Criticality of AI-based Systems
Artificial intelligence systems are increasingly being adapted to human activities. Fed by data humans generate, they are working ever more closely with us in real time. The search for artificial intelligence is also a search for human values. An important part of our research was on how knowledge can be integrated into technologies and how it can be derived from them.
This research group conducted research at the Weizenbaum Institute from 2017 to 2022. In the newly launched research program, research will henceforth be organised in 16 research groups. These will be flanked and supported by the new Weizenbaum Digital Science Center.
Abilities and Limits of Artificial Intelligence
Ubiquitous stealth technologies have the ability to expand and support human knowledge, thereby optimizing people's potential and productivity, but also to exacerbate overlooked distortions and errors in increasingly complex, interconnected systems. To the extent that society is dependent on autonomous, intelligent, and critical software systems, new strategies must be found to ensure not only their existing quality aspects such as safety, efficiency, reliability, and security, but also the ethics of human-machine interactions and the associated socio-political implications.
Artificial intelligence systems are increasingly being adapted to human activities. Fed by data that we humans generate, they are working ever more closely with us in real time. It is becoming increasingly obvious that the search for artificial intelligence is also a search for human values. How knowledge can be integrated into and extracted from technologies was an important part of this research. Central aspects such as systematic discrimination (bias), trust, transparency, responsibility, sustainability, human-machine interaction, and the need for new forms of education were core topics of the research group.
Although they already influence the everyday life of citizens in a variety of ways, many AI systems have up to now been black boxes. Their public perception is significantly shaped by mistrust, but also by ignorance of the theoretical framework of these systems. Therefore, another goal of the research group was to involve the public in the scientific discourse on AI. We wanted to sharpen the awareness of citizens of the capabilities and limitations of AI technologies and to break new ground for scientific research through social dialogue and inclusion.
Methods of Psychology, Computer Science, Social and Cultural Sciences
Both quantitative and qualitative methods of psychology, computer science, social sciences and cultural studies were applied: metrics of quality engineering, literature research, application case studies, discourse analyses, questionnaires, interviews, workshops and interactive events, public discussion forums and debates.
In our research, we considered intelligent algorithms as technologies on the one hand, and as cultural artifacts on the other hand, determined by the ontological position of the individual developer.
We analysed the state of the art of artificial intelligence from a technical point of view as well as the discourses (on AI), the symbolic violence involved, and the biases of such systems.
In addition to developing a theoretical framework for cultural technology research, we were working to broaden the horizons of the current state of user-oriented AI, AI for the common good, and inclusive AI. Our approach is characterized by questioning and deconstructing the current design principles for development of intelligent, critical systems and the development of alternatives to such systems.