Dissappearance
Sensors
Ad Hoc Networks
Probability and Stochastics
Proactive Computing
Interface
Social Networks
Social Computing
Emotion
Education
Flow

Lighting and electricity are the models -- technologies that we don't notice unless they go wrong. "The most profound revolutionary technologies are those that disappear. They weave themselves into the fabric of everyday life until they're indistinguishable. The web is an excellent example of such technology -- it's no longer exciting, because it has become part of our life." (Vincent Tao, in this article.) But of course the web has not dissappeared; it is still a primarily textual medium that requires us to focus our attention on a small screen. This will change, in several ways. Greater bandwidth will enable more and better quality (technically anyway) video, audio, and telephony. The multiplication of raw, real-time or near-real-time data uploaded to the Internet will require new ways of sorting, interpreting and visualising it all. Ambience is about offloading the visualisation of data from the small screen to new, peripheral ways of monitoring -- perceptually available but invisible in use. This is an important distinction. "At the very heart of the vision for Ubicomp, the notion that 'computers...vanish into the background' [30], lies a serious problem for interaction, which is communicating to the user which objects the potential for possible action is embedded in." (Belotti et al, 2002).

Tolmie et al (2002) describe this as "unremarkable" computing. Augmenting artefacts and/or spaces, as they point out, may fundamentally change them, or require them to be completely redesigned. Perhaps it will change them for the better; just because we are accustomed to ways of doing and using things does not mean that these are the best ways. Is it necessary to adhere to everyday routines, or might we simply adapt to new ways of use?

With large amounts of small data collectors scattered around, some of which may move from place to place with people, animals, or objects, static base stations cannot be relied upon for communications. Instead, each device can become a node in an ad hoc mesh network, in which data jumps from device to device until it finds Internet access. If sensor-enabled objects and/or spaces can collect and share some information about their state and that of the environment, what results is a kind of emergent intelligence. There are myriad variations and implications, but the point is that objects and spaces become media, and the interpretation of their "content" becomes a subject of interest. This emphasizes and broadens the study of media ecologies. See Mattern, 2003.

A similar idea is Howard Rheingold's concept of smart mobs (not to be confused with "flash mobs"), which works in an identical way except that the nodes are not devices but people, where each is a network node. William Mitchell carries this idea further in his new book Me++, literally defining people by their relation to networks: "I am part of the networks and the networks are part of me. I am visible to Google. I link, therefore I am." (Quoted in this article.) This may sound silly, but in a world with no distinction between online and offline, the virtual world weaves further into the fabric of reality.

With large amounts of complex data from sensors and elsewhere, delivering results as probabilities represents a new model directly opposed to the traditional, logical model of computing. This view is held by David Tennenhouse at Intel (see this article). In practice, this is as simple as multiplying each input by some number to weight its strength; if the strength passes a certain threshold, it is measured. The weight can be derived from signal strength, or some other factor related to the characteristics of each input. Taken a step further, each weighting can be related to preceding inputs.

Stochastic, rule-based programming such as is used in particle simulation or the a-life algorithms of Stephen Wolfram may be useful for deriving emergent knowledge from complex data. "Such systems as we know can generate, under certain conditions, self-organization phenomena at a macroscopic scale in the form of observable patterns (spacial patterns, temporal rhythms,...)." (Mathieu, 2003) In contrast to traditional, rigid programming, rules give freedom and support creativity.

Part of making computing more "human-centered" may mean giving up more control to them. While this may immediately stir fears of "intelligent" machines dominating humans, here it merely means computers interacting with eachother autonomously -- not consipring, only sharing data and processing. "Proactive" stands opposed to "interactive" computing, the goal being to minimize the need for us to deal explicitly with them (David Tennenhouse talks about this here). In fact, I have always been deeply skeptical of "artificial intelligence" efforts which seek to emulate human thinking or behaviour. Humans and computers are simply not good at the same things, and to our detriment we have created a world in which computers demand precision and memorization from us -- the very things we are not good at. Don Norman has an excellent chapter on this in The Invisible Computer, which you can read here.

The ambient model transcends traditional notions of computer-human interaction because the computer is merely mediating or enhancing human interaction with the physical world. Ailisto et al (2003) identify three distinct cultural differences in approaching interface design for ambience: The European approach focuses on making the technology, including the interface, disappear as much as possible; this is probably directly related to the EU's Disappearing Computer initiative. The American and Japanese approach is to use existing display devices such as televisions and PCs. The Nordic approach is to use primarily mobile and location-based devices such as customized publicly accessible terminals. This division is not rigid. American researchers, notably Hiroshi Ishii, have pioneered naturalistic, tangible interfaces which are meant to "disappear" into their surroundings. And Nordic researchers such as Tobias Skog and Johan Redstrom have utilised fixed, wall-mounted displays as "informative art."

