Direct Observation
Direct Observation
Direct Observation
Direct observation
What does this technique do?
The term nonintrusive is often used to characterise this technique: Users do what they normally do without being disturbed by the observers. One of the advantages of this method is that users can be observed in the environment where the system is normally used. This is why direct observation is said to have high face validity, also referred to as external validity or ecological validity. However, one must always be aware of the possibility of the so called Hawthorne Effect; the fact that people usually perform better under observation because of the attention paid to them. During direct observation it is common for an observer to be present who sits passively and records as accurately as possible what is going on. Usually it is the behaviour of one or more persons that is recorded, and an advantage of the technique is that a number of people interacting with each other and the same piece of equipment can be observed. A variation on this technique is to have a video camera mounted at the point of usage, which records interactions which can later be watched and analysed by an observer. The observation can be totally free or more structured i.e. where observers record events as belonging to one of a number of discrete categories identied. The number of categories adopted largely depends on what the observers intend the data will be used for, and very broad categories may be used for some studies, whilst detailed categories will be used for others. In some investigations a more free approach may be used where the observer records all of their impressions during observation rather than trying to group them in some way. However this introduces a high degree of subjectivity into the evaluation process, and in practice it is usually better to try and dene the categories of behaviour that will be observed. One way of achieving this is to perform a pilot study where free recording
SPECIAL CONSIDERATIONS: Mental impairment Communication impairment Blind/visually impaired The elderly user
The Hawthorne Effect: People tend to perform better under observation because of the attention they get.
usert Tools
33
34
When to use it
To use direct observation for product evaluation requires that there is a fully operating product in use. One may also apply direct observation early in requirements capture, but in these cases the emphasis is on observing the difficulties facing the person in everyday living or the activities that the future product is intended to support. Such observation may also reveal problems with existing products already being used, which in turn may suggest ideas for future developments. For ethical reasons it is not recommended that users be observed without being informed that such an investigation is taking place.
35
usert Tools
36
Selection of users
Although the ideal is to let a representative sample of the user population try out equipment, this is often difficult and expensive to achieve. After having dened the user population the testers should decide whether they would try to approach the extremes of the distribution (best case worst case) or the mode (the typical case). See the Procedure section below on advice for number of users to be involved.
Special considerations
General
To perform a direct observational study usually requires getting access to the user at home or at work. In both cases one should take careful considerations that the necessary permission is obtained, and not only from the user. This may be helpers or other persons present in the same location, or it may involve parents or guardians having a responsibility for supporting the users decision in these kind of situations. More than ever the principle of informed consent applies when using direct observational methods, as it is unethical to make observations without such consent. It may be tempting to use a candid camera technique where interactions are recorded without the awareness of participants. It is not advised to do this under any circumstances as there are serious ethical problems with such hidden monitoring. One should also be aware that many disabled people are embarrassed when asked to expose their disability to strangers. Special attention should therefore be paid to the motivation of users to participate as it may not be obvious why users would want to participate voluntarily in these kinds of studies. As with other forms of investigation nancial incentives can be considered, but the indications are that this is unlikely to be a major incentive for most people prepared to take part in investigations. Careful explanation as to the value of the work can often be more effective, along with appropriate feedback as to the results of the study and its practical consequences. Many elderly and disabled people are lonely, and might see this as an opportunity to engage in social interaction. This can be an incentive for the user to participate and can also provide the user with some benet from the investigation. The practical implication of this is that the investigator needs to allow time for such interactions to take place, and build this into the evaluation process. It is a good idea to establish ground rules for when observations should be allowed to take place uninterrupted, and when social interactions are to encouraged. Without a careful discussion of these issues, the investigation can suffer from unwanted interruptions which limit the value of the data obtained.
usert Tools
37
Mentally impaired
The main concern when mentally impaired people participate in direct observational studies, is to explain the presence of a person (usually a stranger) that sits watching and without taking part in what is happening in the same room. For this reason careful preparation is needed so that the user gets used to the observers presence. It is a good idea that the observer tries to explain their own presence to the user, and in addition any other parties that may be present also need to be informed in advance where possible. Typically the observer should be prepared to allow at least one day to let the participants get used to the situation. On the second day they should try to get gradually less involved with the participants, so that the end user becomes comfortable with the presence of a passive observer.
Communication impaired
If the user applies sign language or a communication device, the observer should be able to interpret the users communication. This may in some instances be rather straight forward, but in other situations knowledge of the communication system of the user is absolutely necessary. To settle the requirements for the observer, interviews with the user or someone close to the user must be done in advance of the observation. In some cases limitations in communication ability will make it necessary that someone who is familiar with interacting with the user e.g. a member of their family is used as an observer.
