We've covered a lot of ground. But in doing so, a few common themes have started to emerge.
The document is the fundamental unit of information exchange on the web. Research findings are (usually) collected into IMRAD reports that are (usually) formatted as PDFs. Those reports—whether PDF or HTML—are referenced by a single URL and are (usually) categorized via a hierarchical taxonomy.
All those factors make research outputs frustrating to use as new research inputs.
A better framework would store individual concepts as fully addressable pieces of content. Doing so would make it easier to create connections between ideas, to reuse—or transclude—text across projects or even across research organizations.
Storing content at the level of concepts/ideas also allows for...
Documents are a poor way of visualizing the links between texts. Yes, a good report follows a logical sequence. But that sequence is merely one of many possible logical sequences. There simply is no one right way to organize a piece of nonfiction.
The document is a particularly poor way of thinking through possible arrangements of ideas. (There's a reason that high school English teachers make students take notes on index cards when you're learning to write your first research paper.)
Unfortunately, research websites largely bake in all the disadvantages of documents.
When you've stored content as ideas, you can easily shuffle those ideas around. A block of content held on a digital card can be moved around the screen, combined and recombined with other items, stacked a bit like Lego bricks.
This means having multiple windows at once. Not switching between tabs in a browser, mind you, but interacting with multiple lexia within the same UI context.
Building in new visual metaphors—ones in which text can take on spatial arrangements more complex than the traditional before-and-after of linear print documents—enables the creation of knowledge graphs, much like the ones available in Roam Research, which can in turn enable researchers to surface connections that they hadn't even realized were there.
Better hyperlinks would have a few defining features.
They're bidirectional by default. When you're exploring connections between ideas, knowing what is linking in is every bit as important as knowing where you're linking out.
They're visible. No blind jump links. You should be able to see context before you click on a hyperlink.
They overlay the text. Hyperlinks that are embedded in the text show only one set of connections—namely, the ones that the author happened to care about at the time. Hyperlinks laid over the text (via a link database) allow for different authors to make different connections, or even for a single author to link one lexia to multiple other lexia.
Allowing external users to add their own links is one piece of making...
Link databases make it possible to create personal webs on top of content. But frequently, we want more than just a hyperlink—even a hyperlink that encodes relationships. Often we want to add our own notes to the text, to comment or engage in the margins, much as we would inside a printed text.
Annotations, like links, would sit on top of the text, allowing readers to collect notes on any piece of hypertext.