The first dot-com boom has come and gone and we now find ourselves in the gritty interim period, where only innovation, cunning, and realized profits will save us and bring us back to prosperity. Gone, thankfully, are the lucrative days when anyone with a webpage could call themselves an "ebusiness" and acheive an obscene level of rapport with investors. A lot of new business models have been tried and a handful succeeded, but most failed.

What I thought particularly sad about the situation is that there have only been a handful of attempts to really think outside the box in terms of the computing experience and a user interface. Some of this is understood, since so much of that which was new was being thought up on the web and it was natural to think of the user experience in "HTML mode." HTML being just designed for document formatting (and not a user experience akin to a program), the set of operations we could perform using web applications actually has decreased. There's only so much you can do in the way of novel interfaces if they have to be in HTML and rendered in a web browser. Some creative folks have figured out how to give us a bit more power in these experiences, such as having popup calendars to fill in a date, but support of even such basic interface interactions as "drag and drop" or uploading a folder of information at a time are made impossible with HTML. HTML centrism has really made narrow possibilities for user interfaces; the move to try and shift all local applications to web applications has reduced the potential richness of user experiences.

To its credit, Macromedia has done a lot to attempt to fight this battle and enable genuinely new and interactive experiences on the web. However, such experiences have been primarily focused on entertainment and artistic applications. There aren't very many practical Shockwave or Flash applications out there. Part of this has to do with the fact that although both enable rich interaction experiences, neither is designed as a full API for programming fully functional applications. They make it easy to display data, but not to obtain it or to interact with resources present on the local computer.

Thankfully, there is at least one trend that may bring hope to new user experiences: peer-to-peer (P2P) computing. With most P2P computing models, a program has to sit on the desktop that gives you access to information through a network of peers, as opposed to a web server. Since each P2P network uses a different client access program (and indeed some of these networks have multiple clients!), there are a rich array of opportunities to create novel end-user experiences. We can only hope that some companies will take advantage of this opportunity to experiment with making novel interfaces.

Now what is it that I mean by "novel interfaces?" It's not that I mean novel in the sense of entertaining, but instead, novel in the sense that an interface that is unique, innovative, and takes advantage of the client platform.

Most user interface designs are static. They are taken from the standpoint of the text publishing world - to have an interface that is nicely laid out is a high compliment. This aesthetic, as mentioned above, works well for web design, which is largely a collection of pages with a certain static visual arrangement. (This site is no exception.) It is what I will call a "paper aesthetic."

But paper is not by any means a natural aesthetic. Whether you're a creationist or a scientist, humans weren't built to stare at flat sheets of paper (or, indeed, flat monitors) for days and days on end. It just isn't how we work. We were built to focus in on things that need our attention, keeping aware of our peripheral surroundings for changes, quietly noting subtle changes and becoming alerted to objects moving quickly and visibly.

This concept of motion bringing attention has been used primarily in one place on the web: banner ads. Since in most cases they are the only thing moving on the page, our attention often drifts to them. Taking a look at news.com's new user interface as applied to particular pages, we see that they have restructured their site such that the only motion on the page is dead center and occupies a good percentage of screen real estate. It sits in the middle of the text, forcing your eyes to come to it as the text wraps around it. You can't help but look at it: it's their new series of Flash ads. Naturally, these ads pay much better since the use is forced to view them.

But what if we took some of these base concepts, these notions of what attracts our attention, what distracts us, and what informs us, and created a new style of user interface more closely adopted to an ideal aesthetic - an interface that makes more sense for humans? What would such an interface look like?

First off, in the real world, when we are working on something, we bring it up close to us: it occupies the majority of our field of vision. We get up and close to the screw needing tightening; we push up close to the blueprint to consider each line. The watchmaker does not engage his task from arm's length. Consequent to this, a good user interface will cause the primary activity to occupy the majority of a user's field of view and will clear away irrelevant data to let us focus on the task at hand. While some applications support a "full screen" mode that allows for focus on only the task and nothing else (IE, MS Word, Excel, Adobe Acrobat, etc.), the user experience should allow for every application to be a central focus. Apple made a good stab at this by allowing a user to hide all interface elements not pertaining to the currently run application, including the finder. Most Windows users I know of (including myself) have a very crowded desktop - while in one sense our own fault, it ends up leaving a permanent and distracting cruft on which I must perform all my work. In addition, most menubars and icon trays are always visible and can't be set to auto-hide. This makes it considerably more difficult for a user to just focus on what they've got to do. So the user must be able to concentrate on a single task without being disturbed.

