Exit Archive

Fall 2005 | v.01 n. 01 | Invitational Issue |

[<] Table of Contents ... [PDF] Print Version

Metaphoric, Invisible, Intuitive: The Adaptive Interface
Michael R Salmond

Center for the Arts, Elon University

Defining Adaptive – The Death of the Desktop Metaphor
Adaptive user interfaces are software artifacts that assist a user in accomplishing some task and, in the process, construct a model of that user's preferences so as to serve him better in the future. Examples include systems that help users sort and file electronic mail, filter news stories, fill out repetitive forms, recommend movies and television shows, select destinations and routes, and construct schedules.1

This approach to personalized services is relatively new but has great potential for improving the effectiveness of interfaces and the ‘invisibility’ of the human-computer connection. However, these ‘agent’ systems may increase efficiency but do little to make the computer interface itself any easier or intuitive. Artificial Intelligence has come a long way and yet computer users still face essentially the same barriers to usability we’ve had for twenty years: the desktop metaphor, the windowed environment and the complexity of the computer exemplified in the interface. In this paper I propose calling for interfaces that not only learn from our behavior but ones that are designed to exhibit complex outcomes through simplified, playful metaphors and environments.

Narrow-Casting and Habitual Reflections
There are two recent examples of innovations in adaptive, or ‘learning’ interfaces: the Tivo digital video recorder and Amazon.com’s product recommendations. Via a set of learning algorithms the Tivo box becomes accustomed to its user’s choices of programs and will actually begin to record programs on their behalf without input. This is not some Asmovian artificial intelligence; instead Tivo works within the realm of habit and genre identification. It does not know you, but it can identify patterns and even suggest or record programs from a genre you watch repeatedly. The Tivo will rarely surprise you and cannot account for truly idiosyncratic tastes. However, the black box does adapt slightly and, perhaps most importantly, does seem intelligent. It works very well as a filter for many and can present its user with items of interest outside of a normal viewing pattern.

Conversely the e-tailer Amazon.com has a subtler adaptive interface. Your Amazon account recommends other products based on your previous searches and purchases: 

Rather than matching the user to similar customers, item-to-item collaborative filtering matches each of the user’s purchased and rated items to similar items, then combines those similar items into a recommendation list.2 

Amazon Collaborative Filter
[>] http://www.amzon.com  

In effect the website acts as a sales assistant with a good memory, allowing you to know about other products similar to ones you have shown interest in. This adaptive feature or agent software isn’t always going to be accurate, but sometimes even knowing ‘What other people who bought this item have bought’ is enough to persuade a customer to purchase more – and that is its purpose. 

This adaptive software is useful for Amazon and the customer as the e-tailer has the ability to offer millions of products. The adaptive solution of collaborative filtering enables the customer to find other products based on their own history and the patterns of others. The irony of the Amazon adaptive agent is that the more you buy the more accurate it becomes. In a world of information glut and choice overload, the intelligent agent, a term coined more in the eighties and early nineties than heard now, is fast becoming a much-required filter. Visually, however, there is little to see. Neither Tivo nor Amazon offers a ‘butler’ front end or three-dimensional avatar for us to relate to or communicate with. The usual visual frippery of a representational emotive character has been ignored and instead they rely on purely hypertextual data.  

These software solutions, while relevant to their respective technologies, rely heavily on interfaces that already exist. Personal computers have advanced, companies are beginning to develop software programs that adapt to our needs and requests, but the main arena of conflict remains; the operating system itself. Our computer-in-office based metaphors are no longer appropriate. Does anyone truly view the OS entry screen as a ‘desktop’; a metaphorical representation of what our computer sits on? It is no less ironic given that increasingly all that sits on our desktop is a computer. What is the relevance of this metaphorical dinosaur beyond acting as an obscure throwback to the past and one that serves only to confuse? Increasingly the office metaphor of folder and file becomes obscured as these terms are now recognized as appertaining to binary data, not paper in an atom-based filing cabinet. When confronted with these redundant interfaces daily, be they Windows or Macintosh, on phones or in automobiles - we are constantly failed in our connection with our technology. The learning curve and ‘feature-creep’ are just too high and noticeable. We have come to rely on an incorrect use of metaphor and our interfaces lack the spark of intuition that humans require to become enmeshed with our technology: “Their constraints have limited the imagination in thinking of ways that digital information systems can be integrated into human life.”3 Operating systems, though much advanced, actually work in much the same way as in their primitive youth. Our systems do not learn or adapt to our needs, we as users are forced to adapt to them. 

Microsoft Windows 95

Apple Macintosh OS9 

The promise of adaptive interfaces and software is one that allows us, as end users, to work and play in ways best suited to our individual needs. Customizable interfaces are a start but one that ultimately frustrate and are of little use to the ‘newbie’ or uninitiated user. Changing colors or desktop wallpaper does nothing to reduce the frustration of computer use. Our machines need to learn from us. We require interfaces that seamlessly merge and blend our idiosyncratic selves with our technologies. This has happened before, with television and the telephone. The interfaces are now so easily defined and understood that we forget they are there at all, as Nicholas Negroponte stated back in 1995 they become ‘Invisible’. Not by becoming more complex, but by becoming more minimal, his hope was “that computers will become more like people”.4

