In standard hyperlinks, Document P points to Document Q. That's it. It's a one-way link from P to Q. Document Q knows nothing at all about Document P. If we were to diagram it, it'd look something like:
P -> Q.
Those standard links are mono-directional. By contrast, a bidirectional link does...well, it does just what it says on the tin. A bidirectional link between P and Q links P to Q and also links Q to P. Those look more like:
P <–> Q.
Up until the advent of the web, pretty much everyone who talked about hypertext assumed that links would be bidirectional. Indeed, plenty of hypertext theorists are still mad about mono-directional links.
When you're building out a distributed web—like, say, Tim Berners-Lee—a mono-directional link makes a lot of sense. The world, after all, is full of bad actors. Imagine managing a website in which 4chan could freely leave links on your site pointing back to ... whatever they wanted. *shudder*.
Of course, a lot of those logistical concerns go away when you think about putting bidirectional links between pieces of your own content. As long as you can avoid trolling yourself, you can be reasonably secure in linking your content in both directions.
The digital gardening movement picked up steam with the release of Roam Research, a note-taking tool that incorporates automatic bidirectional links, using an authoring system that is more like a wiki than like a traditional note-taking app. Roam's linked notes surface new information—not unlike the graph database.
I’m going to talk only about unidirectional links here. Bidirectional links are probably the way the web should have been, but the HTML hyperlink is unidirectional, so that's what we're going to talk about for right now.
When Sir Tim Berners-Lee built the three protocols that underlie the web—the hypertext transfer protocol (or http), the uniform resource locator (or URL) and a hypertext markup language (or HTML)—he prioritized linking between resources over embedding content. Indeed, in some ways, "linking between resources" overstates things. Properly speaking, HTML doesn't link resources so much as it provides an address where a relevant resource should be located. And even then, those links run in only one direction.
Bidirectional links were a staple of every hypertext system from the 1960s on. They are enjoying a resurgence today, though they are largely limited to niche platforms like Roam.
But what it boils down to is that Roam allows users to add bidirectional links to notes, which results in a set of massively interlinked notes. And because those notes can also store relationships between notes, they can surface connections that the note-taker wasn't aware of.
Ted Nelson coins the term "hypertext" as part of his proposed Project Xanadu, an ambitious hypertext system that would allow readers to create “zippered lists” that form compound documents from various pieces of other documents. Unlike today’s hyperlink, Xanadu’s links were bidirectional.