The Brain Attic

Knowledge Management Systems, once a pioneering segment of consumer technology, have largely disappeared from popular attention. Their revival can help us to be better problem solvers.

After Watson has explained that the earth does, in fact, revolve around the sun…

Holmes: “Now that I do know it I shall do my best to forget it.”

Watson: “To forget it!”

Holmes: “You see, I consider that a man’s brain originally is like a little empty attic, and you have to stock it with such furniture as you choose. A fool takes in all the lumber of every sort that he comes across, so that the knowledge which might be useful to him gets crowded out, or at best is jumbled up with a lot of other things, so that he has a difficulty in laying his hands upon it … It is of the highest importance, therefore, not to have useless facts elbowing out the useful ones.”

How much do you read every day?

As a software developer, student, and a curious member of humanity, I read a lot. I’ll be willing to bet that you do, too. We all seem to have our own world of Twitter feeds, newsletters, textbooks, cookbooks, comic books, blog posts, Facebook groups, and journals to keep up with. It can feel overwhelming, even when we make good decisions to filter what’s worth reading and what isn’t. We might feel like we should be reading more, and sometimes we do. But …

How much do you forget?

Forget that, how much can you simply no longer find? A couple of days ago, I spent thirty minutes trying to find an article outlining a specific anthropological theory on the role of religious rituals, but I never found it. (I’m fairly convinced I wasn’t hallucinating, as I can clearly picture the title page). Time would fail me if I recounted the many smaller ways in which this scenario plays out.

It’s almost flattering to explain this as a byproduct of our modern, well-connected age: “Look at all this great stuff we have, we can’t even remember it all!” However, the problem is much older than that. Vannevar Bush was motivated to dream up the Memex because he wanted to keep track of the sea of publications released in scientific journals. Older still, we are reminded of Sherlock Holmes, berating Watson for not being more careful with his precious attic-space.

To be better problem solvers, we need information in our domains of interest. This “problem domain” is where we search for solutions. We start searching other domains when we form analogies, especially when confronting novel problems. This means that, just like Holmes, we should be carefully organizing, curating, and probing this body of knowledge. How can our computers help us to do this?

Knowledge, Managed

In the 1980s, there was a software system that set out to do just this. It featured a blank slate that you could draw on, add text to, and link to other documents. You’d be forgiven for thinking that it was some kind of Figma or Canva. In reality, though, it was something quite different. This was KMS (Knowledge Management System), a product of Knowledge Systems. Originally, it was a cognitive science project at Carnegie-Mellon University, but it was later spun out into a commercial project.

The authors stated their goal in a 1988 paper: “KMS is designed to help organizations manage their knowledge.” In their efforts to reach this goal, the KMS team eschewed many of the user interface features that had become standard by that time — windowed desktops, dropdown menus, and more. They started from scratch: if something did not contribute to the creation, navigation, discovery, and application of knowledge, it was out.

Because KMS catered so well to organizations and large projects, you’d be forgiven for thinking that this was purely a professional tool or a 1980s version of Notion. However, KMS was intended as a personal tool, as well. Besides KMS, there was a vibrant community of “Hypermedia” or “Hypertext” tools that developed around the same time. One of them, Apple’s Hypercard, was designed for individual use and still receives high praise. Tim Berners-Lee has recounted some of his attempts to convince other, existing hypertext editor projects to join the nascent Web. These early tools represent a brief foray into personal knowledge management.

Alas, we know the rest of the story. As hard as Tim Berners-Lee and other early pioneers pushed for it, the Web lost many of these priorities as it scaled. In many cases, internet providers (who were vital in distributing early browsers) were more interested in providing access to lucrative services than in providing powerful tools. As beautiful as the Web and modern Web Browsers are, they now only help us to navigate existing structures created by others. What’s more, with the development of our modern Web architecture, KMS and other tools like it dried up. It seems as though the notion of Knowledge Management dried up with them.

Knowledge, Revisited

But then, last week, I read something very interesting. It was a piece by Dan Shippers. At first, I thought it was yet another article weighing in on the eschatology of knowledge workers in the age of AI. However, he used some thought-provoking verbiage. He claims that “people who organize, store, and catalog their own thinking and reading will have a leg up in an AI-driven world.” His thesis is that, by curating a private collection of personal knowledge, they’ll become “gold mines” if we turn AI models lose on them to help us generate insights. He goes on to share a demo of Rewind AI, a tool you can run on your Mac to “find anything you’ve seen, said, or heard” — even conversing about it.

