![[Main Interface Patient]](../figures/patient4.gif)
Fig. 1. Screen-dump: Overview Counselling Situation for Patient 3.
(1) Department of Social Science Informatics (S.W.I.)
(2) Multimedia Expertise Center (M.M.E.C.),
(3)Department of Clinical Psychology,
Faculty of Psychology,
University of Amsterdam,
Roetersstraat 15, 1018 WB Amsterdam, The Netherlands
Phone: +31-20-525 6788, Fax: +31-20-525 6896,
E-mail: bert@swi.psy.uva.nl
Publication details:
Webnet World Conference of the WWW, Internet & Intranet.
AACE: Association for the Advancement of Computing in Education.
H. Maurer and R.G. Olson (eds), Charlottesville, USA, pages 86-91,
November 1998.
There are a number of reasons why the counsellors that we are working with want to investigate the usability of the WWW as a communication tool between patient and counsellor [Lange and Emmelkamp 1997]. One reason is the assumption that the availability on the WWW will lower the barrier for people to engage in a counselling situation and that thus more people will be able to get help in solving their emotion problems. This lowering of the barrier can be explained by more then one argument. For example, as soon as counselling can be done using the WWW people can take therapy while staying at home. So it simply takes less effort to get started. An other issue concerns the fact that people don't easily go to a counsellor because of the anxiety that such an activity brings about. This is mainly a social or a cognitive barrier. A second reason for investigating counselling via the WWW follows from the fact that many people spend a lot of time communicating via the WWW (chatting, etc.). Apparently this way of 'talking to someone' is appreciated. There is also evidence that people 'tell more' on the WWW then they do in face-to-face meetings [Postmes 1997]. Consequently, using the WWW might be beneficiary for the therapy itself, because one of the key ideas of the counselling procedure is to have patients write extensively about their emotional problems.
Whether the counselling procedure works, and works better using the WWW, is not part of the research discussed in this paper. Here we focus on the design question, i.e. can we build an application for the WWW that facilitates task-based support for counselling situations. Many applications on the WWW are search-oriented'. Users surf the WWW in order to obtain information. Not many applications on the WWW support the execution of a task, although there are some examples in the domain of education [Parker Roerden 1997]. The realisation of task-based support for counselling situations is complex. At each point in the therapy the patient is required 'to specifically do certain activities' and 'may not do other activities'. The same holds for the counsellor. This situation is further complicated by the fact that what 'must' or 'may not be done' depends on the previous activities of the patient and the counsellor.
A second research topic concerns the difference between a task-based application on the WWW and a more regular 'stand-alone' application. Many things are known about the latter, which is not the case for applications on the WWW. Typically users can access the WWW when it best suites them. This is a problem for task-based applications on the WWW when the type and frequency of the interaction between the different parties is crucial. Almost needless to mention that this is the case in counselling situations.
Fig. 1. Screen-dump: Overview Counselling Situation for Patient 3.
When consulting the INTERAPIE WWW site a patient has two options: reading general information about the system or signing up for counselling. In the case of the latter INTERAPIE starts the screening activity. Each patient has to fill out a number of questionnaires. The questionnaires have two purposes. They are used to deny access to the system for patients that cannot be helped by the kind of treatment implemented in the INTERAPIE system. The second purpose of the questionnaires is concerned with information gathering about the patient that might be of use during the counselling. The screening is largely automated, i.e. the answers provided by the patient are stored in a database and procedures have been implemented to decide whether a patient can be allowed to use INTERAPIE. In the case of exclusion, the patient is presented a message explaining why he or she cannot be allowed to use the system. Additional information is given about where the patient may seek help instead.
Only one part of the screening is not automated. This part concerns the use of medicine by the patient. It was felt by the counsellors that this issue was too complicated to automate. In the current version of the INTERAPIE system the patient has to fill out a questionnaire about use of medicine. The information provided by the patient is presented to a counsellor who decides whether the medicine situation of the patient is problematic for the treatment as it is implemented in the INTERAPIE system.
Before the treatment may start a patient has to fill out an 'informed consent' (IC). This is a formal document including a signature of the patient by means of which the patient formally agrees to be part of the research project, and thus that the information about the patient's treatment (although anonymous) may be used for research purposes. Filling out an IC is a standard procedure in the Netherlands for doing psychological and clinical research with real patients. The handling of the IC could not be done fully automatic, because it requires a written signature of the patient.
