3. Design background
The product will be in two parts:
"There was a time when a flyer [pilot] sat at the centre of a complicated works. Flight set us factory problems. The indicators that oscillated on the instrument panel warned us of a thousand dangers. But in the machine of today we forget that motors are whirring: the motor finally, has come to fulfil its function, which is to whirr as a heart beats - and we give no thought to the beating of our heart. Thus, precisely because it is perfect the machine dissembles its own existence instead of forcing itself upon our notice." (de Saint Exupéry, 1967, pg 47)
To fulfil the criteria of use without distracting the user’s attention, the device must avoid presenting its-self as a separate entity that has to be ‘dealt with’. The security officers need to apply their full attention to the task at hand - dealing with an alarm or an incident. They do not have the time to stop what they are doing and ‘use’ the device - in the manner of a (paper) note book, personal organiser or camera. Ideally, the security officer would not be aware of ‘using’ a computer at all. The officer would simply get on with his job, while the device got on with its job, while at times the two jobs would coincide and information would be exchanged.
This idea that computers should be ‘seen and not heard’ was expounded more recently by Mark Weiser in his 1991 paper. He refers to ‘ubiquitous computing’, where the computer its-self becomes transparent - it ceases to command attention as a computer, but instead is seen only for what it provides. He cites the ‘technology’ of writing. In its early days, users of this technology (scribes) needed to have a good working knowledge of ink and clay tablet manufacture. Now that the technology has matured we use it with barely a passing thought for the pen or paper involved. It is the information that is important; the pen and ink have become ‘transparent’. So too with computers. Weiser looks forward to the day when computers are ubiquitous - appearing everywhere, in all shapes and sizes. They would be seen only for what they do, not what they are.
The concept of transparency is an important part of the device being designed here. The device will provide information, as far as possible, without any ‘fuss’ - no searching through menus, no typing of requests. Ideally the device would volunteer information automatically; in practice the design needs to remain simple and cost-effective, and will employ one ‘start’ or ‘help-me-out’ button. Similarly, the device will record events automatically - no pointing a camera or starting a tape. As with an alarm clock - one does not expect to have to watch it to make sure it goes off. The focus is the recording, not the recorder.
If the device is to become truly transparent, it must find its own way to be in the right place at the right time, without relying on its human partner to consciously carry it around. The simplest solution is to be attached to the human. A precedent has been set here - the watch. This piece of technology began as a separate object - the carriage clock. Queen Victoria used to take one on train journeys to check that she arrived on time. As the technology became smaller, it became more transparent. Today, the watch has achieved transparency - it is perceived, not as a machine, but as a source of information. It has solved the ‘being there’ problem by becoming an item of jewellery or clothing - the user wears the technology.
This notion of ‘wearable computers’ has been extensively researched by Steve Mann (1997) of MIT. Mann has experimented with a variety of devices and functionality. His latest device looks very like an ordinary pair of glasses, but is capable of creating virtual images for the wearer. Images can be text, or pictures such as a ‘virtual name tag’ that follows the ‘wearer’ as (s)he moves around. Currently the system requires a two-way radio link to an external computer, but advances in technology should render this unnecessary in due course. The important issue in Mann’s work is his device’s continuous presence. New possibilities arise when a system is always connected and can work without requiring the user to focus her or his attention on making it happen.
Bradley Rhodes (1997) discusses a ‘wearable remembrance agent’. This, like the proposed device, is a wearable computer that captures memories and prompts the wearer with context relevant information. Rhodes’ system is text based, with all new information being fed in by the wearer. Like Steve Mann’s devices, feedback to the wearer is visual, via a ‘hat-top display’. Thus the gathering or capture of information is a manual process. The information returned to the wearer consists of text related to that being entered. So, for example, as the wearer types that (s)he is talking to Miss X, the system will display information about Miss X (such as details of the last meeting). The big drawback with this system is that the user has to enter information manually before any useful response is obtained.
A more advanced system has recently been developed by Chris Baber of the University of Birmingham (Dobson, 1998). This system is designed to provide information to paramedics. It comprises a similar display system mounted on the paramedic’s helmet. The system allows paramedics to talk to the hospital, and to see treatment instructions without being distracted from their work.
These systems highlight the usefulness of being able to obtain information ‘naturally’ - without having to specifically ‘interact’ with a device. The missing element - which the current project hopes to address, is for the system its-self to select the information which is to be presented to the user. By monitoring its environment, the system will obtain clues about the user’s situation, from which it will be able to select a sensible piece of information to present to the user.
Attempts have been made to use automatic image capture to help with subsequent recall. The video diary project at Rank Xerox (Eldridge, Lamming & Flynn, 1991) is one such attempt. Cameras were set up in offices, corridors and meeting rooms. Each person involved in the project wore an ‘active badge’ - a small device that transmitted a signal to a network of receivers throughout the building. This allowed the control system to switch cameras as the person moved around, giving a fairly continuous record of a person’s activities. Images were taken at about 6 to 12 frames per minute. Eldridge et al found that giving participants access to the video record produced a large improvement in their recall of the days events, allowing a 200% improvement in the number of incidents recalled. It is also worth noting - and relevant to the current project - that a 60% improvement could be obtained just by providing participants with information about where they were, at what time and who else was present.
