The research project Human in the Loop? investigates how human participation can improve automated decision-making processes.
Automated decisions are not always free of errors. This is because they are based on training data that may contain unintentional biases or lack human contextual understanding. Thus, individual machine decisions often do not do justice to the individual situations of people. A well-known example is lending, where banks use technological systems to automatically assess the creditworthiness of applicants. This is why there have been calls for a long time to integrate people into such processes so that they can monitor the decision-making processes and thus contribute to improving technological systems.
The project poses the following questions: How should meaningful interaction between humans and machines be designed? What role do human decisions play in the quality assurance of automated decisions? How can we ensure that this interaction is not only legally compliant, but also transparent and comprehensible? And what requirements apply to the interaction between humans and machines when considering the technical system, the human decision-makers, their context and their environment?
Project Focus and Transfer
Four Case Studies
Analysis of human participation in automated decision-making processes through field analyses, workshops, and dialog formats in four selected scenarios.
Taxonomy of Influencing Factors
Investigation of the factors that influence human decisions and identification of the errors, vulnerabilities and strengths of all technical systems and people involved in decision-making processes.
Recommendations for Action
Development of practical solutions to optimize the collaboration between humans and machines and to improve the implementation and interpretation of existing legal and regulatory projects (GDPR, AI Act and DSA).
Image by Luís Eusébio on unsplash.com
