1st Panel Report

Why Memories for Life?

The opening panel was concerned with the fundamental question Why Memories for Life? and included panellists Steve Whittaker (University of Sheffield), Abigail Sellen (Microsoft Research, Cambridge), Alan Dix (University of Lancaster) and Christopher Brewster (University of Sheffield). The panel discussed issues relating to individual and social memory and also presented a projection of future research questions relating to the Memories for Life (M4L) research agenda. The main questions that the panel addressed were:

• What exactly is it we wish to support and what do we expect the user benefits to be?
• What research needs to be performed under the guise of the Memories for Life Agenda?
• How could technology support an individual’s memory? What could digital memory do for us?
• What technology to use and how could it support user goals?

Abigail Sellen (download PowerPoint slides) comes to Memories for Life as a psychologist and is interested in supporting individual memory. She noted various diverse claims to the benefits of memory systems, and outlined the following four different forms of memory:

• Relive (episodic): to mentally re-experience memories for practical or therapeutic reasons
• Reminisce (episodic): to re-experience for pleasure
• Retrieve (semantic): accessing personally relevant information
• Reflect (semantic): accessing relevant information

Episodic memory enables one to mentally re-live a past event, whereas Semantic memory is concerned with facts and concepts relating to general knowledge.

These two types of human memory (episodic and semantic) were discussed, and it was concluded that current technology mainly supports the latter. The retrieval and access of memory were seen as important tasks supporting these and the act of reminiscing past experiences were presented as key motivators to the future of M4L. Abigail went on to describe Microsoft’s SenseCam, a wearable digital camera that takes photos at predetermined intervals. It was suggested that by looking at a photo it aids one’s memory to a recall a past event. Abigail presented research that aims to identify whether or not the SenseCam images actually help the act of reminiscing.

It was suggested that we currently do not know enough about recollection, and in order to drive research we need to identify which human-centered tasks we are attempting to support. That is, there is no need to try and create a model of human memory – this is too difficult a task – but we can aim to support human memory in a task-oriented fashion. In order to accomplish this we must identify the human-centered problems/tasks we are trying to support.

Abigail doubts the value of creating a complete record of everything she's ever experienced, like a “file system on steroids”. She also went on to suggest that not all the information should be generated by the subject. She speculated about a future in which a form of collaborative/social memory made up of different people’s experiences will exist.

Abigail presented the following potential benefits to come out of M4L related research:

• Individual recall – to help support questions such as "where are my keys?"
• Sharing information with families and friends – the construction of social networks, and the notion of collective memory.
• Prospective memory – reminding about future events; and the possibility of predicting future actions, such as "I must remember to keep my dentist appointment tomorrow afternoon".

Finally she suggested that we need to go beyond the current model of "Saving Everything". We need to support situations where memory functions poorly, and work in synergy with human memory rather than attempting to replace it.

Steve Whittaker (download PowerPoint slides)] from Sheffield University’s Information Retrieval Group began by presenting 4 definitions of digital memory:

1. A record of everything I have ever experienced: "Digital Me".
2. A collection of my personal digital artefacts - pictures, documents, music, etc: "Digital Garage".
3. A collection of items regarding me (including information generated by others).
4. A collective record of the past: "Digital We".

Steve presented M4L as a means to support research into the task of individual recall/reflection in the form of developing aids for absent-mindedness by utilising prior experiences.

Steve then went on to suggest that the ability to share one’s personal experiences with others will reap the most benefits.

Social sharing is important because:

1. Stand-alone digital photos are too passive
2. Sharing with others motivates and bonds groups
3. Allows us to add a Narrative
4. Allows us to add an Event Structure

Steve concluded by presenting the following research agenda:

1. We that we need to go beyond saving everything.
2. We need to identify when people use Digital Memory rather than their own memory.
3. Need to find ways of supporting the notion of forgetting.
4. Must exploit other disciplines such as the arts, e.g. linguists must be used to help build narratives.
5. And need to find new models to push the field of Information Retrieval forward.

Alan Dix introduced himself as a user interaction specialist with a mathematical background. He suggested the need to define the difference between Memories OF Life rather than FOR life. And then went on to stress how important it was to look into memories for "supporting" life, as apposed to memories for "documenting" life.

He commented on the possibility of being able to share memories. He mentioned that events are linked in human memory (i.e. everyone can remember where they were when Princess Diana was killed, Kennedy was assassinated, the World Trade Center destroyed, etc) and that this could be an affective way to build links between peoples' memories using common real-world events.

He perceives online communities as digital memories, and how the news is now driven from the individual level, i.e. at the YouTube level. And how memories are at the human level, but can be amalgamated into a shared digital memory of society. This allows for "history" to be documented by the masses as opposed to by libraries and institutions. He described this concept as the "memory matrix".

Christopher Brewster described an agent, a personal agent or a cuddly toy that knows its owner, and presents it as a life long companion. Christopher is presenting the vision of the Companions project headed by the University of Sheffield.

Christopher went on to discuss the concept of storing too much data, and then poses a research problem of regarding the concept of forgetting: how do we identify what information needs to be forgotten? He also commented on the ethical issues of storing the data, where and how should it be stored, and who should have access. The retention of data beyond a human's lifetime is an issue that needs to be considered, should access to person’s personal agent (companion) be given to the police, the state, or to the person’s family?

Questions:

Wendy Hall posed a question about how to begin acquiring a data set to support M4L research. She asked how one would start to collect such data. Nigel Shadbolt then went on to state that he does not believe that there will be "one" data-set, and how M4L is interested in emerging data, and that new personal data is being generated and identified all the time.

A question was then presented to the panelists, asking them how often they think Gordon Bell of Microsoft actually ever queries or checks his personal data collection. Alan Dix answered by stating that it is important to look into identifying the cost of storing data, i.e. we need to evaluate how you would access and the query data, before deciding whether it is worthwhile recording it.

Hamish Cunningham stated that it is important to make computer systems associative and able to answer questions such as "what was I listening to, when I was emailing my colleague X?" Alan then went on to point out that outsourcing memory is an ever-evolving process, and that humans are used to it and have done it as long as record history. 25 years ago, when writing papers you remembered all the references, nowadays you have a bibtex file. Photos are the same; do we store photos or just search the web for pictures of a certain person? Hence, we should identify whether we can actually effectively use information before we decide to store it.