Collaboration between humans and AI systems: Potentials and design approaches for a reorganization of knowledge work and mutual learning
Advances in the field of artificial intelligence (in particular, machine learning and speech recognition) offer new design options for reorganizing knowledge work at the interface between humans and AI. AI systems provide potential in the automation of routine tasks and support employees in solving complex tasks as new "team members". They can help complementary pool skills in many areas. People perceive AI-based systems as social actors, but therefore also have similar expectations of the quality of their solutions and their communication behavior, which are often not met and can lead to dissatisfaction, rejection, or non-use of the systems. The differences in humans' abilities and skills (i.e., human intelligence) and machines (i.e., artificial intelligence) create new design challenges in collaboration and learning processes for human and machine learning.
The junior research group aims to develop, test, and validate socio-technical design requirements and patterns for the development of AI systems in knowledge work. These implement collaborative working practices of human-AI cooperation, particularly for the division of labor, for the transparent, comprehensible transfer of tasks and work statuses, and for promoting learning between humans and AI systems according to their respective strengths.
Representative collaboration scenarios in knowledge work are surveyed and modeled in an application-oriented manner by empirical requirements elicitation with companies to achieve the objectives. Based on this, the junior research group develops a taxonomy for labor division between humans and AI systems. Besides, techniques for transfer orchestration between humans and AI as well as techniques for the promotion of AI (or human-) supported human (or machine) learning are explored and transferred into design patterns. The developed techniques and design patterns are prototypically instantiated and socio-technically evaluated in the laboratory, field, and online studies. The project thus follows a design-oriented multi-method approach of iterative development and evaluation.
The planned results are proven techniques and design patterns for the cooperation and mutual learning of knowledge workers* and AI assistance systems, considering aspects of transparency, ability to act, and autonomy of the participating employees*. It includes evaluation results on the effects of the different design variants within the prototypical pilot project framework and gives recommendations for research and practical users' actions. An action guideline for the design of human-AI cooperation scenarios will be provided.
The transfer of the project results is achieved via the unsubsidized practice partners in the research association as multipliers in other organizations and user publications as well as utilization in research (joint final volume, scientific publications, and presentation at conferences) and teaching (courses and supervision of theses). The junior research group's establishment serves the long-term strengthening of human-centered AI research at the two university locations Hamburg and Kassel.
The project aims to develop, test, and validate socio-technical design requirements and patterns for the development of AI systems in knowledge work, in which collaborative working practices of human-AI cooperation are implemented. In particular for task definition/allocation and division of labor between humans and AI systems according to their respective strengths; for transparent, traceable transfer of tasks/work status in the collaboration process from AI assistance system to knowledge worker and from knowledge worker to AI assistance system; for the promotion of AI-supported human learning and human-supported machine learning.
Usage and transfer objectives
On the one hand, this differentiated objective enables a fundamental research approach in the field of hybridization of human and artificial intelligence. On the other hand, the meta-objective, the development, and testing of socio-technical design requirements and patterns provide a practice-oriented focus. Design requirements and patterns represent an ideal bridge to enable a transfer to user companies.
Funding policy objectives
About the research topic, the HyMeKI project outlines two of the three research topics mentioned in the BMBF-Announcement - Fundamentals of AI and Machine Learning. Practical relevance is ensured by integrating transfer partners (aiconix, IHK Hessen innovativ, and Lycronize, among others). With regard to the funding purpose of having junior research groups led by women, this purpose is fulfilled in two ways by the co-leadership of the junior research group. The co-leadership can also set impulses at two German universities.