Designing for Mobile Devices: Requirements, Low-Fi Prototyping and ...

1 downloads 340 Views 89KB Size Report
This paper describes the design process of a set of ubiquitous applications for critical ... can be used in mobile application's design. Keywords: low-fidelity ...
Designing for Mobile Devices: Requirements, Low-Fi Prototyping and Evaluation Marco de Sá and Luís Carriço LaSIGE & Department of Informatics, University of Lisbon, Edifício C6, Campo Grande, 1749-016, Lisboa, Portugal {marcosa,lmc}@di.fc.ul.pt

Abstract. This paper describes the design process of a set of ubiquitous applications for critical scenarios (e.g., psychotherapy and education). Accordingly, we address the problems that occurred on the various design stages, particularly those that pertain to the mobility and ubiquity of the devices. Regarding these, we detail the various solutions that were adopted along the way, particularly on the data gathering and requirements’ assessment, prototyping and evaluation stages. We introduce a set of dimensions to the concept of context and how it is utilized on the design of mobile applications. We describe new prototyping techniques and explain how they improve usability evaluation. Overall, we aim to share the learnt lessons and how they can be used in mobile application’s design.

Keywords: low-fidelity prototyping, ubiquitous computing, mobile devices, usability evaluation.

1 Introduction Mobile devices possess outstanding capabilities that set them apart from common fixed technologies. Their size, weight and corresponding portability, interaction modalities and even multimedia features allow them to be used ubiquitously on several situations and contexts for a wide and diverse set of purposes [8],[12]. They have propelled the surfacing of concepts such as ambient intelligent and context aware applications and defined new usage paradigms on various levels. However, these same potentialities, together with inherent hardware limitations (e.g. small screens, battery life, input capabilities) introduce greater challenges when designing specific applications that target these devices. Consequentially, new requirements and usability guidelines are paramount in order to provide users with applications that enhance their tasks and offer support to pervasive computing, but still maintaining the same usability levels that users are now accustomed to. While attempting to overcome such difficulties, designers often choose to port existing applications, available for larger platforms, to mobile devices, adjusting the necessary details. However, such approach generally leads to cumbersome and

unusable applications that, even if containing brilliant content, quickly become obsolete and avoided by users [3],[11]. A major factor for designers to adopt such strategies is the absence of specific methodologies for mobile, handheld and ubiquitous devices [14]. Although some guidelines, available in current user centered design (UCD) methodologies, might apply to this specific design process, the unique features and constraints that ubiquity and pervasiveness, require new or, at least, deeply refined approaches. Three design stages are particularly sensitive and present a wider set of difficulties [15]. Gathering requirements is the essential bootstrap for every development and design process. For this purpose, current design methodologies include methods such as ethnography, task analysis, inquiries amongst others. However, constantly using a mobile device and application through several tasks and settings introduces details that are hardly detectable with the simple use of such methods. The influence of the context on a specific scenario might be acceptably assessed. Still, the implications that changing from one context to another, and how this might influence the user and his requirements towards the applications fail to be understood. On the sketching process and evaluation of low-fidelity prototypes, several problems that retract from the process are also evident. Low-fidelity prototypes are, as currently used, poorly suited to the peculiarities of mobile devices and their usage scenarios [15]. Using paper sheets and a set of detachable cards to simulate a UI while a designer uses a Wizard of Oz technique is an acceptable way to test a future desktop application. However, it his likely that such technique fails to provide a clear perspective on usage problems for a mobile device used on a multitude of locations [15]. Furthermore, the technique per se seems difficult to apply on such conditions. These problems extend themselves to later stages of evaluation as well. Overall, the difficulties inherent to the design process of applications for such devices urge for different approaches and extensions on current user centered design processes. This paper describes the design process of two different, proactive, mobile applications and how common user centered design approaches were applied. The problems faced throughout the process are stressed highlighting the specific challenges that mobile platforms imposed; how current methodologies fail to properly support it; and how these facts motivated the proposed solution. We start by enumerating the problems that motivated our work and explaining how existing work was insufficient to solve them or, when adequate, facilitated the process. Afterwards we detail our experiences and the contributions that emerged. Finally, we discuss future work and draw some conclusions.

