Pointable Computing
07.12.02
Abstract
 

Imagine a room filled with electronic devices—appliances, thermostats, computers, portable music players, PDAs—all of which can communicate wirelessly. This may not even require an act of imagination. Now imagine yourself with a handheld device in the middle of all this wanting to use it to communicate with just one of these devices. How will you do it? Will you select its name from a menu on a screen? Will you tap on an icon? Either way your attention will be focused on the device in your hand, not the true target of your intention. Your selection will be a selection by metaphor. Why not simply point at the thing you want to communicate with? Pointing is a natural extension of the human capacity to focus attention. It establishes a spatial axis relative to an agent, unambiguously identifying anything in line-of-sight without a need to name it. This brings our interactions with electronic devices closer to our interactions with physical objects, which we name only when we have to. Pointable Computing proposes that a visible laser be coupled to a handheld device and cheap sensors attached to potential objects of communication. Information may be conveyed along a laser beam, establishing a channel of communication along a perceivable axis of pointing. Ultimately the goal of this project is, through a subtle change in the focus of interaction, to foster decentralized computation in which individual computational objects scattered in the environment may be controlled directly rather than through a single hub. This kind of electronic autonomy curtails the ability of faceless authorities or proprietary standards to shape the future of computation.

 


In a working prototype, Word Toss, two handheld devices are able to send short words back and forth on visible laser beams.
 
 

Motivation

When I started using computers, they did not have mice. I felt close to the machine because I learned to understand all the layers between the hardware and me. When GUI-based operating systems hit the desktop, they completely unnerved me. Suddenly the computer that made sense to me was couched behind a semi-spatial metaphor that elucidated neither its workings nor the nature of space. I have, of course, come to accept it. I have taken the time to try to understand the new layers of operation that have been interposed between me and the machine. But fundamentally, the spatializing of computational objects in screen space remains an affront because it engages the user as a two-fingers and a wrist. Anyone who has ever spent ten perfectly good seconds looking for a cursor that she purportedly controls with her hand has felt the difference between screen space and real space.

My work takes computation away from the screen and into the realm of full physical involvement. It is by no means the first effort to do so. The difficulty and expense of tracking points in space or the vagaries of computer vision have hampered many previous attempts. Many of the successful prototypes require very strict conditions of operation or a special setup not widely available to the public. By contrast, Pointable Computing requires only a very cheap and intuitive layer of technology to be applied to a handheld device and an object in the world, and derives from those a robust and useful spatial relationship without calibration.

There is an irony in the use of the words “active,” “interactive,” and “reactive” to describe computational objects—both physical and virtual. It is a common practice, as though nothing had those qualities until the computer waltzed in and started endowing ordinary objects with buttons and microphones. The truth is that non-computational objects are far more active, interactive, and reactive than any working computational version of the same. The reason is that in order to consider an object computationally, we must derive data from it, and that means outfitting it with sensors in some way. As soon as we do that, we chop away all of the interactions we have with that object that are not meaningful to the specific sensor we have chosen. No matter how many sensors we add, we are taking a huge variety of interactive modalities and reducing them to several. How could a simulation of a cup ever be as interactive as a cup?

Some argue that adding sensors to a physical object does not constrain its existing interactivity, but augments it electronically. I believe that is true as long as the object remains primarily itself with respect to the user and does not undergo some metaphoric transformation into a virtual representation of itself (visual or otherwise) or into a semantic placeholder. That is difficult to achieve, and cannot be done as long as a user must consult a secondary source to determine the quality of his interaction. To check a screen or even to listen to a tone to determine the pressure with which I am squeezing an object supercedes my own senses and reduces any squeezable object into a pressure sensor. In order for a physical object to be augmented rather than flattened by computation, the computation must occur (or appear to occur) inside the object and the consequences of the computation be registered by the object. This is a strong argument for ubiquitous computation, which challenges the supremacy of the desktop computer, placing localized computation in objects distributed throughout the environment.

Pointable Computing takes as its starting point an emerging reality in which everyday electronic devices communicate wirelessly. These devices already have identities tied to their functions, be they headphones, storage devices, or building controls. They are not crying out for an additional layer of interface. How can we address the new capacity of things to talk to each other without further mediating our relationships with them? We need the remote equivalent of touch, an interaction focused on its object and containing its own confirmation. Pointable Computing offers that by way of a visible marker, a bright spot of light. You do not need to consult a screen to determine if you are properly aligned. It is apparent. The receiver may also indicate that is has acquired the beam, but that indication will always be secondary to the visual confirmation that the object is illuminated.

