The badges register position and identity with each room by infra-red. This raw data is converted to 'significant events' and stored in a user's diary. At least 90% of events are correctly identified.
This is accurate enough to encourage regular use.
[Slide: Pepys Diary - searching]
The info is purely location based, so a meeting is registered as 3 people in the same room for more than 10 minutes. This does not give information about what was going on in the meeting.
Adding video snapshots from ubiquitous cameras improves recall. So you see what the meeting room loooks like a several times during the meeting
A note taking pad captures the time notes are made, so that later on, the notes may be matched with voice or video recordings. These notes can be attached to the diary info.
Paperwork is a problem, since it is outside the computer's domain. Experiment involves using a camera over desk to capture images of paper, and comparing with a database of pre-scanned documents. Remember we are not trying to read the doc, only guess which on it is.
Major problem is in processing the raw data into higher level episodes that match the user's recollections. There is a lot of raw data.
Since tabs know what's going on, they can monitor the user to see if the planned action is carried out. i.e. If the calendar says a meeting and you are alone, then you need to be reminded.
Central feature is ability to sense aspects of users environments and make records which can then be used later to enable recall of things they did not know they were going to need to remember.
However: analysing and presenting the data is a challenge. Too little analysis, too much data, overwhelms the user. Too much analysis, too little data is not going to trigger recall.
The data triggers recall of what the user needs to know, even if that information is not in the database.
A memory prosthesis must sense its external environment. The more information a system can gather, the more help it can be when information needs to be recalled later. Many intentions are related to particular people or places.
i.e. a reminder next time you bump into someone.
The furture is unknown. in the future you will forget something. The problem with current systems is the need to recognize ahead of time what is going to be important to remember. Unobtrusively the tab should capture as much data as possible without intervention. Structure will be imposed later during retrieval.
Manual note taking or sketching. It must be simple and fast, and easily attached to a meeting or other event.
The point is that you can enter future reminders and associatie arbitrary notes and events.
Focus on relevant information
There will be a lot of data in the database. Different memory problems
require different strategies for solving them. The device should support
a range of strategies.
In large databases, getting lost is a serious problem. having a context-based searching method should make it easier to navigate the history of events. For example: we may remember that we last had the lost document when we were with a particular person in a certain room and so wish to start from person and location information.
This is similar to the searching techniques used for annotated video at the Media Lab.
A memory prosthesis should associate a rich variety of cues with any note so that we can still retrievei it even if we never categorized it or named it.
Ease of use is fuzzy
using the device must be easier than using other means of recalling an event.
It should capture many different kinds of data and present them in a uniform way.
The memory prosthesis should be at least as available as a day-timer or dairy. Ubiquitous availabilty is achieved using Tabs: small, portable pen-based clients.
Applications should move gracefully between devices in a mixed environment. E.g. when you sit down ata a terminal, a more detailecd view of what you are seeing on the tiny Tab screen should be available.
A successful prosthesis will intergrate seamlessly into the users everyday activites.
Fail safe:
If a user is depending on this system to remember stuff, it must
be very reliable and fail in safe ways. It must be obvious when it is
not working.
System failures are also significant events which must be captured.
Privacy:
This device is capturing all sorts of data about you and all the people
you meet. Think about the ethics of storing someone's personal information
just because it is publically available and you happened to bump into them?
Conclude:
A persons memory for details of past events can be triggered by
providing partial information about it.
People can carry out intended actions by the Tab providing timely context-sensitive reminders.