In terms of individual interaction with data, in a world where computing has disappeared and we are free to move about in an informaion space, the concept of interface is necessarily broadened to a more human scale. "If the body is our general medium for having a world, then we are, as individuals, condemned to meanings that are more or less finite and local, measured by scales appropriate to our earthly bodies." (Arike, 2001) "Our worldview relies as much on the body's senses as it does on the environment itself. Standing, we view the world at eye level. Our posture establishes verticality by day, horizontality by night. We see, hear, smell, even taste with a frontal bias, owing to our physiognomy. Up and down, left and right, front and back describe our orientation to the world....We come so much to expect some sensations that they become transparent to us: the pressure on our feet as we walk, sunlight overhead, the horizon uniting earth with sky....Too much new information causes distraction and stress. Too little will at first calm, then bore us. This knowledge is useful in the creation of information environments." (Anders, 2001)

Knowledge doesn't come from machines, it comes from people. Computers may be able to derive emergent patterns from complex datasets, but humans are much better than computers at producing reliable knowledge. This is why the Amazon model works so well -- it produces knowledge based on the actions and opinions of other people. Instead of trying to emulate the human brain with computers, it is much more productive to harness the global collective consciousness. "The revolution we are in right now is not so much about the digital revolution, the computer revolution, the internet or the telecoms revolution. The revolution is the social interaction revolution and it is all of these things put together in one." -- Don Norman, quoted in this article.

Related to this is the notion of narrative. This is what makes projects such as Annotate Space or Urban Tapestries interesting -- the focus is not on the technology and what it provides so much as on the content, the stories that other people provide.

A related concept is social computing. The notion of "interactivity" is based on a 1960s model of one-to-one human-computer interaction. This is seen as a dialogue in which the human issues a request and the computer is meant to respond appropriately. An ambient computing environment is likely to be populated by more than one person at least some of the time. My experience in putting technology in museums and other public spaces has shown me that much richer forms of interaction and knoledge creation result from multiuser experiences. Christian Heath has written a good article about this, available here.

Social interactions are necessarily framed and constrained by physical space (architecture and urban planning); by culture, and by context. "The persistence of buildings reveals and reflects patterns of cultural behavior....For example, an elaborate picture frame ill-serves a subtle painting by drawing the eye from the canvas. Similarly, the distractions of ornament and detail compete for attention with focused activities in architecture. For this reason, house lights go down at the start of a play, focusing our attention on the illuminated stage. Electronic visual media also affect the status of objects through changes of state. An unplugged television is quite different from one conveying a program. While the former is a rough box of plastic and glass, the latter isffectively a hole in space, its screen is a window cast onto another space entirely." (Anders, 2001). I would add that the screen also has the ability to transcend time as well. A projection, Anders continues, has the effect of breaking down an entire wall. "For this reason, the defining role of architecture as a frame for social conduct is challenged -- if not obviated -- by the virtualization of its surfaces."

Even beyond multiuser experiences, social science research provides a valuable foundation for computer-human interaction involving ambient technologies. The work of Erving Goffman, in particular, sheds light on human-human interactions in various contexts. See also Tolmie et al (2002), who use ethnographic methods in CHI research. "Sometimes what is 'natural'," they write, "is highly situated and thoroughly social."

A related concept is "interactionist AI," which is differentiated from traditional artificial intelligence in that it does not imbue technology with a preset body of knowledge but with social knowledge -- knowledge on how to learn from interactions with people, other machines, and the world. It views knowledge as external instead of internal, modeled on social insects. Traditional AI often ignores the body, while interactionist AI recognizes that a body lives in the physical world. For more see Mateus, 2001.

I implied above that computers are not good at emulating human emotion and creativity. I believe that. However, this does not mean they cannot deliver these things. As a medium, the computer can not only access large databases of human-created video, audio and other media; it can also process and mix these media to create new and emergent works. Perhaps surprisingly, this relates directly to learning, for emotion profoundly affects memory. Don Norman has radically suggested (here) that music can enhance learning, in the same way that it adds emotional power to movies; he is writing a book on "emotional design". Research into computers and emotion is a new area, but sure to produce interesting results. My colleague at the Institute of Education, Pam Meecham, is doing related research.

In a recent paper presented at the 1st European Symposium on Ambient Intelligence , M.A. Neerincx and J.W. Streefkerk described an experiment in which they showed people films meant to invoke various emotional states, then measured their performance in using PDAs and laptops. Unsurprisingly, performance varied depending on the emotional state.

Here is a simplified model of a generalised learning process: Data --> Information --> Knowledge --> Creativity/Synthesis --> Theory --> Practice. A more learner-centered model is perception --> awareness --> reflection. Learning is constructed actively in interactions between subject and object (Piaget, 1954), influenced by cultural background, context, and the media and tools employed. (Vygotsky, 1978) Knowledge without context loses meaning; individual bits of knowledge mean little if disconnected from other bits to form an ontology.

