human needs while using the AI technology, e. g. fairness, accountability, reliability, trustworthy or distinguish the collaborative role of the human beings, e.g. in terms of collaborator or supervisor respectively face the potential for enhancing human skills through AI. Further distinctions will be introduced in the full paper.
Zeitfenster der Veranstaltung (1)
Wilkens et al.