2 Design problems The main motivation behind the work hereby presented resulted from the difficulties felt during the design process of two different mobile applications. The first application is directed to psychotherapy (e.g., Cognitive Behavioral Therapies for depression and anxiety) [17] while the second supports educational activities [16]. Both systems target mobile devices with particular focus on PDAs and were designed using a user centered design methodology [2]. These two choices resulted from the

need to support the highly pervasive and ubiquitous activities that compose the whole of these problems, and the need to understand the critical requirements that users have. Data Gathering: Difficulties emerged immediately with the initial requirement assessment and data gathering stages on both projects. UCD methodologies [2],[13] encourage data gathering and analysis of requirements from a user centered perspective. Accordingly, they suggest the use of ethnographic sessions and interviews to gather the necessary information. These two techniques usually bring designers the possibility to understand the work flow on various activities and detect problems that might be solved by introducing a technological system. However, on pervasive and mobile activities as those involved in psychotherapy and education, these methodologies fail to provide sufficient detail and guidelines on how to achieve such results and on how to apply the techniques [15]. On the psychotherapy project [17], ethical issues prevented us from conducting ethnographic sessions and analyzing in-therapy procedures. In order to understand common practices and activities we were required to review videos [5] and interview therapists. Therapy sessions were necessarily simulated and conducted within controlled environments with the participation of therapists and simulated patients. No direct communication with real patients was allowed. For Test-IT [16], the second project, which supports a wide role of educational activities on different settings, the same techniques were adopted. Ethnographic sessions were conducted within school classes and interviews were made to teachers and students. These processes introduced multiple problems. Firstly, we were not able to accompany any of the users (e.g., therapists, teachers or students) on their outdoors tasks. Secondly, as we tried to introduce some scenario-based simulations [8], covering a set of activities and contexts (e.g. complete specific tasks on various locations), we noticed that these were frequently incomplete and too artificial, failing to cover important details. Thirdly, since scenarios were mostly self-contained and specific, no context changes were assessed or even noticed. For instance, on simulated sessions, as patients continued their work outdoors and at home, no information could be retrieved, creating a large gap on the workflow and gathered data. Although this scenario-based exercise provided insight on some problems that could occur on several activities, it still falls short to cover most of the possible scenarios and situations. Prototyping: Once the gathered information was modeled and analyzed, we started to design and prototype possible solutions. Various examples on building low-fidelity prototypes were reviewed, some particularly applied to mobile devices [1],[8],[19]. However, these few examples of low-fidelity prototyping for mobile devices that were found, used classic prototyping approaches, with simple cards or paper, and low accuracy regarding the screen measurements and available components (e.g. drop boxes, text fields, etc) [18],[19]. This introduced a set of problems on the usage of these prototypes. Components and their size confounded users and resulted in unrealistic and unviable prototypes.

Since screen measurements were disregarded, the information that each card started to include was excessive and impossible to include on actual devices. Moreover, together with the design fidelity details (e.g., size, components), more realistic tools than post-its or cards, as exemplified in [2],[18],[19] were needed. These materials started to mislead users regarding weight, size or the interaction with the devices. Furthermore, the cards started to deteriorate during outside evaluation sessions, especially each time a user tried to put a paper PDA prototype in his/her pocket. This affected the experimentation and evaluation of the sketches that were available and confused the UI design problems with those that depended on the device and used materials. These problems retracted from contextual evaluation sessions, requiring the construction of new prototypes each time new ideas emerged. Once again, alternatives had to be introduced so that the process could provide useful information. Usability Evaluation: Usability evaluation out of the lab, especially on early design stages, which proved to be essential, is not yet a common procedure [10]. Accordingly, the used methodology provided no insight on how to conduct any type of multiple context/setting evaluation. Although some recent work addresses this subject [7],[8],[10], stressing the need to evaluate the prototypes on the real situations where they will most likely be used when completed, these practices are rarely applied [10]. On the majority of cases, applications are tested within labs, generally taking advantage of simulation and role playing [10],[18]. Furthermore, despite the existing literature, no substantial advice on this topic was found. For instance, the use of Wizard of Oz techniques [9] with mobile prototypes was extremely demanding. Following a user while he/she tests a prototype on every imagined scenario might be impossible to achieve. The procedure itself might cause ethical and social problems. This problem was further aggravated when trying to evaluate the software applications. Here users required privacy and long periods of time using the applications. Thus, no observation and detection of usability flaws could be made. Once again, the used methodology provided no guidance on such procedures and solutions had to be arranged.

3 A Design Methodology for Mobile Devices In order to overcome the stated difficulties, we were required to use alternative techniques on all these design stages. Furthermore, this process propelled the development of new methods which coped with the user centered design of the two user interfaces. The following sections detail how and which techniques were used and some findings that emerged along the way.

