Social AI are machine learning models used to create meaningful predictions and subsequent actions based on social media data and such models are becoming important engines of the data-driven society. Social AI automate what kind of news is being presented to what kind of people and the algorithms automate how people are depicted through social data. This AIAS project will critically scrutinize and discuss to what extent Social AI are able to create meaningful predictions that are sustainable both to our understanding of the social human being and to society. Through six case studies of empirical uses of different AI models, and a historical account of central AI problems, the project proposes a pragmatic theory of social AI. The emphasis is on contextualizing data as depictions of the human in which classifiers and training data and navigating conflicting ambiguity play central roles. The project proposes that making visible political dimensions of model training, reasoning and the connected interpretative work flows are together with a close eye for knowledge that can be derived from outliers, important steps that need to be taken in order to advance the further development of Social AI.
Anja Bechmann is research director of AU Datalab, assoc. professor at the Media Studies Department and was a JCS Fellow at AIAS from 2017-2018. In the framework of various research grants she conducts multidisciplinary research at the intersection between algorithms and media sociology, entangling how algorithms create meaning from digital human communication and behavioural data, and the challenges in doing so both regulatory, ethical and in relation to sociology and information design.