Back to selected work

Case study

Human-Centred AI Research

Research into how people form expectations of AI — and what that means for trustworthy design.

Perspectives

Experiencing AI

Method stages

AuditPathwayPrototypeTest

Organisation and timeframe

MA research at Elisava School of Design and Engineering (2024–2025), inclusive user research for the i-Kfz app at Bundesdruckerei, and Seeing the Algorithm, a workshop programme for a Berlin school (2026).

The situation

People's expectations and mental models of AI decide whether they trust it, use it well or avoid it. Those expectations are rarely researched before systems are built, especially for people who are older, have disabilities or are younger and growing up with algorithms they can't see.

Audit

What I found

In my MA research, I used critical journey mapping across helpdesk, identity and authorisation, and cross-system credentials, and analysed gaps in literacy, explainability, recovery and accountability in public sector AI systems. The systems were technically strong, but people's trust didn't follow accuracy. It followed whether they could understand what the system did, question it, and find a way forward when it got something wrong. In the i-Kfz research, I worked with older people, people with visual and motor impairments, and people with low digital confidence. Designing for the people least confident with technology made the service clearer for everyone.

Pathway

The pathway we designed

These findings became design principles I now use across my work. Recovery first: every wrong outcome needs a clear, human path forward. Design for the least confident user first: if they can understand the system, everyone can. Explain in plain language what the system did and why, and always keep a way for a person to step in. For Seeing the Algorithm, I designed a method to surface how children picture the algorithms around them: they sketch on paper, turn the sketch into a story, write a prompt from the story, and compare the AI-generated image with their own drawing.

Prototype

What we made tangible

The research principles became part of the Trust by Design framework and the TrustAudit self-assessment. For Seeing the Algorithm, I designed a five-phase programme: a discovery workshop with teachers and parents, three student workshops and a closing exhibition. It could grow into a living archive of how children imagine AI. I also built the project website, seeing-the-algorithm.org.

Test

What testing showed

For i-Kfz, the inclusive approach was confirmed at launch: the app met all 98 BITV 2.0 criteria, and over 1,000 app store reviews mention its user-friendly design, including from people who say they struggle with technology. The MA principles were presented to more than six communities within German government, where operational teams recognised the gaps once they were named. Seeing the Algorithm hasn't run yet, so there are no findings from it so far.

Scale

What scaled, changed or stopped

The MA principles grew into Trust by Design and an independent practice. The i-Kfz approach became a benchmark for accessible public digital services in Germany.

What the organisation can do now

Teams can use recovery first and design for the least confident user as concrete design criteria, not general values. Seeing the Algorithm adds a ready-to-run method for surfacing mental models of AI with children, families and teachers, which can be adapted for other groups.

My role

Sole researcher on the MA research. UX and accessibility research lead on i-Kfz. Designer and facilitator of Seeing the Algorithm, and builder of its website.