Over at KMWorld, you can read a great interview with David Weinberger, author of the new book, Everything Is Miscellaneous.

I haven’t read his book yet, but some of the insights contained in the interview are very dear to my own heart:

“…if your business depends upon information, as all businesses do, then by using tools that allow that information to be broken out of its assigned categories, you will discover relationships you didn’t know were there. You’re going to spur innovation, you’re going to discover efficiencies and you’re going to enable people across your organization to find other people who share their passions…”

Within the enterprise, Our traditional approach to information management (if you like, the “Enterprise 1.0 thinking”) is that “Whoever owned the information organized the information.”

Why? The answer comes from paper, perhaps still the primary information medium for the worlds information. Paper could only exist in one place, at one time. It wasn’t possible for the consumer of the information to determine how they wanted to consume, classify, or present it. And so, a whole discipline of information management sprung up around the way that we classified physically finite things. That Dewey bloke has a lot to answer for…

But digital information is different. It can be re-purposed, redesigned, and integrated into all kinds of different contexts. It can exist in multiple (or infinite) places at once. As an example, I’m pretty sure my elementary school librarian wouldn’t have let me rip all the pages out of my choose your own adventure novel so that I could lay them all down and make a story map of the path I chose – but I could easily do that with digital information, and in real time, without affecting the thirty other people who were also reading the novel. Surely then I could figure out how to avoid being eaten by that pesky snowman…

This one crucial difference – the ability for digital information to be stored once, and then represented in many user definable, varied contexts is perhaps the most exciting notion amidst the Enterprise 2.0 hooplah.

(Turn to page 4. )

Continue Reading

Lots of people have asked us, “So what exactly are you guys doing?”

We’ve been pretty cagey when it comes to answering that question so far, because – well, when all your company has is a great idea, you can’t help but feel a little protective of it.

But Tim O’Reilly gave away some great clues two weeks ago at the Web 2.0 Expo in Berlin. You can see the slides from his presentation below.

Thanks Tim – we can save ourselves half an hour of explanation now. We’ll just ask folks to watch the slideshow!

Continue Reading

The New York Times reports that Thomas Edison’s original DC Grid in Manhattan is finally being switched off.

As Nick Carr suggests, information systems have layers of history, just like their mechanical engineering ancestors. Nearly all of the enterprise solutions I’ve built have involved building a new solution on top of, around, or at least loosely connected to an existing one. Usually these were big, scary mainframe systems, written in COBOL, and most seemed to do batch financial transactions.

While it’s nice to know that good ideas tend to stick around, this baggage can also slow the rate of adoption and the pace of innovation. You’d think there would be higher turnover in enterprise systems.

Why are legacy systems such a prominent feature? Some possible theories include:

1. New ideas grow from old ideas
   “Standing on the shoulders of giants”
   Marketing nowadays seems so obsessed with ‘innovation’ and ‘differentiation’ that it often seems as though every product is a brand new idea that was never thought of before. This is patently untrue. As an example, we need look no further than Amazon’s recent Kindle announcement. They’re trying to market e-books to people, using an electronic reader. Their inspiration? The actual book (A-Book?).
2. People hate change.
   “You can’t teach an old dog new tricks”
   On the whole, people really don’t like it when things change. While this cliche is perhaps a bit too pessimistic overall, it’s certainly true when it comes to the workplace. Remember that a lot of people aren’t primarily working because of their passion for whatever menial task they do. They’re in it for the money. Employees might feel that having to learn a whole new system is a waste of time. It might make people feel uncomfortable with a job that they felt they understood just fine. So, in order to appease people, new solutions are ‘phased in’.
3. Because they don’t need to be replaced.
   “If it ain’t broke – don’t fix it”
   Newer isn’t always better. There are numerous reasons for adopting a new system, and a near-infinite range of human motivations for making a change. Sometimes these systems are actually made worse. As an example, consider Microsoft’s OOXML format, released with Office 2007. Initial investigations indicate that most people are still saving their documents in the old, .DOC format. The reason for this is that the documents look exactly the same, and they can share them with lots of people. If Office 2007 didn’t support the new format, I suspect that nobody would care very much.
4. Because in some niche somewhere, someone is wedded to them.
   “Where you stand is where you sit”
   It’s easier to sell a system to a select few — and have them defend it and represent it — than it is to sell to the masses. When the new direct current electrical grid came out of beta in 1882, companies interested in profiting from a good idea began building for the new power system. The resultant ecosystem of companies and customers proved pretty hard to shake — there are still elevators in Manhattan that run on DC power, 125 years later. Enterprise systems are platforms, and they become the part of the foundation of the organizations that own them.

Which ever way you look at them, legacy systems are a crucial part of the way enterprises work. Part of me wants to assume that eventually, one brand new shiny solution will rule them all, but the reality is a long long way from that. Just as we needed the initial electrical grid to show us what was possible in 1882, we need these older enterprise systems to help us grow and improve the work that we do.

Continue Reading

Tree view controls were very much in vogue when I first learned to write code, some 10 years ago. They’re a common user interface convention that still features heavily in software:

I don’t like tree views because they tie you to a hierarchical world. Every element has to be described relative to its parent — which assumes that each piece of information has one direct ancestor and potentially multiple descendants. So if I put my album collection into a tree view, all my song files would be directly related to the album folder.

