SLOP3203: Autocomplete Studies

A semester spent taking predictive text seriously: from T9 keypads to code completion to language models, tracing what it means for a machine to keep finishing your sentence, and what that has changed about how we write, code, and decide what to say next.

• predictive text
• HCI
• language models

What you will do

Every week pairs a lecture on one predictive-text domain with a lab where you build or interrogate an artefact from it: a toy T9 disambiguator, an autocorrect you deliberately break, an audit of your own inbox's smart replies, a side-by-side of three forecast UIs and how each one hides its uncertainty. You leave the semester with a small portfolio of builds and case studies, and a final project that turns one of them into a proper audit of a real product's suggestion design.

Who it is for

A student a couple of years into a CS or HCI-adjacent degree who already writes small programs without help, and who is more interested in why a piece of software suggests what it suggests than in shipping a polished app. If you have never once been annoyed at autocorrect, or "it's just next-token prediction" feels like it ends the conversation rather than starting one, this course is aimed a little past you.

Where to go next