Skip to content

About

How do you design AI that people can understand, trust and act on?

Explainability Design is a research project by Christian Laesser about that one question. This page explains why it exists, how it works, how it’s made and where it stands.

Origin

It started as a way to learn.

Portrait of Christian Laesser

I’m Christian Laesser, an interface and data visualization designer based in Austria. For more than 15 years I’ve worked on one kind of problem: making complex systems legible to the people who use them and are affected by them. Mostly through data visualization and interface design, for institutions like WHO, the World Bank, UNESCO, Mozilla and Deutsche Bahn.

AI systems are the most consequential complex systems I’ve worked near, and the least legible. I wanted to understand explainable AI properly, and I learn best by building. So this project began as my own learning practice.

It is also, openly, how I’m moving my design practice toward AI. Both are true, and I’d rather say so than pretend it’s only one of them.

Method

Diagnosis comes before interface.

If you searched for explainability: this site covers the design side, not model internals. Every Exploration starts with five questions, and the answers depend on the type of system, whether it predicts, classifies, recommends or generates.

  1. What decision is the person making?
  2. How do they relate to uncertainty?
  3. How much time do they have?
  4. What can they actually do with an explanation?
  5. Who is accountable when things go wrong?

How this is made

AI drafts a plausible version. I look for the realistic one.

A project on explainability that hides how it was made would be a strange thing. This is the process behind every pattern and Exploration.

  1. AI drafts

    Research, pattern content, taxonomy and source synthesis. Fast, broad and often convincing.

  2. Human checks

    Framing, editorial decisions, and every claim against primary sources, such as the pattern files and the AI Act text.

  3. Still open

    The Explorations use mock data. Testing with real users and real deployments is the next step.

Research log

A work in progress, published as it’s finished.

Each piece ships when it’s finished, not before. The project grows with new patterns, new Explorations, and revisions when the vocabulary needs to change.

  1. Started as a learning practice

    Reading, building, and collecting the questions people ask about AI decisions.

  2. Pattern library

    24 patterns in four pillars: Understand, Judge, Act and Govern.

  3. EU AI Act reference

    A map from legal obligations to design patterns, plus a page on Article 50 disclosure.

  4. First Exploration

    Municipal Benefits Navigator: one resident, two AI systems, two deadlines.

  5. Explainability Design

    Renamed, restructured, and reframed as an open research project.

  6. Next

    Into practice

    A second Exploration on a high-risk credit decision, and first tests with real teams and users.

Working together

Let’s test this on a real product.

I’m looking for teams who develop or deploy AI systems and want to make them explainable to the people they affect, in the design sense rather than the documentation sense.

  1. Review and redesign

    A review of one decision flow in your interface, followed by a design of what should change.

  2. Design sprint

    A short, focused sprint on a single high-stakes decision, from diagnosis to tested interface.

Testing the patterns on a real product is what this research needs most, so I’m open to collaborations that serve both sides.

Write to me View my portfolio