3.1 Pervasive Data Gathering For designers to understand currently faced problems and define more suited solutions, they must work together with future users and understand their needs so that created alternatives are innovative and usable. As mentioned before, existing techniques provide designers with some directions to gather this information and analyze it. However, these techniques focus mainly on a specific setting and scenario. Even if they are repeated for every imagined scenario and setting, none of these methods provides means to relate the gathered requirements when settings are changed. Furthermore, the requirements and their causes are commonly lost during these transitions. This problem is generally non existent on fixed solutions since scenarios are usually immutable or usage restrictions, and their influence on the user’s behavior, are constant. With mobile technology the case is hardly the same. Due to its pervasive and ubiquitous nature, the context, settings and usage scenarios are constantly changing, such as the requirements and needs that these imply. Therefore, it is crucial to understand the changes between settings and scenarios, which is possibly attainable with current methodologies, but it is also paramount to understand why and how these changes occurred. For example, using ethnographic observations, one can notice that during meetings, only two or three tasks are accomplished by a user on a certain setting and with a particular sequence. On the other hand, while at home only one of these tasks is done and arranged on a different sequence, within a set of other specific tasks. Furthermore, even within one of these settings, depending on privacy or social settings, the sequence may be changed once again. However, there is no explanation on why and how these changes on usage behavior happened and how this should be modeled into a future solution. Moreover, when modeling an application, the UI’s changes and triggers are also difficult to picture. To tackle these problems, we tried to use alternative methods to evaluate and assess requirements on mobile settings. Particular care was directed to understand how usage and user’s behaviors are affected by the surrounding environment, emphasizing not only the changes but also their causes and effects. Firstly, we established a high level conceptual framework, which pointed specific directions and highlighted crucial concerns that had to be taken into consideration. We started by selecting some scenarios which included a set of details that needed be taken into account while designing our applications, each including various contexts. Initially, we used dimensions of context that had been previously defined [6],[14]. Nevertheless, we extended the notion of context and included two other dimensions that seemed important to our study and which focused more the users and their behavior on the real settings and not only the system. First, to assess the influence of personal, social, cultural, and cooperation issues on each context, we tried to establish the notion of user-context. Mainly, this dimension focuses details that affect the user directly or that depend on his/her activities. This allowed us to categorize details such as: (1) the users’ positions while working (e.g., walking, standing, seating, etc); (2) different social/cultural events and distractions (e.g., conversation, school, work, meetings, on the bus). Secondly, we also wanted to detect influences that were directly related to the location in which the user was. Among others we included loud and quiet locations

with and without constant distractions, lightened and dark locations, etc., to understand how these affected the usage of the applications. The aim was to provide guidance on the requirements analysis’ first steps focusing our efforts to scenarios and designs that were particularly relevant for each of the case studies. Still, the amount of scenarios was not extensive. However, the selected ones contained the majority of aspects that we wanted to assess. Afterwards, we relied on actual users to gather information and requirements on realistic scenarios. To achieve so, we provided users with adequate questionnaires, containing contextual questions and that needed to be filled-in on specific situations. This allowed us to gather contextual data without following users to analyze their work on realistic settings. A similar technique, called the Experience Sampling Method, has been used to evaluate ubiquitous applications and systems by Consolvo [4] with positive results. Our approach was similar but focusing on data gathering instead of evaluation. Furthermore, we used fewer resources (e.g., paper and pen instead of PDAs). These methods allowed us to detect activities that extended through several settings and acknowledge those that were specific to certain locations. Moreover, we were able to detect how the users’ needs changed according to their surroundings (e.g., social and physical). 3.2 Early Stage Design and Prototyping Since common prototyping techniques started to mislead users and hinder the overall design and evaluation process, we tried some new techniques. These resulted in a set of guidelines which can help application and UI developers to create suitable lowfidelity prototypes for mobile devices. While prototyping our applications we created a frame using a light wooden material with about the same weight and size of a Palm Tungsten T3 (Figure 1). The frame had a small opening on the top, which allowed the cards (screens) to be switched very easily. It took about forty minutes to build. Even so, this allowed the future users to have an accurate notion on the device and interaction techniques that they would use. Furthermore, it was quickly noticeable that this procedure allowed us to solve a list of UI problems that would not have been found using just cards. For instance, buttons on the bottom of the screen had to be moved to the top for easier access and so that they would not move or disappear if the screen was extended (screen extension is a T3 feature). Also, we detected that buttons and lists that extended to, or were placed near, the borders of the screen were difficult to read and interact with. This rigid frame and the realistic prototyping techniques also provided additional benefits. Using it, users were able test the applications while walking from one place to another, switching screens at will. This showed that the size of the buttons that were used needed to be larger when the user was walking, which was a common situation in some tasks. Moreover, users that were involved on the initial tests, with traditional techniques, clearly noticed the difference. They all preferred the frame we created since it gave them a realistic feeling regarding interaction and problems while using the devices.