But what does it mean to be disturbed? One example of a bad behavior in this example is Eudora's default action when fetching mail - if you've set it to go grab your email every five minutes and put it in the background, when you do have mail, it whips you out of whatever you're in the middle of, takes you to Eudora, and proudly displays a dialog box proclaiming that you have new mail. This is obviously non-ideal behavior, but it's also not a very dramatic example simply because so many applications exhibit similar characteristics. The classic thing that frustrates me is that if I'm in the middle of wading through programs on the Start bar and some program calls itself to attention, the computer drops the submenu navigation (i.e., the Start bar menu goes away). The computer's designers are clearly making the statement "that which my program has to say is more important than your navigation or input." In this modality, the computer can be considered an ouiji-based entertainment center. You merely consent to the computer's understanding of how you should spend your time - it's not designed to take your interaction seriously.

A computer's crisp, jerkily static imagery is poorly suited to human interactions - we're used to a world of "soft changes" and things that move slowly, with a more or less smooth derivative, instead of dialog boxes that jump out of nowhere and exclaim sounds at us. We should come up with better interaction mechanisms for keeping a user up to date with the status of a system without disturbing them from a task. That is to say, an interface should be designed such that with a cursory look, the status of the system can be determined and changes actively reported, but not in such a way as to be distracted. One way to do this would be through color fades. For example, you could imagine a mail interface where mail was sorted into various inboxes. Instead of having a list of numbers of the quantity of unread messages in each inbox, coloring each inbox in accordance with the importance of the unreplied or unread messages in an inbox would allow for smooth color fades when new messages came in. It would be subtle enough not to distract you from a task, but by glancing over at the colors, you could tell what you probably should look at next.

A database-backed file and email system is also called for. Instead of putting a file into a specified category (i.e., folder), files simply exist and can be filed under multiple categories (a job lead from your friend Kevin about a Linux company could simultaneously be filed under "job leads", "Kevin", and "Linux" without having three copies). Data could also be retrieved via a variety of mechanisms (looking for how recently it was composed, key words in its title, its size, its type, etc.). As an added bonus, if implemented as a "journalling" filesystem and coupled with a bit of data synchronization sofftware, you could guarantee against losing documents and even automatically have an infinite-level "undo" mechanism built in on the filesystem layer, allowing you to version any files on your hard drive. Modern hard drives are sufficiently large (80Gb hard drives now cost $300) that there is no longer any feasible reason why textual, pictoral, and even sound data should ever be deleted. Old copies should be automatically kept around, but quietly so (e.g., they wouldn't turn up in a general search or clutter your views of which files existed, but you could still bring them up explicitly).

The skills of television producers in drawing people's attention to information, actors, and scenes are particularly ingenius; us programmers should learn from them and exploit the tricks they have discovered. In the same way that many computer interfaces accidentally draw attention to the wrong things in the wrong ways (as mentioned above), we should consider how we could purposefully bring the user's attention to certain things to guide and assist their focus. Apple had this in mind when designing how alerts work in OS/X - they pull, transparently, out of the top of a window. The motion lets the user know to look there without surprising them, and the transparent but attached nature of the alert lets the user know that the alert is tied to that particular application.

The main reason why these interface elements have not previously been incorporated into common programs is inertia: with the current set of APIs, it's much easier to piece together a program with usual menus, square windows, and a generally consistent feel relative to other applications (this is a good thing) than something really groundbreaking. (Sonique, with its next-gen UI, has taken years to cobble together.)

Secondly, however, it's only pretty recently that one could assume that an end user's computer would be fast enough to handle smooth transforms, 3D effects, and fades in a complex environment. But with even the cheapest of computers shipping with high-speed 3D accelerators these days, that assumption is being shown to be false. But we haven't really taken the time to question the basic parameters by which we decided upon the current mode of interfacing with a computer whose design really originated in the late 70's and early 80's at Xerox PARC - and that actually originated from Stanford work done in the mid-70's! Clearly, (hopefully?) there is a larger set of possibilities for human-computer interaction enabled by faster and more advanced technologies. Now we only need to harness it and conceive of the next wave of interfaces, in turn to be swept away in another 10-20 years when hardware will permit another rethinking of interaction.

This essay has only touched on a few possibilities for next-generation user interfaces. I plan to update it with more ideas over time, and will amend it here. Please feel to mail me your comments on it or comment below.