If the screening is successful, the next main activity in the procedure is carried out by the counsellor. The counsellor has to construct an introduction ('right-arrow' icons in figures 1 and 2) for the patient. Most of the text needed for this can be taken by the counsellors from sets of pre-canned sentences, but the counsellor is free to deviate from this. Notice that, patients when they enter the INTERAPIE system are assigned to a counsellor ad random, although the assignment procedure makes sure that each counsellor has approximately the same number of patients.
Next the patient has to plan the first sequence of four essay writing sessions. This activity is shown to the patient in the interface by means of a 'clock' (figure 1, top row). By clicking on the icon a new WWW page is generated that allows the patient to plan the writing sessions. The planning tool implements certain constraints. For example, two essay writing sessions have to be separated by at least one 'free day' whereas at the same time there is a limit in the total amount of time that the four essay writing sessions may take (two weeks).
After the planning, the counsellor has to construct the specific instruction for the first two writing sessions (illustrated by the 'loudspeaker' icon). Next the patient performs two essay writing sessions followed by feedback and new instructions for the next two essay writing sessions. This first part of the treatment finishes with a feedback constructed by the counsellors on the third and fourth essay writing session. The second part of the treatment (Schrijf-opdracht 2) smoothly follows the first part and has the same set of activities, although the instructions, feedback, etc. will now have a different contents.
After finishing the second part, the patient proceeds with the ritual (figure 1, Eind-opdracht2). In figure 1 only a fraction of this third part is shown. The patient has just received the instruction specifying what to do next and is currently reading a previously written essay (nr. 8).
The counsellors may use pre-canned sentences for these activities. In fact, the counsellors that we are working with favour the use of such sentences because it improves the similarity between the treatments of different patients and thus provides them with a better situation for evaluating the usefulness of the treatment. INTERAPIE also supports the counsellors in how to read the essays written by the patients. However, none of the activities that have to be carried out by the counsellors are automated. Hence it is always the counsellor who interacts, although via the WWW, with the patient. This was also seen as an important requirement by the counsellors.
![[Feedback Workspace for Counsellor]](../figures/coun-feedback.gif)
Fig. 2. Screen-dump: Overview Counselling Situations for Counsellor.
A number of other things are worth mentioning. The set of icons in the patient interface that illustrate the activity that has to be carried out, and those that have been carried out, grow while the treatment is underway. In the beginning there is only the introduction icon. When the patient has made the planning, and the counsellor has 'replied' by constructing the instruction, the related planning and loudspeaker icons appear in the patient interface, etc. So the patient doesn't know beforehand how the treatment will proceed. Instead this is made clear to the patient, step-by-step, while following the treatment. At the left-hand side in the patient interface, five icons are shown that illustrate general issues. The topmost icon, currently showing a book and a pen, changes during the treatment, i.e. it always illustrates the activity that has to be performed. So we can infer that this particular patient has to write an essay, probably his or her final one. When clicking on the 'book and pen' icon the patient is presented a new WWW page which is basically a full-screen text-editor. When the patient is finished with writing, the essay can be saved and the patient can go back to the overview pages. The 'essay' icon then appears next to the instruction icon, meaning that this activity has been completed.
The counsellors mainly construct feedback and instruction. For these activities a dedicated interface is available. A counsellor usually has a set of patients (C=client) he or she is counselling at a certain moment. Therefore the counsellors must have means to easily overview the different counselling situations. This is realised by the interface part shown in figure 2. In fact, this is the screen the counsellors starts with after logging on to the INTERAPIE system. As all patients go through the same protocol an overview is generated by making a kind of table matching activity to patients. The black squares under the icons show activities that have been completed. At the end of each row, the current situation is highlighted. The red circle for patient 1, 2 and 4 means that the counsellor has to do something (feedback for 1 and 2, instruction for 4). At the end of the row for patient 3 no special cue is provided, which means that the patient has to do something. Sometimes a patient runs late with respect to his or her schedule. In such situations, a special icon is placed to illustrate this being late (patient 5). By clicking on one of the rows the counsellor gets into a workspace that matches the activity that has to be carried out. Finally, at the top of the interface five icons are shown that illustrate general activities. Respectively from left to right they represent: general help, new patient who's medicine situation has to be analysed (medicijn wachtkamer), new patient who's informed consent has to be checked (IC wachtkamer), the possibility to remove a patient from the INTERAPIE system (for instance because he or she has not followed the agreed upon schedule) (verwijderen), and logging out.