This is all well and good, but it assumes that the only way I care about organizing my music is by album. But I’m a complex individual. I classify my music in many different ways: music of a particular genre, music I like to listen to in the mornings, etc. Sometimes I use less tangible criteria that I can’t really explain to iTunes — like music I like right now, or music that a friend recommended to me that I think might suck but I might listen to later…

When you consider all the ways that I think about music in my head, the two-dimensional tree view seems remarkably quaint. It will only tell me that each track exists against a particular album. Which is nice, but, well — you know…

The problem here is something that academics call multifaceted classification. Simply put, it means classifying things in lots of dimensions, not just one. This creates a huge problem for our two-dimensional tree: Suddenly each track could be in thousands of different categories at the same time.

The Web 2.0 way to approach this problem is with tags. Just stick a tag on all the things that you care about: jazz, morningMusic, etc. This works well because you can tag multiple things with the same tag and put multiple tags on the same item. This is a form of multifaceted classification.

The drawback comes from the fact that this tag-based classification is hard to fit in your head. You can’t display tags very nicely. (The tag cloud is about the closest you can get, and it doesn’t really show you anything of much use.) A tag itself doesn’t have a concept of timeliness or uniqueness. I’m forgetful, so I mis-tag things. (Did I use blog or blogs last time?) Sometimes what seems like a great tag never gets used again, so I have lots of orphans and one-offs in my del.icio.us account.

Computers can do math in multiple dimensions — that’s not the hard bit. The hard bit is trying to fit the whole thing onto a two-dimensional screen in a way that doesn’t hurt people’s heads or require them to get a library science degree.

Right now, infovark is thinking about the best way to solve these problems. We know that people intuitively understand multifaceted classification, even if they don’t know what it’s called. I do multifaceted classification with my music library and I could explain my system to you if I tried. People seem to cope just fine with multiple ways of organizing things. We want our software to be able to do that, too.

At least, that’s our challenge for today. If you’d like to check out some newer approaches to information visualization, head over to Information Aesthetics — they seem to always have great stuff!

Continue Reading

I couldn’t resist checking out the Nerd Handbook on Rands in Repose this morning. If you’re a nerd, or have a nerd in your life, it’s a funny article. If you don’t have a nerd in your life, you can just imagine me instead. (Case in point: I began reading Rands’ blog after reading his essay in Joel on Software’s Best Software Writing I. That’s nerdy.)

Rands asserts that nerds have an efficient relevancy engine inside their heads. This relevancy engine can be extremely annoying, since every item of information given to a nerd has to pass an “Is it interesting?” test before it is remembered or acted upon. And as we all know, a nerd’s definition of interesting can be quite different than the rest of humanity.

While many nerds suffer from having an extreme form of this relevancy filter, I think all of us have one. Human brains are wired to be pattern-matching machines. It’s what lets us recognize faces from a block away, a song from a handful of notes, or the darn alarm clock by fumbling in the dark. We can even see patterns where none exist, which can cause us to believe in things like the abominable snowman or Nessie. We’re naturally good at separating the signal from the noise.

Unfortunately for anyone that works with computers, this is precisely not the way that computers think. When people first built computers, they built them to compute. Performing complex calculations is a tedious and error-prone task, so we created a tool that was very good at it. This allowed scientists, engineers and mathematicians to get on with the more interesting parts of their work while letting the computer crunch the numbers.

Despite all the fancy bells and whistles that have been added over the years, the heart of a computer is still its central processing unit (CPU). The CPU is really nothing more than a souped-up pocket calculator. It expects detailed instructions about what to do, and given the right instructions and accurate data will produce a correct result.

(For some reason, I have this Kraftwerk song playing in my head right now….)

As I mentioned in a previous post about emergence, one of the main problems with enterprise software today is that it is built for the way computers think, not for the way that people think. Most enterprise tools assume that the user can articulate a detailed series of steps to follow and provide good data. But today’s knowledge worker is not merely crunching numbers or following standard operating procedures. In fact, we’ve engineered most of the raw data processing out of people’s jobs, shifting it to the computers that do it so well.

So knowledge workers need tools that help them do the kinds of work that they do today: research, analysis, synthesis, and composition. These tasks are all relational in nature. They require us to do things we’re naturally good at: find patterns, weigh evidence, determine relevance and execute judgment. And that’s where the problem lies when it comes to enterprise software. Computers just weren’t designed to help with these sorts of tasks.

One example: Recent research has shown that bees are good at facial recognition, even when the face is partially obscured. Computers struggle with this problem; it takes powerful, modern computing hardware and advanced software to approximate the accuracy of a tiny insect. In fact, it’s enough of a struggle that it made the list of Grand Challenges. The reason? Pattern matching and number crunching are fundamentally different tasks. Just as it’s hard for humans to repeatedly execute a complex series of calculations, it’s difficult for computers to draw inferences and determine relevance.

This situation leads to a Catch-22 in enterprise software. Since knowledge workers intuitively grasp patterns and relations, they often have trouble articulating their thought process or decision model. But since computers are best at executing instructions, knowledge workers are forced to lay everything out in meticulous detail. The end result is that both the software and the employee wind up doing tasks that they aren’t very good at. Is it any wonder people find working with enterprise software frustrating?

To fix the problem, we have to make software that either thinks the way we do or that can leverage the power of our power of our wetware — our relational, pattern-matching brains. Would your computer associate enterprise software with experimental ’70s synth-rock? No, but I’d bet I’m not the first person to do so.

(Yeah, the other one was probably also a nerd. )

Continue Reading