Introduction
Timo Baumann remembers the red corrections on school essays less for the errors than for the feeling they produced. The words were fixed, but he rarely understood why. With dyslexia, correction without explanation can become another kind of barrier.
Timo lives in Bern and develops AI tools for accessibility while studying. He does not treat accessibility as an add-on. For him it is the place where design either becomes honest or exposes its laziness.
Story of the Path into AI
Timo used read-aloud and writing tools for years, but he did not want to remain only a consumer of support technology. He asked why accessibility tools are so often designed without the people who need them every day.
He learned user-experience design, language models and testing with people who have different learning and sensory profiles. His first project was a writing assistant that not only corrected sentences but explained why one phrasing was clearer than another. The early version was too blunt. It fixed text in a way that made users feel judged.
Timo redesigned the feedback: several suggestions, no automatic replacement, user choice at the centre. That small change altered the tone of the tool.
Current Work
Today Timo works in a student project on inclusive learning software. In tests, he discovered that automatic correction can shame users even when the result is technically helpful. His tool therefore shows alternatives and explains trade-offs, allowing the user to decide.
Lecturers have also begun using the prototype to make their own assignment wording clearer. Timo likes that reversal. Accessibility is not only for a minority; it reveals where ordinary communication was too careless.
Personal Advice
“Do not build accessibility afterwards. By then half of it is already lost,” Timo says. His advice to developers is to test with real users before the interface has become too expensive to change.
Key Facts
Age and place: 20, Bern.
Background: dyslexia, young age, family rooted in manual trades.
Entry into AI: writing assistant that explains clearer phrasing instead of only correcting.
Focus today: inclusive AI.
Typical tools: UX tests, language models, assistive technologies.
Werkstattnotiz
Timo’s test notes include the sentence “the correction is right, but I feel smaller.” He keeps returning to it. His next prototype will measure not only error reduction, but whether people understand the change well enough to keep control of their own text.