Do Life-Logging Technologies Support Memory for the
Past? An Experimental Study Using SenseCam
Abigail Sellen, Andrew Fogg, Mike Aitken*, Steve Hodges, Carsten Rother and Ken Wood
Microsoft Research Cambridge
7 JJ Thomson Ave, Cambridge, UK, CB3 0FB
{asellen, v-afogg, shodges, carrot, krw}@microsoft.com
*Behavioural & Clinical Neuroscience Institute
Dept. of Psychology, University of Cambridge
mrfa100@hermes.cam.ac.uk

ABSTRACT
We report on the results of a study using SenseCam, a ―life-
logging‖ technology in the form of a wearable camera,
which aims to capture data about everyday life in order to
support people‘s memory for past, personal events. We find
evidence that SenseCam images do facilitate people‘s ability
to connect to their past, but that images do this in different
ways. We make a distinction between ―remembering‖ the
past, and ―knowing‖ about it, and provide evidence that
SenseCam images work differently over time in these
capacities. We also compare the efficacy of user-captured
images with automatically captured images and discuss the
implications of these findings and others for how we
conceive of and make claims about life-logging
technologies.
Author Keywords
Personal digital archives, life-logging, SenseCam, episodic
or autobiographical memory, images, capture
ACM Classification Keywords
H5.m. Information interfaces and presentation (e.g., HCI):
Miscellaneous.
INTRODUCTION
The idea of a ―life-log‖ or a personal digital archive is a
notion that can be traced back at least 60 years [5]. The
vision is that technology will allow us to capture everything
that ever happened to us, to record every event we ever
experienced and to save every bit of information we have
ever touched. Indeed, in recent years, this vision has been
given a new lease of life, recent impetus and enthusiasm
coming mainly from a number of technological
advancements. These include the development of smaller,
lighter-weight capture devices and sensors (capturing
everything from image, location and ambient sound to heart
rate), advances in wireless networking, and massive
increases in digital storage capacity making the archiving of
huge amounts of personal data possible. Now, as never
before, technology offers the possibility of capturing data
from everyday life both continuously and unobtrusively.
As a result, a number of new efforts to build systems and
devices to support life-logging have emerged in recent years.
A case in point is a technology developed in our own
laboratory called ―SenseCam‖ [18]. SenseCam is a device
containing a camera and embedded sensors worn around a
user‘s neck which automatically takes a series of still images
over time as well as capturing other aspects of life events
such as ambient light levels, temperature and movement.
In addition to SenseCam, there are many other life-logging
systems being developed both in research laboratories and as
commercial products. Much of this work is only concerned
with recording users‘ activities in the digital world, focusing
on interaction with electronic media such as documents,
photos, sounds, videos and so on [e.g., 1, 9, 13, 15].
However, another strand of work is more concerned with
recording aspects of life ―out there‖ in the real world, away
from the desk. Some of the earliest research in this vein can
be traced back to the Active Badge and PARC Tab systems
at EuroPARC, Xerox PARC, Olivetti and AT&T [e.g., 20,
27, 32] in which users carried or wore small devices which
were tracked within the confines of an area covered by
networked sensors. More recent instantiations of this
approach can be classified either as wearables, portables or
instrumented environments. Wearable systems are based
mainly on head-mounted still or video cameras [e.g., 17,
24,], or on wearable audio capture devices [30]. Portable
systems largely make use of specialized software on existing
devices such as PDAs, notebook computers or cell phones
[e.g., 21, 22]. Cell phone data have also been used to infer
human activity after the fact [10, 19] by analyzing location-
based data. An alternative approach is to rely on
instrumented environments which capture activity through
installed sensors or local wireless networks [e.g., 25].
Many potential benefits have been put forward for such
systems, but by far the most common proposition is that, by
capturing data about our daily activities, life-logging systems
will offer effective support for memory of our own personal
past [e.g., 2, 17, 20]. These arguments range from
supporting the reliving of or reminiscing about personal
events, to more specific functional support for memory

Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise,
or republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee.
CHI 2007, April 28–May 3, 2007, San Jose, California, USA.
Copyright 2007 ACM 978-1-59593-593-9/07/0004...$5.00.