38
Procedure
Planning
Nature of observations The nature of the problem will decide what kind of observation are required. If the project is in an exploratory phase, the observations might be unstructured with an aim to discover what is important. With a more precise set of problems, one should prepare an observational form, allowing the observer to record the occurrence of different behavioural categories. One would normally construct this after a period of unstructured observation.
Example:
A product developer may suspect that there is a need for better devices for keeping track of the time of day among people with mental impairment. She decide to visit a group home where four mentally impaired adults live with the support from a team of helpers, one always present. After having made the necessary appointments she is invited to visit the home. The rst visit must be used to be introduced to the users and the helper on duty that day. She may explain that her company makes watches and that she is interested in knowing what kind of watches are needed. She will probably have many reactions to that, which is strictly speaking not raw data from direct observations, but rather should be treated as opinions. However, such opinions may be valuable in their own right. When the residents are used to the observers presence (which may take a day), they will be in the position of being able to observe without causing disruption to the activities of interest. more
usert Tools
39
With such broad categories, it is not sufficient just to indicate the category, the persons involved and the actual behaviour must be recorded, for example "Peter asks what time it is". Some of the state categories may be: Person x is in a time-critical phase (for example about to take the bus, get dinner ready) Person x is waiting for something (description of the situation.....) No time critical activities in the group home
usert Tools
40
Observations
During the observation it can be useful for the observer write down the type of activity being engaged in and the time at which it took place. Errors or misunderstandings that took place in using a product can also be of interest to record, and in some cases exactly how long a particular activity took to perform. If the observed events occurs very frequently, it may be wise to sample observations during the day, to ensure that a representative sample of use is observed. The observation periods should be determined in advance, and should not be decided on an ad hoc basis. The length of time of each observation period should be carefully considered, taking into account the frequency of the events of interest, and the difficulty in observers maintaining concentration for long periods. Evidence suggests that most peoples span of attention is for about twenty minutes at a time, which means that regular rest breaks are needed if an investigation goes on for any length of time. Observation sessions should be distributed either periodically (for example ve minutes each half hour), or randomly, for example for ve minutes beginning at 1000, 1032, 1119, 1250 etc. The decision to set up observation sessions randomly or periodically is often based on how much the frequency of activity varies periodically during the day. If there is some short term variation it can be a good idea to spread out the observation periods on a random basis. The success of observations are dependent on the how unobtrusively the observations are performed. Therefore it is important that the participants attention is not drawn to the observer each time a record is taken. The written log of observations are the raw data that normally must be analysed further, though an additional stage of analysis is needed if video recording is used. Videos of interactions can be of particular value when complex activities are being observed, and it proves difficult for an observer to record events in real time. One solution to this problem is to use more than one observer and for each observer to have different responsibilities. However this is not recommended for use in home or work settings due to the additional disruption caused by having more than one observer. In addition video recording can be particularly useful when observing activities with a strongly physical component e.g. the use of a wheelchair. Video recordings can be played back at the observers leisure and where appropriate also in slow motion or a frame at a time.
usert Tools
41
Data analysis
There are several ways to treat raw data from direct observational studies. The simplest way is to count frequencies and durations of different categories of activity. These categories are often the same as those used to focus
Example
The observations described in the group home example might be sampled several times during the day. Since the activities are obviously not very frequent, one must be assure that observation periods are long enough to cover relevant activities. It could also be expected that the frequency of relevant activities would vary periodically, therefore it was decided to sample activities at irregular times of the day. It was also decided that observation should go on for two days, and that a total of 5 one hours sessions should be sufficient and that they should be distributed differently the two days. Then 10 different starting times were generated (drawn randomly) and distributed so that the rst, third etc. was on the rst day, and the second, fourth etc. on the second day. This resulted in the raw data forms seen below.
Event Peter asks Paul about the time Paul does not answer Mary tells Peter that it is 0830 Peter says that he must go to school Mary tells Sara that she must go to the bus The taxi arrives to take Peter to school, the taxi must wait Sara leaves to catch the bus, (bus arrives in 30 mins, Sara needs ve minutes to get to the bus stop) Mary tells Paul that it is 9 o'clock. She shows him the clock on the wall and explains how the hands point to the time
Ends
08:47
09:00
09:06
usert Tools 42
Example
From the above example a tabulation of frequencies of some selected categories occurring in the specied periods during one week, might look like this:
1 16
0 33
From this tabulation it may easily be inferred that Peter and Paul have little understanding of time, and is constantly reminded by the helpers. Sara worries a lot, while Ben likes to tell others what time it is. The helpers very often tells the residents what time it is. It might have been concluded that Peter, Paul and Sara will benet from some automatic device that on pre-programmed hours reminds them of the time. Setting up a table of how the time is communicated might have shown that time should be stated in relation to an event, not as an hour of the day. If such a device would decrease Sara's anxiety is still to be seen. However, even in a case like this, with a rather clear conclusion, the total size of the potential user group remains to be assessed.
usert Tools 43
usert Tools 44