← writing

What my thesis was about

1 min read

I finished my bachelor’s thesis this month: User Experience 3.0: Applying Artificial Intelligence on Web Platforms. It’s written in Serbian, and it’s a review with one case study, not original research.

The question was simple. What changes in user experience when AI isn’t a feature added to a product, but the main way you use it?

My answer was that part of the interface now gets created at the moment you use it. So the design work moves from buttons and screens toward intent, context and trust. I used my own project, pripremi.rs, as the case study.

Two things from it stayed with me.

First, these products make it easier to say what you want, and harder to know whether you got it. That’s the idea I wrote about in my first post, and the thesis is where it came from.

Second, the reliability in the prototype didn’t come from asking the model nicely. It came from checking its output in code. One small example: Serbian is written in two scripts. Telling the model which one to use was the first layer, and an unreliable one. What actually keeps a lesson plan in the chosen script is 127 replacement rules applied after the model is done.

The honest limit is that my conclusions are reasoned, not measured. I didn’t test any of it with users. That’s the obvious next step, starting with the teachers who already use pripremi.rs.