Thought processors struggle with multiplicity
We found many of our interviewees wrote about the same topic in multiple forms across stages of development, often even rewriting full drafts of content once they reached a certain point, see also People process complex information in multiple levels and stages of processing. B12 R10 B2
When people write about the same concepts repeatedly, inevitably their words take different forms. This is often a positive. One participant told us that this helps her speak more fluidly about ideas with others. For her, writing multiple outlines and drafts is a practice that helps her figure out how to articulate her thoughts and learn what her points are. In the process, she will often generate new insights. For her and many others, multiplicity enables synthesis.
However, this multiplicity comes with challenges. Current thought processors are not well equipped for multiple frames around the same idea. Some participants experienced difficulty identifying the canonical version of an idea, overwhelmed in their search results with information that was largely repetitive (albeit with different wording) across drafts, outlines, and snippets. Their tooling is unable to recognize that all of that content is about fundamentally the same thing.
# Ideas for overcoming this challenge:
A logical solution might include continuously editing a single document, dodging multiplicity entirely. One of our participants was borderline offended at the notion that she would repeat her work by creating multiple versions. However, this may create a premature structure that adds friction at the point of idea entry, depending on discipline in user behavior. For example, if people come up with a new idea in a flash and they need to think about where it fits into an outline before they write it down, they might lose their train of thought. A common refrain we heard from participants was: If I’m thinking too hard about structure before I write the note, I forget what I was going to write down. The effort put into this structuring may not be worth it.
Many people don’t like to update original copies, often preferring to ossify an old outline as still potentially useful while making a new one for the purpose of iteration. People don’t intentionally review old notes.
This is comparable to save points in an RPG game – when you have a difficult choice about a document, you can create a fork to see what the outcome would be, and revert back. Multiplicity enables playfulness and experimental behavior, a key driver for synthesis. Requiring users to mutate the original copy can lead to a loss of provenance to the initial idea or source.
Multiplicity also presents an issue for people who need to manage multiple versions of the same document. For example, the authors all share a folder of markdown files. They each have personal notes, which use the same titles, for half-baked ideas not ready yet. Then they have more refined versions of a subsection of those notes in another directory to publish as a hypertext notebook. There is currently no way for the authors to reference and maintain all of the versions of a note at once as semantically the same “thing.” Transclusion is not the solution, as each version necessarily presents different and has different permissions for different people. Each page has its own backlinks section, and there is no simple way to compose all of the backlinks for each version into one view.
- Leaving ideas and notes in various places is a common default action that keeps us in flow, able to work at the speed of thought. Provided we return to write about the same concept, this multiplicity leads to diversity and playfulness in our ideas. Of course, this makes it difficult to accrete knowledge and can feel overwhelming when we sit down to compile our work.
- Our tools don’t help us here! Direct search fails to find all related instances to our ideas, and presentation of multiple versions is a UI challenge. Potential ideas to overcome this take advantage of implicit structure.
- searchable blobs as unnamed groups enable unnamed formalization
- Workspaces as a primitive, as seen in Codex OS, can infer structure from what entities are present in a workspace for later retrieval.