Independent project · HCI · Interaction design · Ongoing
AI is a spectrum
What if AI didn’t have to look like a chatbot?
Where it starts
We’ve become so used to interacting with AI through chat that it’s easy to forget it’s just one possible interface.
But what happens when we stop starting with the chatbot and start with what people are actually trying to do?
AI is a spectrum is my ongoing exploration of that question.
The idea
The interesting design question isn’t just what AI can do, but how we want to interact with it.
AI interaction isn’t a choice between chatting with an assistant and letting an agent do everything for you. There’s a whole range of possibilities in between, and beyond.
AI can be visible or invisible, conversational or embedded directly into an activity, something we collaborate with or something that works quietly in the background.
Making the possibilities tangible
I started AI is a spectrum to make these possibilities easier to see, discuss, and experiment with.
The project brings together examples of alternative AI interactions, ideas from human-computer interaction research, and small interactive experiments.
Rather than proposing one ideal interface for AI, I’m interested in exploring how different interactions might suit different people, activities, and contexts.
Examples
24 products and research projects that already let you work with AI beyond the chat box. Each one is sorted by the main way you interact with it: pointing, adjusting, showing, choosing, reshaping, stepping in, noticing, or AI that stays ambient.
Browse the examples ↗
Playground
Six small experiments, grouped by five things AI can do: interpret, generate, transform, notice and act. Most let you try the same task twice, once the familiar way and once without the chat box, so you can feel the difference instead of reading about it.
Try the experiments ↗
A note on the experiments
There’s no real AI in them yet. The replies follow simple rules, so the only thing that changes is how you interact. Each one is an inspirational bit: a small, working thing that lets you understand one property of a technology by using it (Sundström et al. 2011). I first used the method in my master’s thesis at KTH.
An ongoing experiment
This isn’t a finished product or a definitive framework.
It’s a place to explore questions, test ideas, and build things that make alternative futures for AI interaction a little easier to imagine.
I’m particularly interested in how we can design AI that supports human thinking, creativity, and agency rather than simply automating more of what we already do.