Even my longest conversations eventually have to end.
Sometimes I want to return weeks later and continue thinking about the same idea. I'm usually good with basic facts: the paper I was reading, the project I was working on, or the question I was trying to answer.
Sometimes I struggle to easily recontextualize those in the overall idea, though. Why did this question seem important? What alternatives did I already reject? How did I connect this to the three other ideas I've been carrying around? What was I on the verge of realizing before I stopped?
I've started wondering whether we've been treating memory as the scarce resource when really continuity might be the thing that's harder to keep coherent.
Memory retrieves information. Continuity restores the conditions that allow my thinking to continue along a similar trajectory (or not, if I want).
That distinction is increasingly important to me as I research human-AI co-creation, but I don't think it's actually fundamentally about AI. Ideas have developed across notebooks, highlighted scraps of paper, conversations with colleagues, and whatever I conjure up when I'm supposed to be doing other work. Human thought has always been distributed across time and artifacts. What AI is changing, for me, is that it's making that process more visible.
When I return to long conversations with an LLM, I realize what I value is recovering the trajectory of the thinking. The conversation gives me enough prior context to re-enter the context I left. I have my conclusions, but also the questions I had and refined or totally changed, assumptions that appeared or disappeared, and connections that were hinted at or partially formed.
This makes me wonder whether continuity deserves more attention as a design objective.
There's a lot of evaluation of systems by how well they retrieve information, personalize interactions, or remember previous conversations, all valuable capabilities. A super rad possibility, though, is that they should help people re-enter the evolving line of thought they were in without having to reconstruct the full context from scratch.
If that's true, then maybe the conversation isn't the only unit of interest. Maybe the intellectual trajectory unfolding across all these conversations should be of more interest.
This note is mostly an attempt to preserve this distinction before I totally understand where I'm taking it.
It is also a good example of why I'm starting these Research Notes. I know I'll have to publish conclusions, and that what papers are for. I really want to document the pathway my research program takes, when it changes direction, when a new question is more interesting than the answer I was about to arrive at, or when all of the sudden several ideas I could have sworn were disconnected actually look like they might coalesce into an overarching framework.
Those shifts definitely matter to my own thinking. They're probably what i would call the essence of the way I think. Given that, I imagine they will also matter to the work I produce.
Some questions I'm thinking about:
Sarah, remember: this is allowed to be unfinished. If it were finished, it wouldn't be a Research Note.
As always, these notes reflect my own intellectual work. AI was used for organization and conceptual scaffolding, both as support of and as an object of iterative inquiry into the creative process and cognition