Computers are very good at storing data and performing repetitive functions very quickly. Humans, on the other hand, can be very good at intuitive leaps and recognizing patterns and structure, even when not actively attending to an event. Thus, an interface where a wearable computer helps the user remember and access information seems profitable. In general, 95\% of computer time is dedicated to word processing. Since wearable computing allows such ease of access to a keyboard, this percentage may be even higher for such machines. However, word processing requires about 1\% of the cpu power of the system. Instead of wasting the remaining 99\%, an information agent can search the user's personal text database for information relevant to the current context. The names and short excerpts of the closest matching files could then be displayed. If the search engine is fast enough, a continuously changing list of matches could be maintained, which would increase the probability that a useful piece of information will be recovered. Thus, the agent can act as a memory aid. Even if the user mostly ignores the agent, he will still tend to glance at it whenever there is a short break in his work. In order to explore such a work environment, the Remembrance Agent was created.
The benefits of the Remembrance Agent (RA) are many. First, the RA provides timely information. If writing a paper, the RA might suggest relevant references. If reading email and scheduling an appointment, the RA may happen to suggest relevant constraints. If holding a conversation with a colleague at a conference, the RA might bring up relevant work based on the notes taken. Since the RA ``thinks'' differently that its user, it often suggests combinations that the user would never put together. Thus, the RA can act as a constant ``brain-storming'' system.
The Remembrance Agent can help with personal organization. As new information arrives, the RA, by its nature, suggests files with similar information. Thus, the user gets suggestions on where to store the new information, avoiding the common phenomenon of multiple files with similar notes (e.g. archives-linux and linux-archives). The first trial of the prototype RA revealed many such inconsistencies on the sample database and suggested a new research project by its groupings.
As a user collects a large database of private knowledge, his RA becomes an expert on that knowledge base through constant re-training. A goal of the RA is to allow co-workers to access the ``public'' portions of this database conveniently without interrupting the user. Thus, if a colleague wants to know about augmented reality, he simply sends a message to the user's Remembrance Agent (e.g. thad-ra@media.mit.edu). The RA can then return its best guess at an appropriate file. Thus, the user is never bothered by the query, never has to format his knowledge, and the colleague feels free to use the resource (as opposed to knocking on an office door). Knowledge transfer may occur in a similar fashion. When an engineer trains his replacement, he can also transfer his RA database of knowledge on the subject so that his replacement may continually receive the benefit of his experience even after he has left.
Possibly the most striking use of the Remembrance Agent is its ability to seemlessly share knowledge in a work group. Instead of simply using one member's notes, the database is expanded to include the members of a small work group. This allows personal experience to be shared quickly and conveniently. For example, such an interface is useful if one member of the workgroup is in charge of repairing the team's computers. When the team member receives new information about an obscure bug in the operating system, he puts it in his personal information files which can then is available to the rest of the team. If other members then experience this bug, the appropriate file might be suggested to the member. This helps identify the problem quickly and greatly reduces the overhead involved in diagnosing and assembling a well-formed question to the rest of the team, whose appropriate member may be unavailable.
To experiment with this idea, three volunteers have been assembled. These volunteers have pooled their knowledge into one RA database which runs continuously in their text processor (emacs19). Originally, logs were to be kept on which of the RA recommended files was accessed, but due to problems with the software, a successfully running team system was not available in time for the week long study planned.
The current Remembrance Agent uses the Savant information retrieval system developed in-house by the Jan Nelson and Bradley Rhodes. The Remembrance Agent runs through emacs, a popular text editor. The user interface is programmed in elisp, and the results are presented as a three line buffer at the bottom of the window. Several considerations have gone into the design of the RA. First, the RA should not be distracting unless unusual circumstances arise. To that end, the RA does not use boldface or highlighting and is run at a low priority. Secondly, if the RA recovers something of interest to the user, the full text is accessible with a quick key combination (e.g. C-c 2 to get the file associated with the second line). The RA is configureable to search in any size of context. In this case, the highest ranked result from a search on the last 10 words (first line) and the highest 2 on the last 100 (second and third line) are displayed. The local context search is done ever 10 seconds while the larger context is searched out of phase every 30 seconds
Figure 1 shows the output of the Remembrance Agent. The reference database for this screen shot was Brad Rhodes's e-mail archives. The first number on each line of the RA output is simply a file label for convenience. For example, to view message 2, the user would simply press ``Control-2'' The second number on each line refers to the ``relevance measure'' of the message.
A significant artifact of this test is that each participant does not have an equal sized database to share with the others. Thus, it can be expected that the one with the largest database would be sited most by the Remembrance Agent, and such is the case so far. However, the usefulness of the system in everyday e-mail was immediately shown. Upon receiving an e-mail request for help on file io under UNIX, the RA immediately suggested a recent serendipitous message from a different newsgroup. Another point of interest was that it was often not clear whose memories were being recalled at any given point. This may be due to working relationships of the test subjects but brings up a point of view issue. What happens when your RA brings up other's impressions of the same event? Does this allow for easier collaboration? Will other's viewpoints be accidentally accepted as one's own?
Using this system on a multiple person database has revealed several areas of improvement for the system. Frist, the server nature of savant must be improved to allow faster update speeds. Better descriptor lines should be included to indicate out of whose memories a given recall comes. Finally, several improvements in the emacs user interface have been identified. Once these are fixed, a larger test may be attempted with detailed logging on what information is referenced and when this information is provided across different users.