The task analysis was very important to the development of the INTERAPIE system. During this phase the tasks that the INTERAPIE system must perform were defined. Were needed these task were further decomposed into sub-tasks, until for each sub-task a detailed data-flow was available that clearly showed how the task could be implemented. Task analysis is not just concerned with specifying the tasks. Also important is task-allocation and the construction of a data-model. The former defines for each task the agent that has to perform it. In the case of our counselling situation, three agents can be defined: the INTERAPIE system, the patient and the counsellor. The data-model provides the basis for the construction of the database. It also specifies the data types that can be input and output of the tasks. The complete task analysis consists of specifications for approximately 50 data-flow diagrams for the leaf-tasks, i.e. the tasks at the most detailed level of specification.
While constructing the task analysis a number of remarkable issues emerged. Particularly, between counsellors conflicting statements were made about how certain treatment details should be realised in the INTERAPIE system. The counsellors also had a tendency to sometimes change their ideas during the constructing of the task analysis. Both facts illustrate the fact that the counsellors were still in the process of refining their ideas on the INTERAPIE system. In fact, many of the low level details have been decided upon while constructing the data-flow diagrams for the INTERAPIE system. When following a user centred design it is the task of the knowledge/software engineer to make sure that the domain experts reach a consensus on these issues. There were no major bottlenecks in this respect. The data-flow diagrams were finalised by defining a control structure on top of the tasks. The control structure specifies the exact order in which, and conditions under which, tasks have to be executed in the INTERAPIE system. To further minimise the step towards the implementation, this control structure, was then re-written in a semi-formal language.
The next crucial step in the design of the INTERAPIE system concerns the interface design, i.e. how do we visualise the interaction of the system with the patient and with the counsellor. This was a difficult problem to solve. One discussion focused on the question whether the interface should be 'text-oriented' or use mainly graphics. It was decided that the latter was the best option. An important reason being that icons can be used to illustrate many concepts, while usually only using little space in the interface. It was soon recognised that both the interface for the patient and for the counsellor would become very full otherwise.
In order enforce that both patients and counsellors follow the activities as scheduled in the treatment protocol, little cues have been used to highlight the next activity. First, the interfaces only visualises the activities that have to be carried out next (the last icon in the row), and the ones that have been done before. The current activity is further emphasised both in the interface for patient and for the counsellor. In the case of the former an additional icon is used to point to the current activity: in figure 1 the book and pen icon illustrating that an essay must now be written. In the case of the counsellor the current activities are emphasised by a red spiral (figure 3).
The treatment requires a strict schedule for carrying out the required activities. To enforce this regime a planning tool is used. After the plan has been constructed it is not directly visible in the interface, i.e. the patient has to click on the 'clock' icon to read his or her planning. The complexity of a plan and the planning tool made it impossible to show the results on the screen continuously. Also note that a plan is fixed. That is, after the patient has made it, and saved it, it cannot be changed anymore by the patient. If the patient gets behind on the schedule the interface will show this to the counsellor by placing the 'late' icon in the overview screen (figure 2). The patient on the other hand receives a standard text message in the case of being late.
Finally, note that the icons in the patient's interface are considerable larger then those in the counsellor's interface (compare figure 1 and 2). The main reason for the latter being smaller was lack of space. However, the idea was that for the counsellor this should not pose any problems. The counsellor spends more time working with the INTERAPIE system and is also more experienced with the set of activities within the treatment. Moreover, in the actual system the icons can be 'read' easily in the counsellor's interface.
The website of INTERAPIE was implemented using Macintosh computers. The webserver of INTERAPIE is WebStar. All the information that is used by the INTERAPIE system is stored in the relational database Butler. The interface of INTERAPIE is written in Tango. Tango is a program that communicates between the database Butler and by doing so can dynamically create HTML pages which can be sent to the webserver WebStar, and thus be made available for the outside world. For example, when a patient wants to see/read a previously written text, he or she has to click on the essay icon. This selection is presented to Tango by Webstar. Tango generates an SQL-query for Butler. As soon as Tango receives a reply from Butler it constructs a HTML page and presents this page to Webstar. Webstar puts the page on the patient's computer screen.
A difficult problem was the use of the many icons and the dynamic way in which each screen had to be constructed using a continuously changing set of these icons. Care had to be taken that the INTERAPIE system didn't take too long to put the next WWW page on the screen.
The experiment consisted of video-taping users while using the INTERAPIE system. The camera recorded the computer screen showing the interface and the sound (spoken words) made by the user. Users were asked to comment on the system as much as possible. Particularly, they were asked to mentioned unexpected phenomena, issues that confused them or they didn't understand, and things they thought were interesting or good. Next to video-taping, users were asked questions after a single INTERAPIE session (for instance, after writing feedback). Not all sessions were recorded completely, selections were made of periods of which we thought that the use of the system was most critical. For instance, when making the first planning, or reading the first feedback. In contrast, writing the 5th essay will probably not present much new insights on using the INTERAPIE system. However, during all sessions an observer was present to take notice of important events happening.
