Research topic

Explainable AI.

Only what we understand can we use responsibly. Explainable artificial intelligence shows how a system arrives at its result – and gives people the opportunity to check it, put it into context and correct it.

← Research

What is explainable AI?

Many powerful AI systems – neural networks, large language and vision models – learn from data without their internal decision paths being visible. They deliver results, but no reasons. Explainable AI (XAI) refers to methods and design principles that close this gap: they show which data influenced a result, how confident the system is and where its limits lie.

Why it matters

The more AI feeds into decisions – in medicine, public administration, industry or research – the more important trust becomes. Trust does not come from promises but from traceability. Explainability helps developers find errors; it helps users assess results correctly; and it helps organisations take responsibility. The European legal framework, such as the EU AI Act, also requires transparency and human oversight for many applications.

Our approach

For us, explainability is not an afterthought but a matter of design. Three principles guide our work:

  • Traceable chains instead of a black box. We break tasks down into verifiable steps – recognition, structuring, plausibility checks, human review. Every result can be traced back to its source.
  • Combining model knowledge and data. Data alone explains nothing. We combine learning methods with expert knowledge and models of the respective domain so that results acquire meaning.
  • Explanations for people. A good explanation suits the person who needs it: developers need different insights from users. We study which form of explanation actually helps in which situation.
How this works in detail is described by our shell model →

Explainable AI in our projects

  • Historical printed works: Language and vision models read Fraktur texts; automatic checks and spot checks by experts make the quality visible, and every entry refers back to the original page.
  • Satellite data analysis: Machine learning combined with expert knowledge of landscape and vegetation – so that it remains clear which data leads to which statement.
  • Visual feedback in therapy systems: Feedback that people understand immediately and can respond to.

From Chemnitz, for practice

We do not research explainable AI in a vacuum, but together with partners from Chemnitz and Saxony – where AI is already in use today or is set to be used tomorrow. If you would like to make an AI system more understandable, more verifiable or more trustworthy, get in touch.

Any questions?

Interested in working with us or would you like to learn more about our research? We look forward to hearing from you.

Get in touch →