As Norbert Weiner pointed out, any system containing a human being is a feedback system. As a user, a person automatically adjusts his behavior based on the overall performance of the system. What makes the Pointable Computing a robust communication system is that the feedback loop containing the human being is direct and familiar. The human eye has an area of acuity of 1–2°, implying that narrow, beamlike focus is the norm, not the exception for human perception. The rest of the visual field is sampled by eye movements and then largely constructed in the brain. Tight visual focus is the way we solve the problem of reference without naming in a spatial environment. The feedback loop that enables the act of looking entails our observing and correcting our proprioceptively-sensed body attitude to minimize error of focus. It happens so quickly and effectively that we do not even notice it. The same visual feedback loop can be applied to a point of focus controlled by the hands. It is not quite as immediate as the eyes, but it is close. And, as it turns out, it doesn’t suffer from the kinds of involuntary movements that plague eye-tracking systems.

I have developed a few use cases to demonstrate the potential of Pointable Computing:

Universal remote
The most obvious use of Pointable Computing would be to make a universal remote. Pointing the device at any enabled object would turn the handheld into a remote control for that object. On the face of things, this seems to be a rather mundane application, and one that seems to run counter to the program of endowing objects with individuality and escape from metaphor. But consider a simple control consisting of a single pressure sensor. This might be useful as a volume control or light dimmer, for instance. Since there is nothing on the control to look at, the interaction with the controlled device can be directed entirely at it while a thumb controls the pressure.

Further, this kind of control can bring autonomy to a previously overlooked device. For example, speakers are the source of sound. It would make sense that to control their volume you would manipulate them directly. This isn’t, however, the case. Instead we have to reach to a separate box covered with controls and turn a knob. We know this drill because we have learned it, but it makes sense only if understood as a case for efficiency—all the controls are centrally located to save you the footwork of walking to your speakers and the to save money in manufacture. If the speakers were outfitted with pointable sensors, they would be controllable from anywhere they were visible as fast as you could point at them, and they would enjoy finally being addressed as the agents of soundmaking instead of the slaves of a master console. This kind of distributed autonomy and freedom from central control is exactly the condition that Pointable Computing facilitates.

Active Tagging
Imagine yourself walking down an aisle of products. You see one you would like more information about or two you would like to compare. You point your handheld device at them and they transmit information about themselves back to you. Why is this different from giving each product a passive tag and letting an active reader look up information in a database? Again the answer is about autonomy and decentralization. If the information is being actively sent by the object scanned, it does not need to be registered with any central authority. It means that no powerful agent can control the repository of product information, and anyone can create an active tag for anything without registering some unique identifier. Note also that in this scenario we see the likely condition that a non-directed wireless communication like BlueTooth would be useful in conjunction with a Pointable. The two technologies complement each other beautifully.

Getting and Putting
In a vein similar to the Tangible Media Group’s mediaBlocks project, it would make sense to use Pointable Computing to suck media content from one source and deliver it to another. Here again it is not necessary to display much on the handheld device, and one button may be sufficient. An advantage in media editing that the Pointable has over a block is that there is no need to touch the source. That means that it would be possible to sit in front of a large bank of monitors and control and edit to and from each one without moving. It may even make sense to use a Pointable interface to interact with several ongoing processes displayed on the same screen.

Ad-hoc networking
In this simple application, the Pointable is used simply to connect together or separate wireless devices. If, for instance, you have a set of wireless headphones which can be playing sound from any one of a number of sources, there is no reason you couldn’t simply point at your headphones and then point at the source to which you want to connect them.

Sun Microsystems likes to say, “The network is the computer.” This is a fairly easy formulation to agree with considering how many of our daily computational interactions are distributed among multiple machines. Any form of electronic communication necessarily involves a network. The shrinking and embedding of computation into everyday objects implies that informal networks are being created in the physical fabric of our homes and offices. If we assume that the network of wireless devices around ourselves is essentially a computer, we must admit that we spend our days physically located inside our computers. Being located inside the machine is a new condition for the human user, and it allows the possibility of directing computation from within. A pointing agent, a kind of internal traffic router, is one potential role for the embedded human being.

Reactive surfaces
Reactive surfaces are building surfaces, exterior or interior, covered with these changeable materials coupled to arrays of pointable sensors. They make use of new materials that have changeable physical properties such as LCD panels, electrochromic glass, OLEDs, or electroluminescents. It would be possible, for instance, to write a temporary message on a desk or wall or define a transparent aperture in an otherwise shaded window wall. Such an aperture might follow the path of the sun during the day.