Current e-learning systems make use of discreet packages of knowledge called learning objects. This assumes that knowledge is something static that is transmitted from one person (or computer, as the case may be) to another, roughly following Claude Shannon's (1948) linear mathematical theory of information transmission. But of course knowledge is not static: history is constantly rewritten, scientific theories are updated and replaced, languages and cultures evolve. Learning objects must be continually updated based on new data, new theories, and new interpretations. Yet e-learning is often heralded as a fix to our "broken" traditional educational systems. Perhaps the ambient paradigm holds the key, for it is a bridge between the physical and informational worlds. It harnesses real-time and ever-changing data, and views knowledge not as fixed but as process. In practice, Wisneski et al (1998) concluded that the more a person uses an ambient environment, the more they internalise its workings, like driving a car (though theirs was an early and limited experiment). Maybe devices are the wrong focus, and getting a computer into every student's hands is the wrong approach; maybe the focus should be on computing and not on computers. When the environment is a computer and we are living inside it, computing becomes a physical and social act, not a cerebral and solitary one. Like e-learning as it is currently practised, an ambient educational environment cannot replace a good human teacher. But it may be better suited to supplement him or her.

The promise of digital technologies, as yet unrealised with screen-based presentations, is that they can reveal or convey information or experiences normally invisible or impossible in the real world. (see Price and Rogers, 2003; Resnick, 2000). Physical or ambient technologies can act as a natural interface

How best to put the idea into practice? Research into prior uses of media in education gives some clues. In an early study of multimedia in education, Alty (1991) matched particular media to types of learning tasks -- audio for imagination, text for details, and so on. Rogers and Scaife (1995) subsequently found that the separation of media in a multimedia presentation makes it more difficult to cognitively integrate the various types of information to form knowledge. There has been much debate about the merit of various representations of information, with text being compared to everything else, in terms of "computationaln offloading" -- not referring to a computer but to cognitive processing. For example, the pocket calculator has been alternately embraced and abolished in schools. In fact, as Eco (2003) points out, such debates date back to the introduction of books themselves, which, when introduced, were seen by some as the end of creative thought brought about by paintings and oral tradition. But of course oral tradition, or any of it succeeding traditions, has not died out; it adapts and is re-created in different forms.

If there is anything close to consensus, it is that technology should be kept at arm's length during the formative educational period, until "the old ways" can be learned. Yet this contrasts with the simultaneous push to get computers into school, lest students be left on the wrong side of a perceived "digital divide." Such debates are not likely to be quelled by ambient media, but because of the flexibility of the new technologies we are, at least, in a position to specify just how we would like information delivered, and it could potentially be served in subtle, natural ways that do not inhibit cognitive development.

One often overlooked aspect of e-learning is feedback. I refer not to communication between teacher and student or among students, but between student and computer, or more properly in the ambient era, between student and content. In keeping with the constructivist paradigm, learners can and should actively construct their own knowledge. Yet educational content is often served up for absorbtion, later to be discussed or assessed. The multi-media delivery of information that computers enable is only half of the story; the other is the computer as an input device, and a processor of information and media. In fact, this is not limited to electronic media, for even printed textbooks are not meant to be written on, yet students highlight, underline, circle, draw, and write marginal notes in them. Computer-based learning materials seldom include this ability, beyond a notepad in a separate space. The carefully scripted and designed presentation and the rigid software that runs it are not to be tampered with. But the great advantage of computer-based media is not its ability to serve up various representations, but the ability to process them based on input. Ambient technologies broaden the computer's input capabilities beyond the mouse and keyboard, encompassing the whole human body, groups of people, and limitless objects and processes in the real world.

Directly related to this is an inverse learning process -- applying theories and methods of learning to machines. This is the adaptive and intelligent part of Ambient Intelligence, and it refers to computers which are trained instead of programmed (see Mathieu, 2003). (A broader view of this is evolutionary.) One approach to this challenge is to utilise theories of childhood cognitive development, such as Piaget's. The advanced stage would be pattern recognition. "These 'patterns' of structured data appear at a meta-level and are triggered by collective variables at the lower level." (Matheiu, 2003)

Directly related to learning is the concept of "flow," as defined by Csikszentmihalyi (1991), and in a learning context this amounts to just the right amount of challenge (see Norman, 2002). Castells (1996) defined our networked society as a "space of flows," blurring real and virtual realities. These two bases provide a foundation to put ambience into practice. In this excellent 2002 article, John Thackara provides a blueprint for managing the chaos of data, for designing flows and ways to perceive flows. "Firstly, in order to do things differently, we need to see things differently. We need dashboards for cities and buildings. We need to experience the systems and processes on which we depend, in order to look after them." This is moving from designing things to designing systems, with a focus on process.

A wireless color-shifting abstract decoration
A low-cost implementation of Ambient Devices' Orb