Memory Prostheses

Christian mogensen@cs.stanford.edu http://www-pcd.stanford.edu/mogens/377/reaction-1101.html

Situation: problem

I have trouble keeping track of meetings, appointments and shopping lists. For example: the fact that this assignment is due today, the fact that my refrigerator contains no milk.

The solution I currently use is a small calendar notebook in my hip pocket, in which I note down things as they occur to me. the problem is now that I forget to check my notebook often enough.

This is a prospective memory task - plans to meet, plans on what to buy later. Declarative memory is the focus here - It's the fact that I have CS377 tomorrow morning that is the problem here, not how I get to class.

Situation: Solution

The problem is not in recording data, more in the problem of recall, or even triggering the memory of the need for recall at the appropriate moment. i.e. If I remember to check my little pocketbook every hour I'm fine. That doesn't work well when the schedule is 9-11 am: read email. So it would be better if the notebook could actively remind me to check it.

Failing the embedding of a small voice-recognition unit in a $15 calendar, we have to fall back on manual aids. Since I always need to check the book, a string round the finger is not good enough. I could use the alarm function on my watch, but it is fiddly to set the alarm for the watch while holding the notebook with the information about when to set the next alarm.

Assuming I remember to set the alarm on my wristwatch, I will be reminded to check for appointments. The cognitive load has been lessened, since I no longer have to remember the approximate time of an appointment. I only have to remember that the alarm now means "check the pocket calendar and reset alarm to next appointment".

Of course, if I don't remember to set my wristwatch for the next reminder in the notebook, the fact that my cognitive load has lessened will lead me to forget the reminder completely.

Another weakness is if a new appointment arrives for later today, before my next appointment, I have to remember to reset my alarm.

The context here is the highly lightweight and portable nature of the system. This leads to the problem that active electronics that are portable and light tend to be very expensive. An exception to this rule is the portable phone. Couple this with the fact that a home computer is now capable of doing speech recognition [IBM's dictation system in this week's ComputerWorld], and the consequence is that:

When I need to remember something, I call up the computer and tell it what I need to remember and when. The computer stores this information in a useful format (i.e. a database of appointments).
When the next appointment is about to come up, the home computer calls my mobile phone. If I have the appointment written down, I don't need to pick up the phone (and save a few cents in call charges). If I don't have it written down, I can hear the playback of the stored reminder message recorded above.

It's still pretty expensive though - the basic problem is in recording times you want to be reminded fast and easily on a machine capable of reminding you at that time. A small watch that buzzes would be fine. If the alarm time could also be set with one gesture it would be even better, since the problem of using a watch is that it takes a lot of operations to set the alarm for a particular time. It's as bad as programming a VCR!

Now - what was the next question?

Thought Exercise

I also have a big problem remembering unscheduled activities like reading library books for a class, and ploughing through my CS377 readings. I need a portable nagging device. There is no way I will ever schedule 2 pm: read War and Peace. What I can do, however, is program my PortaNag with the information that I must have finished War and Peace by Tuesday of next week.

As Tueday nears, the PortaNag grows more and more insistent, screeching at me from the ear piece: "You only have two days left to finish War and Paece". Or maybe a visual display would be better: hooked up to project onto the user's glasses. (This is a Nerd device after all)

In any case, I don't think the Nag should be completely under the user's control. It should be a bit paranoid: so being the lying deceitful cad that I am, I tell the Nag that I have finished WaP. Later the Nag queries me: Are you sure you finished War and Piece?

The interesting thing about that Nag is that it requires almost no infrastructure beyond its memory storage. Even if you are the only person with a Nag, it is still useful. The Nag poses some interesting input problems - like how do you fit an input device to something small enough to wear? It must be very simple to set up a nag, otherwise it will not be useful.

The Nag does not have to be smart - a nag is by definition dumb, and stubborn and hard to convince. The AutoNag should inherit these "qualities".

Of course there is the problem of scheduling nags - there may be several nags running simultaneously - and the Nag should make sure you have time to finish part of one task before nagging you again.