Tracing conceptual development in human-AIcollaboration
I wonder if we're missing a vocabulary fordescribing how ideas develop across distributed systems of people, artifacts,and increasingly AI. Authorship and acknowledgments don't really fit asdescriptions of conceptual development.
I legitimately don’t care if we call AI creativeor not, or honestly even if it should receive credit for ideas. Those areprobably important conversations for someone else to follow. I’m much moreinterested in how ideas develop and how we describe that development.
Often academia tends to tell linear stories aboutemerging ideas. First, cite the literature that influenced us, then acknowledgecolleagues who provided feedback, and then describe some contribution as thoughit emerged following one phase to the next.
Maybe not for everyone, but for me that has neverreally captured the entirety of it.
Sarah’s idea development process: do something,notice something, question something, have conversations about something, draftsomething, abandon that draft of something, reread literature, meet withadvisors, draft three other beginnings of articles, incorporate conferencequestions, notice something else, take a few months for unknown reasons, sketchout ideas that I hopefully remember, wake up to realize those scattered pieceswere different views of the same idea. So what can sometimes look like a suddeninsight emerges from months or years of slow conceptual accumulation.
This is compounded with the introduction ofhuman-AI collaboration. There’s not another author suddenly. It doesn’t makeintellectual development distributed. It makes visible something that wasprobably true all along: ideas develop across people, literature, convesations,artifacts, and now AI.
In my experience, AI doesn’t often contribute afinished idea. It will organize, ask questions, provide language, and surfaceinconsistencies. These are challenges to my established thinking and createsproductive friction that expands my thinking. It’s not authorship. But it’salso not nothing. How do we name that? Dismissing human-AI interaction as partof the process also risks dismissing genuine contributions.
Occasionally it will introduce a perspective thatsends me in an entirely new direction, but even then, that suggestion onlybecomes meaningful because it encounters an existing network of experiences,readings, conversations, and partially formed thoughts. The intellectual workisn’t the appearance of the suggestion. It’s evaluating whether that suggestionbelongs within an already developing conceptual system.
One reason I started writing/recording theseresearch notes is that they help make my development of concepts observable. I’mtrying to preserve the trajectory of these ideas. Ten iterations from now, Imight disagree with parts of this note, but that’s kind of the point. When you’restudying process, the evolution is part of the contribution. I’m not trying toestablish ownership or prove novelty (I mean, I’m not trying too hard toestablish ownership or prove novelty), but I’m much more interested in makingconceptual evolution visible.
These notes are meant to capture transitions wherea question changes shape, connections align, assumptions fall apart, newconcepts are adopted, and ideas emerge and are refined.
There are arguments for or against calling AI athinking partner. I do suspect that whatever we call these interactions will maketrajectory just as important as documenting outcomes.
I don’t have a framework for this yet. Somethingsomething… “Sarah, you can look into that AFTER YOU FINISH YOUR DISSERTATION!”But I do think there’s an opportunity to rethink how we describe thedevelopment of coherent ideas in distributed cognitive systems.
Some questions I kick around:
What counts as intellectual provenance in human-AIcollaboration?
How do we distinguish authorship from cognitivesupport?
Should research preserve trajectories instead of onlyfinished products? Is it already essentially doing this?
What evidence should matter when ideas emerge overmonths rather than moments?
Sarah, remember: this is allowed to beunfinished. If it were finished, it wouldn’t be a Research Note.
As always, these notes reflect my ownintellectual work. AI was used for organization and conceptual scaffolding,both as support of and as an object of iterative inquiry into the creativeprocess and cognition.