Some interesting results can be pointed out. Even though the INTERAPIE system had been tested before starting the above mentioned evaluations, some bugs still emerged during some of the sessions. One example was a limit on the number of characters that a certain variable in the program could manage. It turned out that the first instruction written by the counsellors had to be smaller that 200 words. The error was noticed only after the first 'large' message was written by one of the counsellors. Fortunately, the error could be repaired before the first patient wrote an essay. Particularly for the patients, who encounter a program error, this is a strange and sometimes frustrating event. Imagine the situation in which a patient has been writing an essay for almost an hour and now the machine does not want to store the text. This can be very frustrating for patients.
For inexperienced patients there are many standard computer interface issues that are unclear. If someone has never used a computer, things such as a scroll-bar, a hyperlink, etc. are not so obvious. If you haven't worked with the WWW before and you do not know that sometimes it takes a while for the files to appear on your screen, you may get confused. In some case this resulted in repeatedly clicking on some icons (which made the problem only worse). It was in this respect remarkable to notice that the general help pages were almost never consulted by the patients. This may indicate that the system was still easy to understand and use for them.
The counsellors complained that while constructing a feedback or an instruction they had insufficient overview of general facts about the patient. For example, the age of a patient cannot be assessed by the counsellor while constructing a feedback or an instruction. Although this information is available in the INTERAPIE database, the counsellor can 'only' see/read the essays written by the patient. Counsellors also would like to have a kind of patient specific notebook for making statements about the patient. During the treatment the counsellors want to use this as an external memory on how they are interacting with a patient.
Each time a new patient enters the INTERAPIE system, he or she is assigned to a counsellor. In the current version of the INTERAPIE system each counsellor can get 6 patients at the most. If at some point patient 'number 25' tries to enter the system he or she is not accepted. In the current version of the system this means that the patient receives a message saying that the system is full. In future implementations a kind of waiting room should be constructed in which patients can wait a few days (or weeks) until counselling space is available.
Although the INTERAPIE system has continuously been debugged during the evaluation phases, and at this moment may seem 'to be without many problems', this doesn't mean that in the end the above problem has been solved. There may always be some reason for the required interaction between the patient and the counsellor to be hampered. What do to if someone falls ill? At least some form of meta-communication channel should be available next to the INTERAPIE system to inform users about the meta-problems with the system or related issues that effect them.
Related to the above issue is the need for having a kind of super user (cf. UNIX operating system) to manipulate the contents of the database if required. Due to different kinds of circumstances small modifications had to be made to the database entries that the INTERAPIE system uses. For instance, re-allocating a patient to a different counsellor, or modifying the writing schedule of the patient. In the current version of the system changing something in the database that is outside of the scope of the counselling protocol requires detailed knowledge on how the system has been programmed and can thus only be performed by the system programmer. In a future implementation of the system a super-user facility should overcome this problem.
Next to simply improving the current version of the system, plans have been discussed for a larger INTERAPIE system that might work within a city, nation or maybe even on a world scale. Of course, this would require solving many new problems. Language problems may not even be the most complicated ones. Another plan for future work concerns the construction of 'INTERAPIE-like' systems for other psychological or medical problems that can be treated using a largely fixed protocol as in the case for the treatment of traumatic or stressful life events.
[Lange and Emmelkamp 1997] A. Lange and P.M.G Emmelkamp. Laagdrempelige Psychologische Hulp via Internet, Research grant proposal, NFGV, 1997 (in Dutch).
[Parker Roerden 1997] L. Parker Roerden. NetLessons: Web-Based Projects For Your Classroom. O'Reilly, Sebastopol, CA, USA, 1997.
[Postmes 1997] T. Postmes. Social Influences in Computer-Mediated Groups. PhD thesis, University of Amsterdam, Amsterdam, The Netherlands, 1997.
[Preece et al 1994] J. Preece, Y. Rogers, H. Sharp, D. Benyon, S. Holland and T. Carey. Human-computer Interaction. Addison-Wesley, Wokingham, England, 1994.
[Schoutrop 1997] M. Schoutrop, A. Lange, G. Hanewald, C. Duurland and B. Bermond. The effects of Structured Writing Assignments on Overcoming Major Stressful Events: An Uncontrolled Study. Clinical Psychology and Psychotherapy, 4(3):179-185, 1997.
Last modified on May 9th, 2001