Public Interest AI

The use of artificial intelligence for the public interest is repeatedly formulated as a central goal in political AI strategies, guidelines and government funding measures, as well as in projects from commercial enterprises. However, it is still unclear whether and, above all, how the public interest orientation of AI applications can be realised and how it can be checked. The “Public Interest AI” research group is therefore developing an operationalisable analysis procedure that allows defining and assessing public interest AI on three levels: on an argumentative level, on the level of the technical implementation of AI, and on the level of the usability of AI in the application context.

To illustrate the practical feasibility of this concept, we will develop two prototypes of a public interest AI. A first prototype uses computer vision for automatically detecting the barrier-free accessibility of various locations. The second technical prototype is based on Natural Language Processing and deals with simplifying German texts in order to minimise exclusion of less literate citizens. A third research project deals with the development of a design framework for public interest AI with a specific focus on users understood as citizens.

Through collaboration with non-profit project partners, real-world cases are drawn upon to serve the research group in testing the analysis procedure. With this practice-oriented research approach, the research group is developing an exemplary catalog of requirements for assessing to what extent socially oriented AI projects serve the public interest. As a result, this can function both as an important contribution to the discourse on AI development in Europe and as practical implementation guidelines for AI developers.