CHI 2007 Proceedings • Capturing Life Experiences April 28-May 3, 2007 • San Jose, CA, USA
81

including finding lost objects or documents, remembering
names, remembering whom you met, details of
conversations, and remembering past actions or events. In
other words, life-logging applications claim to support a
whole range of ways in which we can look back, re-live, re-
examine, and search through our past experiences.
Such systems generally make some (usually implicit)
assumptions by their very design. One is that the more data
captured the better in helping us to look back. So these
devices aim to capture and store as much data as possible in
the course of daily life. Another is that the more different
kinds of data captured the better. Thus many systems strive
to be both as comprehensive and diverse as possible in the
amount and variety of data capture that takes place. And
another is that capture of data should occur automatically or
at least with minimal effort on the part of the user. Thus data
are captured throughout the course of everyday events, and
as an incidental by-product of a user‘s activities.
Naturally, one result of such undertakings is a huge amount
of data that must be structured, organized and searched
through. Indeed much of the interesting work in this area is
grappling with the challenges this presents for users,
interface designers and computer scientists [12].
However, in this paper we propose that there are equally
important research challenges in substantiating the beneficial
claims made about these systems. Specifically, there is little
systematic evidence that the data offered up by life-logging
systems do in fact support human memory, and little
research to help us understand how they might do so. To
date, we must rely on anecdotal evidence or conjecture.
Indeed the rhetoric in this area often rests on unfounded or
even irrational assumptions about what these systems do. An
example of this is the claim that life-logging technologies
capture day to day ―experiences‖. It is important to point out
that these systems, no matter how sophisticated,
comprehensive, or diverse in their data capture, do not
capture human ―experience‖. Rather such technologies
capture a set of cues (data) which we hope will trigger the
remembering of human experience.
This then begs a number of important questions such as: Do
the cues these systems capture really support memory of past
experience? If so, how much can people recall about the past
given a particular kind of cue? What kinds of cues are best
(e.g., images, ambient sound, location)? Does the
effectiveness of different kinds of cues deteriorate over time?
Does it matter if these cues are automatically captured by the
technology or actively captured by the user?
The study which we describe in this paper is a step toward
explicating and answering questions such as these. Yet, this
is not solely in the service of advancing our theoretical
knowledge--it is also of relevance more generally for the
development of life-logging technologies. Unless we
systematically assess the underlying assumptions of such
technologies, we can only make ―best guesses‖ about what
they are good for, and how they may be most effectively
designed.
RELATED RESEARCH
As we have already alluded to, there is very little in the CHI
literature which explores the link between life-logging
technologies and human memory. While there are many
systems which claim to support memory, most often such
claims are not substantiated with anything more than
anecdotal evidence.
Exceptions include a study, now 15 years old [11], in which
three subjects‘ memory for events during a day at the office
was tested after intervals from one day up to four weeks.
Subjects were given both automatically generated text-based
diaries and video excerpts of their days and found that while
the text diaries did increase recall, video cues were more
effective. A more recent study (again using a very small
sample size) investigated people‘s memory for computing
events within a one hour time slot [8]. In this case, there was
only weak evidence that both video and still images can
serve as effective reminders of these past events after one
day and one month. Research by Vemuri et al [30] also
provides some evidence for the use of audiotaped
conversations (in this case, conference talks) in jogging
memory, although in this case the memory problems were
simulated rather than real. Finally, Carter and Mankoff [6]
explored the efficacy of different kinds of media as cues in
eliciting recall of everyday events within the context of diary
studies. In a study involving 7 participants, they compared
the use of photos, ambient sounds and tangible objects in
cueing the recall of a day-long festival. Not many details are
given with respect to the method and analysis, but in general
they found that photos were best for eliciting detail.
Turning to the literature in the psychology of human
memory, we face a different problem: it is often difficult to
make direct connections from contrived laboratory situations
to the kinds of real-world situations we are interested in.
Having said that, there is much theoretical work that can be
applied.
For example, we can begin with the distinction between
―semantic‖ and ―episodic‖ memory [29]. Semantic memory
is the term used to refer to general knowledge of the world
(e.g., that a dog has four legs, that the capital of France is
Paris). Episodic memory, on the other hand, refers to the
capacity to re-experience specific episodes from your past
(e.g., feeding your dog last night, or the time you visited the
Eiffel Tower). With life logging technology, it is clear that
most of the claims that are made have to do with enhancing
and exploiting the relationship between captured cues and
episodic (or what is sometimes called ―autobiographical‖)
memory.
With regard to episodic or autobiographical memory, a
further useful distinction is made by Conway [7] who
distinguishes the recollection of specific details of recent
past events (remembering what one perceived or felt at the
time of an event) from more general knowledge about
oneself accumulated over time. The former is about
mentally traveling back in time; the latter is more about the
―things one knows about oneself‖ such as the periods in
one‘s life, repeated patterns and habits of activities, and the
CHI 2007 Proceedings • Capturing Life Experiences April 28-May 3, 2007 • San Jose, CA, USA
82