Rigid Frame with accurate dimensions. Slide-in card/screen slot Paper sketch composed by a card/screen

Fig. 1. Low-fi prototype of a PDA with a slot that allows users to easily exchange screens.

The experience showed that low-fi prototypes can be effectively used for mobile devices. However, the following findings and suggestions should be taken into consideration: − Common techniques and material do not apply directly for most mobile devices/applications. − A careful distinction between the device prototypes and the application/UI prototype has to be made. − The device prototype should be created using rigid materials in order to be used in real-life settings and it should have approximately the same dimensions of an actual device. − A slot where cards can be easily and quickly inserted and removed needs to be available. When this option is unavailable, alternatives that facilitate screen substitution should be provided. − Sketches must be drawn with the same size of the device’s screen using similar components and fonts to those available for a real device. − The interaction type of the actual device should also be emulated (e.g. stylus/pen, joystick/small colored drawing pin or a thumb tack). − Although these hints might imply more effort from the designer, they are compensated with more accurate results. 3.3 Mobile Evaluation Evaluation with low-fidelity prototypes: Complementing the aforementioned procedures to create low-fidelity prototypes, the corresponding evaluation procedures were also updated in order to take advantage of these. Particularly, it was essential to emphasize the importance of context and location while evaluating mobile applications. In fact, as already concluded, these are usually tested on controlled

settings and fictitious scenarios retracting from the evaluations sessions and preventing the detection of some usage errors. During the early evaluation of the two previously mentioned applications the context and settings of use proved to be of extreme importance. With the developed frame and prototypes, users carried the sketches and simulated applications from one place to another. Besides evidencing some design problems that could occur, the ubiquitous evaluation sessions promoted participatory design, allowing users to engage actively on the process. In fact, most of the innovations appeared when testing the prototypes on real situations and scenarios. On the sessions that occurred several findings were made. For instance, using different colored pencils to interact with the UI, allowed us to easily trace the users’ behaviors. Furthermore, even the number of clicks or taps the user had done, as well as the user’s accuracy on different settings (e.g. seated, walking, etc) could be verified. Besides, users engaged on the sketching process by themselves, providing input for every feature that they needed according to the situations that appeared. Some of the users even requested the frames that were used so they could use the prototypes at home; and since each user followed the color scheme that was provided, (one color for each task) their performance was also evaluated afterwards and without their presence. This detail is particular interesting for mobile applications where the user can not be followed the entire day. On this stage some of the main conclusions were: − Using rigid materials on the prototypes allows users to carry them, keep them in their pockets, take them home and participate on the sketches’ design directly on the device, conducting evaluation on relevant locations. − Context of use is essential for effective usability testing and evaluation should be done in several possible scenarios. − When using adequate material, users are allowed to participate on the sketching process during contextual sessions, further promoting innovation and adequate ideas that cope with the contextual requirements. − Tasks should be defined previously (e.g. using scripts), but the possibility of creating new features and test them on the spot should be provided. − Color schemes can be used to trace users’ activities after the evaluation sessions. − Questionnaires can also be given to users together with the prototypes, so that they can complete them during out of the lab sessions (e.g., ESM).

4 Discussion and Future Work So far, our work allowed us to detect several shortcomings on existing design methodologies when applied to the design of mobile applications and corresponding user interfaces. These findings, and the “on-the-fly” solutions that were used to cope with them, motivated us to develop an entire methodology with techniques that apply specifically to mobile devices and ubiquitous applications. Globally, these problems led us to believe it was necessary to introduce solid guidelines on the various stages of design focusing on the mobility factor and on the users’ varying context.

Although this paper focuses the initial stages of the design process, we are currently working on new notations that tackle the constant behavior and context change during the specification and even evaluation stages. Furthermore, besides the prototyping and early evaluation techniques that were described, we are developing simple and low-cost methods for ubiquitous usability evaluation on later stages. These techniques take advantage of common equipment (e.g., laptops and webcams) and provide rich and useful insight on users’ behavior without direct observation. Also, with the experience gained and ideas that resulted from this process, we started to develop a prototyping framework for mobile devices. Overall, the tool aims at supporting prototyping and evaluation of applications with the peculiarity of creating the prototypes on actual devices. Some of its contributions are (1) the support for prototypes with various fidelities; (2) an automatic mechanism to gather contextual data; (3) the inclusion of ESM and probing techniques; (4) the support of on-the-spot participatory design and (5) the use of pre-established design guidelines.

