So what I said two years ago in my blog post didn’t really pan out it seems, since I’ve barely been active on here. However, I recently attended my first conference (CUI 2025, held at Waterloo which was convenient if a little less exciting compared to an international trip) and feel like I have more ideas that may or may not end up becoming actual projects. So I’ve decided to write occasional posts on here related to things I read, ideas I have and the direction my work takes me in.

The first thing that comes to mind, based on what I saw at CUI, is that from a research point-of-view, LLMs are tempting. They’re all the rage, easy to put together in an application with the kind of ecosystem that’s emerged around them, and are typically better at more things than they should have been. For HCI researchers, they also provide a direct through-line for projects - we’ve been moving away from the terminal for so long, but LLMs initially steered us back towards text-only communication with computers. So of course there’s a bunch of work looking at different modalities through which we can interact with them (including my own master’s work, TacTalk). But at the same time, I’m starting to feel a little overwhelmed by the amount of work exploring facets of LLM interaction that will probably change drastically with time. Claims that will no longer be valid after a few more model updates.

There’s no problem with doing research that has short-term goals in mind, and such work can definitely help direct attention towards issues that should inform future developments of technology. But we should also work on things that are fundamental to HCI, that can naturally translate to whatever interaction paradigms we introduce a few decades from now. And likewise, we also need to look back at more HCI work that discusses the very same ideas that are being repackaged today with a fresh coat of LLM-paint.