To be clear, I don’t think this is the same thing as generating insights. Rewind seems to be the modern equivalent of raking in “all the lumber” we can find. As an example, I know that I wouldn’t want those thirty useless Stack Overflow threads that seemed superficially similar to my coding problem to interfere with my finding the right one. This is an organizational or ontological problem, though. Shipper’s thesis is still provocative.

In my personal research, I’ve been able to find a few projects that are setting out to do the same thing. The nearest modern versions of KMS and Hypercard exist in the form of extensible note takings solutions, such as Logseq or Obsidian. In the last couple of months, we’ve started to see plugins appear, making headway to use AI assistants as middlemen through which we interact with our notes. Eleanor Konik lists some of them, in this recent post.

This is amazing news. People are interested in their brain attics. I’m excited about the concept of using this personal, highly relevant knowledge to make our artificial assistants that much more intelligent. However, there’s still quite a bit of work to do. In their 1988 paper, the authors of KMS pointed to a fundamental problem that dogged knowledge management tools then and now:

“One of the most pressing problems of human-computer interaction is the ever-growing complexity of software systems. We believe that this complexity is caused largely by the complexity of the underlying data models. The problem is compounded when users must cope with a multitude of different systems that have inconsistent data models.”

“Inconsistent data models.” The knowledge that we’ve been collecting and continue to collect is stored in silos. Do you know where your notes in your Apple Notes app are stored? What about your Google Docs documents? Can you seamlessly link from one to the other? Even if we can get them out of those silos, chances are that they won’t be in a format we can easily wrangle — and that “wrangling” is the important bit. If we hope to turn our collections of knowledge into something that we can comfortably use and gainfully apply, we need to be able to easily collect and connect these disparate pieces.

Stocking the Furniture

The first step is to “own” our data. If we keep our knowledge in a format that is transparent to us, then we can choose the tools we use to manipulate it. There are all sorts of communities that share this goal, mostly for philosophical reasons. You have the anti-DRM folks, the FOSS camp, and pretty much anybody who’s ever used a desktop Linux distribution as their daily driver. The point here is that once we’ve stored a piece of knowledge on our computers, we aren’t inhibited from parsing it.

Formats are the key. Open tools do their best to use well-established, platform-independent storage formats. For instance, open alternatives to Apple’s Notes app often use the Markdown document format — a simple text document with a few extra features. Alternatives to Amazon’s Kindle ecosystem often utilize the industry standard ePub format — making it (mostly) possible to move your books easily from one platform to another. The same is true of many different mediums. In all, between industry standards and open-source tools to work on them, we’re doing pretty well on this front.

The second step is to combine our data, and here we aren’t doing so well. The tools that we use to consume information are often divorced from their means of combination. Here’s what I mean: in writing this post, I consulted the PDF form of the KMS paper and several websites on modern knowledge tools. I created notes analyzing each one. Now, however, if I go back and start reading those websites again, there is no way for me to know or be reminded of my previous work — my analyses and connected articles are hidden from view. Even if I started from my notes, it’s exceedingly difficult to precisely link to a piece of content in a video or a PDF file. The ability to explicitly connect information just isn’t there.

Building a tool that interacts with one of these formats is difficult enough, let alone all of them. I suspect that the key to this exceptional degree of combination will be in building better, more powerful standards. For instance, the URL standard for “text fragments” finally allows us to link to arbitrary segments of text within an HTML document. It’s slowly making its way into modern browsers, but it still doesn’t help us document the nature of our connections. It’s my hunch that, moving forward, the semantic web standard will start to fill this hole.

Conclusion

Just take a moment and think of all of the beautiful, crazy problems we could solve if we get this right. What could we accomplish if we keep track of, find, and creatively surface just the right knowledge at the right time? That’s what a fully-functioning, modern Brain Attic should help us do.

But we can start today. Even while we wait for these systems to mature, there are some steps we can take. We can start by designing personal systems to organize our knowledge. We can find vendors, formats, and storage providers that give us control over our information. This will position us to choose responsible AI solutions that make our data work for us.

With these sorts of systems in the pipeline, the future of Knowledge Management Systems is looking bright. What problems will you solve?

What will you choose to remember?