Learning to Play
Such advanced interface simplification is best observed is within the video game industry. This is an industry that prides itself on non-existent (or extremely simple) instruction manuals:”Most interface design techniques focus on making the learning phase as short as possible.”5 This is not to say that the games themselves lack depth – far from it. The aesthetic and penetrating visualizations are intense but it is in our approach to the interface that, I feel, allows us room to experiment and learn. Play and sleek interface allow us a much more efficient and effective interfacing with our technology. Whilst it may be difficult for us to imagine turning Mario, Doom or Half Life into a spreadsheet interface, the way is clear for a more vital, visual representation of data and computers. We can access our technologies in a more engaging way, once we remove the barriers of functionality and formality that prevent us from learning to play. There is no better way of knowing someone than by watching them play or by playing with them. Hence, the adaptive interface and intelligent agent could learn faster and more intuitively if our interaction is based less on an out-dated business metaphor and more on one of game play. Video games have a visceral immediacy to their interface designs, but even within the framework of a game there is still the issue of continuity across products and the ‘language’ of the game-space consistency.  

Civilization III –Firaxis Games
[>] http://www.civ3.com

SimCity 4000 – EA Games
[>] http://simcity.ea.com

Over the years interfaces have continued to evolve, being driven by the most popular and innovative games. Although video games rely on metaphors, these metaphors are constructed from real-world models and as such diminish the learning curve for the player. “Making invisible interfaces does not mean taking buttons and commands away from the screen, as doing so would make the game hard to play. A better goal would be to substitute every control that can break the game's sense of immersion with an adequate metaphor. If the metaphor we choose integrates nicely within the game environment, the game atmosphere will be greatly improved.”6 Adapting this immersive and minimal approach towards the personal computer could deliver us a much more ‘realistic’ and intuitive metaphor and kill off the desktop environment for good.

We already have a pseudo-model for this new interface, in massively multiplayer games. Inhabiting a three-dimensional world a user is able to store items in realistic ‘physical’ environments within a game world and retrieve them at any point in time. Games such as EverQuest and online environments like Second Life are examples of virtual spaces that become a form of operating system and interface. Why then, have any form of visible ‘external’ operating system  (windowed environment) when we don’t have an external interface in ‘real life’? In effect the metaphor becomes all encompassing: “Interface researchers believe that the future lies in adaptive interfaces and computer systems built firmly on psychological understandings of users and anthropological understandings of social settings”7 

Taking this concept one further, when playing any open-ended game where one communicates with others in social and anti-social settings, a player imbues their online avatar with much of their personality and ideals. Although we control the avatar, there are simplistic cues in place today that allow our online representatives some autonomy. Once developed further the avatar in game space could begin to represent us in ‘meat-space’ as well as in the virtual realm. The adaptation of the Massively Multiplayer Online Game (MMORPG) genre to its inhabitants and the evolution of that game-space based on feedback, demands and player suggestions allows us to see a future potential for the networked self. We tie our ‘character’ or avatar into our phone systems or entertainment choices and they become indistinguishable from us: our agent and interface between worlds.

Re-examining Our Input
Adaptive interfaces that learn our needs and desires are one thing, but with the added subtlety of gesture, a shrug or postural nuance, the software becomes an even sharper reflection of self. Gestural video game interfaces, such as the eye-toy or certain VR Gloves are making their way into more homes. 

Black & White: Lionhead Studios
[>] http://www.lionhead.com

HandVu : Kölsch, Turk, Höllerer, DiVerdi.
[>] www.cs.ucsb.edu/~matz/HGI

These gestural peripherals have the advantage of familiarity and again the element of play. Although body language reading agents may be a way off, avatars in certain online environments already have access to a range of gestures and reactions that begin to communicate deeper emotions. 

Adaptive interfaces have the potential to offer us something that we intuitively understand. Our software needs to better serve us, to really learn from us and develop along with us. Already as computer users we are so deeply embedded in the desktop metaphor environment that to change it seems a conceptual chasm. Changing the way we interact with our technology suggests a new era of convergence between humans and technology: our hardware has finally caught up with our imaginations.

We must set down plans and roadmaps for this new interfaced environment, or we are doomed to forever languish in windowed-environment hell. The models and examples exist, but we must make the conceptual leap allowing us to embrace not only the concept of play, but of new dimensional metaphors and simulacra. The reliance on old metaphors and two-dimensional interfaces seems incongruous to technological progress: why not make computer interfaces as intuitive as a doorknob or pencil? The concept of metaphor is still solid, it’s just that we’re using the wrong one and need to move on.

1 Langley, Pat and Melinda Gervaso. (2001) http://www-csli.stanford.edu/cll/aui.html

2 Linden, Greg, Bent Smith and Jeremy York. (2003) “Amazon.com Recommendations: Item-to-Item Collaborative Filtering”, www.computer.org/internet p76-80

3 Wilson, Stephen. Information Arts (Cambridge, Mass: The MIT Press, 2002)

4 Negroponte, Nicholas. Being Digital (Reading, UK: Hodder and Stoughton 1995)

5 Sanchez-Crespo Dalmau, Daniel. (1999) http://www.gamasutra.com/features/19991108/dalmau_04.htm

6 Sanchez-Crespo Dalmau, Daniel. (1999) http://www.gamasutra.com/features/19991108/dalmau_04.htm

7 Wilson, Stephen. Information Arts (Cambridge, Mass: The MIT Press, 2002)

Copyright © New Media Caucus. All rights reserved.
[ www.newmediacaucus.org ]