Public sector · Medium-high stakes
Municipal Benefits Navigator
A friendly chat in front of a decision it doesn’t make. How do you design that honestly?
UX design for explainable AI
Explainability is a design problem.
I’m researching how to design AI interfaces that make decisions understandable, trustworthy and actionable for the people they affect. With the help of AI, I curate a library of UX and interface design patterns and test them in design Explorations.
Latest Exploration
Why this project
AI systems increasingly decide things about our lives, and most give people no way to see why.
Explainable AI (XAI)
Model-level methods like SHAP or LIME help data scientists trace a score back to its inputs. That work is necessary, but it rarely reaches the person affected by the decision. They ask something else: why did this happen, and what can I do? Answering that is a job for interface and UX design, not for the model.
The EU AI Act
The law now requires parts of that answer. Lawyers interpret the obligations and engineers build the models, but the screen where people meet the outcome is seldom designed with explainability in mind. That is the gap I’m exploring.
AI as a partner
AI helps me research, draft and structure the patterns and Explorations. I am checking the results against primary sources and my own professional judgment. Using AI to study how AI should explain itself is part of the experiment: I learn its strengths and limits first hand.
Patterns
A working library for designing and auditing explainable AI interfaces.
Each design pattern answers a recurring question about an AI result, not with a UI component alone, but with the reasoning behind one. It names the interface building blocks and what they cost. The library is organized around four questions: Can I understand it? Can I judge it? Can I act on it? Who is accountable for it?
Understand
Interface patterns that help users grasp what an AI system did and why.
8 Patterns
Judge
Interface cues that help people decide how far to trust an AI result.
4 Patterns
Act
Flows and controls for those affected to challenge, correct, or override an AI outcome.
4 Patterns
Govern
Design practices that make AI systems transparent and accountable, to their users and to regulators.
8 Patterns
EU AI Act
Navigating the EU AI Act timeline
The EU AI Act turns parts of this problem into law, on two clocks. Disclosure comes first: since 2 August 2026, people must be told when they are interacting with an AI system. Explanation and oversight follow: from 2 December 2027, high-risk systems in areas like credit, employment and public benefits need human oversight, record-keeping, clear instructions for deployers, and a right to an explanation for people affected by a decision.
-
Transparency obligations apply
People must be told when they interact with an AI system, and AI-generated content must be marked.
-
Today ,
-
Grace period for content marking ends
Systems already on the market before August 2026 must now embed machine-readable markers in AI-generated content.
-
High-risk AI systems
Rules apply to high-risk systems listed in Annex III, such as credit scoring, hiring and access to public benefits.
-
High-risk AI in products
Rules apply to AI embedded in regulated physical products listed in Annex I, such as machinery, toys and lifts.
Does your AI system decide about people?
I’m Christian Laesser, a UX, interface and data visualization designer based in Austria. For more than 15 years I’ve made complex systems legible for institutions like WHO, the World Bank, UNESCO and Deutsche Bahn.
Now I’m looking for public bodies, govtech teams and institutions whose AI systems help decide who gets a benefit, a job referral, a loan or a treatment. I design the interface and UX of the explanation those people receive, so they can understand a decision, check it and act on it. Let’s test these patterns on your system.