5 Conclusions Current design approaches, although providing extensive detail on various stages of the design process, fail to support most of the related activities and tasks when mobility of the system is paramount. Existing techniques, particularly for initial stages of development and evaluation, are rarely applied, especially taking into account the volatile settings in which the applications are likely to be used. Furthermore, these techniques aren’t always applicable and, when ill-used, hinder the overall process. In this paper, we described the problems that emerged while using a common user centered design methodology on the development of two ubiquitous systems. Throughout this process, mobility, pervasiveness and ubiquity required the adoption of new techniques and the adjustment of existing ones in order to cope with the challenges that derived from the constant change of context and multitude of settings that could be envisioned for a specific system. Overall, these techniques and resulting findings allowed us to gather data on relevant locations and settings. Moreover, they propelled the detection of user interface problems on both low-fidelity and high-fidelity prototypes during the various evaluations sessions that we conducted. In comparison with existing techniques, which were used on initial stages of this process, these provided insight to problems which would not have been detected otherwise. Finally, the definition of a high level conceptual framework which enclosed criteria and scenarios for data gathering and analysis also improved the design process, on its various stages, and allowed us to focus on significant details. Acknowledgments. This work was supported by FCT through project JoinTS and LaSIGE.

References 1. Ballagas, R., et al. iStuff Mobile: prototyping interactions for mobile phones in interactive spaces. in PERMID 2006, Workshop at Pervasive 2006. 2006. Dublin, Ireland. 2. Beyer, H., Holtzblatt, K., 1998. Contextual Design: Customer Centered Approach to Systems Design, Academic Press, San Francisco, CA, USA. 3. Brewster, S. (2002). Overcoming the lack of screen space on mobile computers. Personal and Ubiquitous Computing, 6, 188–205. 4. Consolvo, S. and M. Walker, Using the Experience Sampling Method to Evaluate Ubicomp Applications. IEEE Pervasive Computing, 2003. 2(2): p. 24-31. 5. Davidson, J., Persons, J.B. & Tompkins, M.A. (2000). Cognitive-behavior therapy for depression. 6. Dix, A., et al., Exploiting space and location as a design framework for interactive mobile systems. ACM Transactions on Computer-Human Interaction, 2000. 7: p. 285-321. 7. Duh, H.B.-L., G.C.B. Tan, and V.H.-h. Chen. Usability Evaluation for Mobile Device: A Comparison of Laboratory and Field Tests. Mobile HCI'06. 2006. Helsinki, Finland: ACM. 8. Jones, M. and G. Marsden, Mobile Interaction Design. 2006, West Sussex, England: John Wiley & Sons Ltd. 377. 9. Kelley, J. (1984) An iterative design methodology for user-friendly natural language office information applications. Transactions on Office Information Systems, 2, 26–41. 10. Kjeldskov, J. and C. Graham, A Review of Mobile HCI Research Methods. Mobile HCI'03. Vol. 2795/2003. 2003: Springer Berlin / Heidelberg. 317-335. 11. Lee, Young E. Interface Design for Mobile Commerce. Communications of the ACM, Vol.46, No. 12, December 2003, pp. 49-52. 12. Luchini, K., Quintana, C. & Soloway, E. (2004). Design guidelines for learner-centered handheld tools. In Proceedings of the CHI2004, ACM, Vienna, Austria. 13. Mayhew, D.J. (1999). The Usability Engineering Lifecycle. Morgan Kaufmann, San Francisco, CA, USA. 14. Raptis, D., N. Tselios, and N. Avouris. Context-based Design of Mobile Applications for Museums: A Survey of Existing Practices. MobileHCI'05. 2005. Salzburg, Austria, ACM 15. Sá, M., Carriço, L., Low-Fi Prototyping for Mobile Devices. In Proceedings of CHI'06 (extended abstracts), Montréal, Canada, pp. 694-699, ACM. 16. Sá, M., Carriço, L., Handheld Devices for cooperative Educational Activities. In Proceedings of SAC'06, ACM Symposium on Applied Computing, Mobile Computing and Applications Track, Dijon, France, pp. 1145-1149, ACM Press, April, 2006 17. Sá, M. Carriço, L., Antunes P., Ubiquitous Psychotherapy. IEEE Pervasive Computing, Special Issue on Healthcare, IEEE Press , 2007. 18. Svanaes, D. and G. Seland. Putting the users center stage: role playing and low-fi prototyping enable end users to design mobile systems. in CHI'04, SIGCHI conference on Human factors in computing systems 2004. Vienna, Austria ACM Press. 19. Weiss, S. (2002). Handheld Usability. John Wiley and Sons, England.