Workshop Goals

A vision - towards ubiquitous peripheral interaction

Internet enabled cell-phones and PDAs with wireless networking capabilities enable continuous access to information for mobile users. A PDA can access information from a server, thus (potentially) keeping users current. However, users cannot be expected to hold their cell-phones or PDAs in their hands, constantly checking them for new information or messages. In other words, these devices support a "pull model" of information access, in which users have to make an explicit decision to seek information. In contrast, wearable devices like wrist computers enable peripheral awareness of information, a "push model". With a wrist computer, for example, information such as messages, reminders, and news stories could be constantly streamed to the device and displayed as a scrolling "ticker". The information is at the periphery of users' attention, and there is no guarantee that any particular item will be noticed. However, from time to time a user may glance at his watch and notice an item of interest. When this occurs, he can follow up on that item simply by pressing a button on the wrist display. This will initiate a program to view that item in detail on the user's PDA. The user then can take out his PDA and interact with the information as desired. This workshop will explore the theme of multi-device interfaces for ubiquitous peripheral interaction in depth. It has a number of specific goals, including identifying requirements for ubiquitous peripheral awareness devices and considering specific devices that can meet these requirements, exploring software techniques and architectures that drive the interaction, and examining designs for interfaces that divide their functionality across several wearable devices. We consider each of these goals in some detail.

Goal 1: devices for ubiquitous peripheral awareness: requirements and candidates

A wearable peripheral awareness device must be always on, always accessible (that is, a user must always be able to receive information from it), and must not demand explicit user attention (that is, it is accessible without a user explicitly "taking it out to use", as one must with a PDA or cell-phone). It would be very helpful if some version of the device already enjoys widespread use (as do wristwatches, for example). And, since in our vision, this device is intended to work together with a PDA, it must be networked. We believe that wireless devices are much more acceptable to users, which means that the networking should be wireless. There are various promising devices that meet some or all of these requirements. One is a computationally augmented wristwatch. There has been a lot of activity in this area recently, both in commercial and research contexts. Companies such as Suunto make sophisticated wrist computers tailored for sports like golf and skiing. Fossil and onHand make wrist PDAs that runs Palm OS. Swedish researchers prototyped a simple 'Reminder Bracelet' (CHI 2001) - LEDs were added to a wrist band, and they were lighted to show several types of information of varying levels of importance. Most notably, IBM and Citizen are prototyping a general purpose wrist computer that runs Linux. Another interesting possibility is the use of audio for peripheral awareness. For example, Sahwney and Schmandt's Nomadic Radio used small, neck-mounted, directional speakers to deliver audio. Headsets that are relatively unobtrusive and still allow users to engage in normal interaction are another possibility;. Grinter & Woodruff (CHI 2002) did a pilot user study that showed some preference for single ear headsets; ear buds also are worthy of more exploration. Other wearable devices also are worth investigating, e.g., computational jewelry or eyeglass displays. When it comes to the networking requirements for a multi-device interface, Bluetooth seems well-suited. It works over short distances, but the peripheral awareness device and the PDA will be well within the working range. While Bluetooth is appropriate for networking the components of the multi-device interface, the PDA will need to connect a server (more on this below). This will require a long-distance wireless networking technology such as WIFI or GSM/GPRS. Another important issue we will explore is how much can be done with off-the-shelf devices. The IBM/Citizen 'Watchpad' is not yet available; if it were, it would fulfill most or all the requirements mentioned above. However, it currently still may be necessary to do some simple hardware prototyping along the lines of the Reminder Bracelet. In general, the workshop will address this goal by trying to reach agreement on appropriate requirements, developing a comprehensive list of candidate devices, and evaluating the extent to which each candidate satisfies the requirements.

Goal 2: enabling software techniques and architectures

For any information delivery system, ensuring that information is relevant to an individual user is a key goal. This is even more important in our context. Networking may be relatively slow, unreliable, and expensive. Most of the time, users' attention will not be focused on their peripheral awareness device, so the system should be careful not to distract users from their current tasks. These considerations pose challenges for the information management software that runs on the server. This software is responsible for monitoring events in a user's computational environment (e.g., email, voicemail, calendar), deciding what information to send to a user's wearable system, prioritizing that information, dealing with user responses, and learning from those responses so its decisions can improve over time. One challenge for the design of this software is to identify the factors it can use to make these types of decisions. One such factor is timeliness. New email or voicemail, upcoming appointments, and changes to web-based information services such as stock tickers or sports scoreboards are some examples of timely information. For mobile use, location also is important in determining relevance. GPS is one technology that enables devices to be location-aware. Location-aware devices could notify people of nearby stores with items they need to purchase, or of nearby friends whom they'd like to contact. A system also should provide users with information that is important. For a simple example, it should avoid streaming spam email to its users. Ideally, it should be able to prioritize all the types of information (messages, reminders, news items, etc.) and use these priorities to guide its presentation of the information. Initial priorities may be based on the type of the information: for example, getting notified that one of your friends is in the same mall or library as you is likely to be more important than most news stories. Finally, an effective system will be adaptable. No two users will have the same notion of importance. A system should learn from user responses what that particular user considers important. For example, if a user frequently reads email messages from particular addresses, messages from these addresses should gradually receive higher priority. Thus, the key enabling software techniques come from areas such as information filtering and artificial intelligence, particularly machine learning. The workshop will explore the range of techniques that are relevant, discuss lessons to be learned from existing research prototypes, and identify open problems where new research is needed.

Goal 3: design of multi-device interfaces

There are several challenges here. One is the general problem of designing for devices with i/o capabilities that are quite different from, and relatively impoverished compared to desktop user interfaces. Devices may have very small displays - or, in the case of audio devices, no displays at all. And their input capabilities may be very limited, in the case of a wrist computer, or inherently error-prone, as with a speech interface. Designing an interface for a PDA or cell-phone is hard; designing for a wristwatch will be even harder. Another issue is when it is appropriate to 'promote' information beyond the periphery, that is, when the system should attempt to notify the user about some information at once. When this is necessary, a wristwatch computer might be able to flash its display, vibrate, or even sound a tone. Of course, the decision whether to seek the user's attention must be guided by the prioritization techniques mentioned above. It's also important to consider how users might respond to items on the peripheral awareness device that catch their attention. One way would be to support brief "canned" responses. For a message, in particular, it should be possible to send a response like "OK" or "I got your message - will reply in detail when I'm back at my desk" without having to take out one's PDA. One research question worth exploring is how to define a small set of generally useful canned responses. Another issue is how to design an interface for a device with very limited input capabilities (like a wristwatch) that makes it simple for users to select a response. Of course, the most interesting and novel interface design challenge is how to divide functionality across multiple devices. The simplest case to envision is one where a user notices something on the peripheral awareness device, gives a single command (e.g., a button press) to indicate interest, then takes out a PDA to explore the information in detail. This model assumes that one first uses the peripheral awareness device alone, then the PDA alone. It also requires very simple coupling between the two devices. We think this model is well worth exploring; however, we also will attempt to identify situations when a more complicated model for combining the devices is necessary. For example, when a user has his PDA out and is using it, should information and notifications still be going through the peripheral awareness device? Or should they now go through the PDA? This suggests that the division of functionality across the two devices may have to be dynamic.