The idea of a system of automatic event recording was discussed in another paper from Rank Xerox (Lamming, Brown, Carter, Eldridge, Flynn, Louie, Robinson & Sellen, 1994). The paper gives guidelines regarding the necessary functionality for such a device:
"Suppose I ... say ... ‘Remember!, Recollect!’ Does your ... memory obey the order ... ? Certainly not. It stands staring into vacancy, and asking, ‘What kind of thing do you wish me to remember?’ " (James, 1899, pg 117)
At present, security officers are expected to retain details of each incident in their memory until they are able to sit down and write a report on the incident. Ideally, this would be immediately after the incident. In practice, however, there may be another incident to deal with before a report can be made out. Indeed, one security officer commented that sometimes it is not practical to write reports until the following day. Furthermore, security officers are given no special training in memory techniques or report writing. It is thus relevant to consider the wider body of research on memory for this type of information - eye-witness testimony - to identify some of the problems that security officers may have in this area, and ways in which these problems can be overcome.
Eye-witness testimony comprises memory for specific episodes in the past. Acquisition of this episodic memory is likely to be incidental to the event its-self - it is difficult while dealing with an incident to make a deliberate effort to memorise details that may be useful later. Further, the security officer may be preoccupied with a threat to his own safety. Stress can focus concentration on just a few items (Easterbrook, 1959). In a threat situation, these tend to be items most related to danger, such as a knife, or blood. This "weapons focus" (Loftus, 1979, pg35) can lead to a greatly reduced memory of other details which would have been useful as later eye-witness testimony.
Even during a non-threatening incident an officer does not merely store events like a video recorder. Frederick Bartlett (1932) argued that we store only the details that stand-out as important or unusual. Yet we seem unaware of this lack of information when we recall the experience. Bartlett explained recall as an active process, whereby we reconstruct memories by fitting the details we do have into an appropriate schema for the activity in question. For example, when recalling a meal in a restaurant, we assume there was a waiter or waitress, a table, cutlery, etc. even though we may not specifically remember each item. Thus, schemata help us organise the details we do remember into a coherent whole; what we ‘remember’ is part specific memory of that event and part general memories of that type of event.
Reconstructive processes have serious implications for eye-witness testimony, as schema theory suggests that some details may appear in memory because they are expected, rather than because they appeared in the original event. This was shown well by Loftus and Palmer’s 1974 experiment (Loftus, 1979, pg77) where participants were shown a film depicting a road-traffic accident and then asked to fill in a questionnaire on what they had seen. Towards the end of the questionnaire was the question "Did you see any broken glass?" In the film, there was none shown, but broken glass is likely to be a component of most people’s schema for two cars colliding. Indeed, Loftus and Palmer found that 14% of participants did report having seen broken glass. This is unlikely to be random error - people would be unlikely, for example, to ‘remember’ seeing a non-existent elephant in the film.
Schema-assisted recall can also lead to inaccuracies when information is acquired after the event, creating an artificial expectation and thus non-existent ‘memories’. Loftus and Palmer demonstrated this using a subtle - but crucial - alteration to their experiment. One of the earlier questions asked was "About how fast were the cars going when they hit each other?". They found that changing this to "About how fast were the cars going when they smashed into each other?" significantly affected the answer to the much later question of whether or not there had been any broken glass. While ‘hit’ yielded a 14% ‘yes’ result, ‘smashed into’ increased this to 32%. Thus a security officer discussing an incident with others before writing the report runs the risk of changing his memories of the incident.
To better understand the cause of such anomalies in eye-witness testimony, and to identify ways around them, it is necessary to examine theories of how long-term memories are stored. One influential theory is Tulving and Thomson’s (1973) ‘Encoding Specificity’. This concentrates on the cues that are present both at storage and at retrieval. It theorises that new memories are stored containing not just the information remembered (eg: ‘Julie Barker’), but also a variety of contextual information (eg: a pink blouse). Encoding Specificity explains the differences between free-recall (eg: "What was that woman’s name?"), cued recall (eg: "Who is the woman in this photo?") and recognition (eg: "Is this ‘Julie Barker’? ") as improvements in the quality of those cues present at both storage and retrieval. Recognition is then easy - the item to be recalled acts as its own cue, evoking the same context related information. In cued recall, only the context is provided. The task is then to find the memory containing the cue and other relevant items. Free recall is like cued recall, in that there is always some contextual information - one is always trying to recall something. Encoding Specificity does struggle to explain all recall - sometimes the relationship between a cue and an item to be recalled may be difficult to explain. Nonetheless, there is a crucial finding here for eye-witness testimony - the importance of context in retrieving memories.
One very effective practical application of this is the cognitive interview (Baddeley, 1990, pg289; Geiselman, Fisher, Cohen, Holland & Surtes, 1986). Designed in the early 1980’s, this technique encourages the eye-witness to relive the event. (S)he is given detailed instructions along the following lines:
Using this technique, Geiselman et al found 35% more correct information was elicited compared to a standard police interview, with no significant increase of incorrect information.
Thus the aim of the present system is to assist the security officer in his recall of events by helping to reinstate the context in which an incident took place. The information to be recalled does not have to exist in the captured data. It is enough to trigger recall, without supplying the actual information. Presenting details from the incident (in the form of still pictures and short sound clips) provides a rich source of cues from which memories may be retrieved, either by recognition or by cued-recall. The extra detail recalled will allow ‘fine tuning’ of the appropriate schema, making the overall memory for an incident much more detailed and accurate.
| Next Section | Previous Section | Beginning | Contents | MSc home |