Abstract
The goal of human-AI symbiosis is a mutual benefit for humans and AI. The challenges of gaining a human-AI symbiosis are manyfold. One challenge is what is communicated to humans and when. Too frequent inquiries by an AI will cause the human to feel distracted, which can lead to negative experiences. However, if the AI takes over some or all of the task solving for humans, there is a risk that humans will become dependent on the technology and loose skills. The right type of communication between humans and AI is important. The goal of our PHD research is to develop a method to optimize human-AI symbiosis through appropriate interaction and explainability. Our approach is to use AI to encourage humans through nudging to discover solution strategies on their own. In this way, the cognitive performance of a human is aimed to be preserved as much as possible. This could also counteract common problems in modern AI systems such as parameter drift. Therefore, we will study the effect of nudging in human-AI interactions and gain insights into what would happen if the machine nudges a human in the wrong direction. For this purpose, a user study will be conducted in which the human interaction with an AI is simulated. Afterwards an AI will be developed that uses nudging mechanisms that reliably nudge the user towards the correct solution without losing the human’s own cognitive effort. The AI should support humans unobtrusively and thus be integrated into daily working life.