The internet is a social innovation as much as it is a technological one. It democratizes access to and the production of knowledge, resulting in a situation where you can find any perspective on any topic. Learning online is a process of free exploration akin to musement (see Peirce), where the credibility of information takes a backseat to the epistemic confidence of the learner. This is embodied by the prototypical online learning community, wikis.

Wikis and linked data

Wikis lean into the fact that anyone can say anything (incidentally, a foundational assumption of the semantic web). As a jumping off point for the systems described below, consider the norms for the first wiki:

Here, information quality and attribution (provenance of assertions) are optional, yet it's one of the richest places I've found for learning online. It achieves this by leveraging the qualities that make learning online unique. It's plain how the semantic web can benefit attribution (e.g. linking several hypertext editions of a book to a common identifier, linking one assertion to all others by the same person). Though semantic wikis exist, such features are a mere novelty without a community of practice that adds frequent high quality information to the wiki.

The same can be said for content federation, which is the direction wiki's inventor has taken with FedWiki. In a similar fashion to Berner-Lee's Solid project for the semantic web, FedWiki lets users store and run their own wikis, adding and forking pages from other wikis. Pages contain version histories, but the branching paths aren't intuitive to navigate in the same way pages within one wiki often are (see my literature review on navigating information space). One can search across the entire federation, but the limited participation makes it preferable to search the rest of the web instead. Additionally, searching on FedWiki often involves wading through several same-named forks to find the best one. The original wiki's WikiVersionTwo includes an extended articulation of how to improve on wiki, but following it would lead us off topic for today.

FedWiki is an example of a WikiHive, a foundational construct behind my proposal for using the semantic web to facilitate learning. In a WikiHive, there exist independent wikis that grow as a collective. In a loose way, the semantic web is working toward a similar goal across the entire internet. Consider this website under that light. It's a corpus of ideas that I'd like to share with others, including pages from other websites (e.g. this Project Gutenberg book). I can store the provenance in RDF or deprive readers of it. Its contents are my choice, but it grows under the influence of others' work.

This website is not, however, collaborative in the sense a WikiHive is. A group of people could manually connect their websites in a way that offers the benefits of FedWiki, minus the ease and social contract that make it easy to participate. Outsiders would be better served by collaborating on a second layer, where the owner couldn't contest their ability to say whatever they want. Enter web annotation.

Web annotation and ThreadMode

Web annotation allows anyone to comment on a webpage, specify the portions to which their comment applies, and control who is able to see it. In contrast with wikis, where the boundary between author and reader is blurred, the system just described is broken into layers where the author has total control and another where the readers have total control. On the annotation layer, readers become authors, and so there's a resemblance to wikis. But the resemblance is limited to discussions. Thesis and antithesis are presented, but the synthesis never condenses onto a unified webpage.

The difference is that each reader has total control over what they author. In this way, the distinction between the first layer and the annotation layer can be recursively applied (though only as an abstraction)1 and the branching layer cake of collaboration comes to resemble FedWiki in that edits are (would be) forks and navigation is difficult.

The wiki construct that embodies both of these resemblances is ThreadMode. From WikiWikiWeb:

The observations InPraiseOfThreadMode focus on the benefits of signing assertions. The reasons ThreadModeConsideredHarmful include:

• Signatures and writing in first person inhibit editing
• It becomes hard to extract meaning from the thread (requires context)
• It discourages a holistic point-of-view, instead considering what what last said
• That it preserves and emphasizes the trivial. Who said what and in which order may not really have lasting significance.
  The author of these points acknowledges that thread mode is a more convenient and natural way to collaborate. All this applies to web annotation, only more so. Read through any of the wiki pages I've cited and imagine following the same conversation through web annotations. Jumping from site to site as the conversation progresses would feel like one of the wild goose chases in the margins of a middle school textbook2.

Annotation's ease of use facilitates getting ideas out in the open. If one wanted to avoid the pitfalls of thread mode, they could collect a list of annotations and write up a webpage that synthesizes them and references the list (per contribution method 4).

Web Annotation and linked data

Leveraging the semantic web here could help solve all of these problems. If the comments in annotations had semantic metadata, one could perform inference on the network of conversations. The same conversation could be encoded by anyone who cared to do so. Which assertions to include for inference is up to the user. Users could even compare different encodings of the same conversation. The same could be done for webpages.

Using annotations for metadata also addresses the problem of a website that doesn't provide its sources. "such is the irresistible nature of truth, that all it asks,—and all it wants,—is the liberty of appearing". If only that had a citation!

Semantic annotation: A potential implementation

This webpage was written in markdown using Obsidian, the personal website uses neocities, and the annotations on this page use hypothes.is. I've characterized the entire internet as a quasiWikiHive. So far, I've linked an online ebook to its author and linked two annotations on that ebook to their common author. Despite technical difficulties, the annotations could have been linked to the webpages they were on. But none of that leverages the semantic web in a novel way.

Semantic annotation for a personal knowledge manager

Obsidian has a rich plugin environment, but none of them integrate it with the semantic web. As a personal knowledge manager, the link structure of one's notes reflects the semantic organization of concepts in their mind. While the software is helpful for personal knowledge management as-is, support for RDF could help with querying relationships among a large knowledge graph.

Personal websites as parts of quasiWikiHive

Neocities isn't particularly noteworthy here except for its ease of use. But I'm acting under the premise that people will have a "their place" where they host (or at least control) all the data the data they want to share. Solid and IPFS act under the same paradigm. If you want something to be around, store a copy.

Annotation and its limits

This page includes annotations that would be better stored as metadata than comments. Their inclusion is still helpful, but only to people on the page for as long as they're on the page. If the web were an archipelago, hyperlinks are more like bridges than maps. It's not linked data.

1 For the purpose of this paper, assume it's possible to annotate annotations (in contrast with replying to them)
2e.g. "Go to page 47 -> Go to page 99 -> ... -> UwU"