Two sides of an encounter
On one side is the human being – with their cognitive processes, the way they take in, organise, abstract and also forget knowledge, with their talents and their impact on the world. On the other side are data technologies – machine learning, neural networks, data science – that find features, structures and relationships in measured data. Human-AI interaction asks how both sides can act on each other so that people are strengthened rather than sidelined.
Three fields of interaction
The encounter between people and AI is not a one-off moment; it accompanies a system throughout its entire life cycle:
- Development. Developers understand the mechanisms of the AI, interpret results and intermediate results and can influence them in a targeted way.
- Use. Users understand how the system works in an adapted form, can follow its results and respond appropriately.
- Further development. Learning systems change continuously. People and system stay in dialogue so that human goals are preserved over time.
What makes good interaction
- Understandable: the system shows what it is doing and how confident it is.
- Controllable: people keep the decision and can intervene.
- Appropriate: feedback and explanation suit the person, the task and the situation.
- Learning in both directions: the system learns from people – and people understand the system better over time.
Human-AI interaction in our projects
- Visual feedback in therapy systems: real-time feedback on movement – here interaction becomes directly tangible.
- Historical printed works: experts and AI work in a shared chain; the machine reads, people check and decide.
Interdisciplinary, from Chemnitz
Human-AI interaction brings together computer science, engineering and psychology. These perspectives meet in our team and advisory board – and become practical in projects with partners from Chemnitz and Saxony.