Model-Based Design and Evaluation of Interactive Applications

Transcription

1 Model-Based Design and Evaluation of Fabio Paternò CNUCE-C.N.R Pisa, Italy Structure of the Tutorial n Introduction to Model-based Approaches n State of art in Task Models n ConcurTaskTrees n Task Models for Cooperative Applications n Exercise n Task Models in User Interface Design n Task Models in Usability Evaluation

2 Why Model-based approaches? Highlight important information Help to manage complexity Useful to support methods Significant Models in HCI Task models Cognitive architectures User models Domain Models Context Models Presentation Models Dialogue models

3 UML and HCI Model-based approach Nine notations Lack of consideration of the user interface How to improve it? Definitions Task Goal Basic Task Task Analysis Task Model

4 Task Analysis (example) Tourist general and clear information in limited amount access by spatial representations guided navigation Art student some basic knowledge choice among different types of information Expert precise requests detailed description of the information requested Engineering task models Flexible and expressive notations with a precise semantics Systematic methods able to indicate how to use information in the task models Availability of automatic tools to use such information efficiently

5 The many possible task models Existing System Envisioned System User Use of Task Models Improve understanding of the application domain Record the result of interdisciplinary discussion Support effective design Support usability evaluation Support the user during a session Documentation

6 Representations of Task Models Hierarchical task analysis GOMS family UAN Different syntax (textual vs graphical) Different level of formality Different set of operators for task composition GOMS Example GOAL: EDIT-MANUSCRIPT GOAL: EDIT-UNIT-Task repeat until no more unit tasks GOAL: ACQUIRE-UNIT-TASK GET-NEXT-PAGE if at end of manuscript GET-NEXT-TASK GOAL: EXECUTE-UNIT-TASK GOAL:LOCATE-LINE [select: USE-QS-METHOD USE-LF-METHOD] GOAL: MODIFY-TEXT [select: USE-S-METHOD USE-M-METHOD] VERIFY-EDIT

7 Limitations of GOMS It does not consider user errors It does not consider the possibility of interruptions It considers only sequential tasks It can be inadequate for distributed applications (such as web-based applications) UAN - User Action Notation The user interface is represented by a hierarchy of asynchronous tasks user action and system feedback are specified at a low level textual notation

8 Example of UAN specification Task: BuildRequest: ((SelR ClearR IconifyR)* --> SpecField+) Task: SelApplication User Action Interface Feedback Interface State ~[x,y in AppICON] (t<tdoubleclick) w!: w -! UnMap(PrevAppliMenu) Map(AppMenu) UnMap(AppICON) CurAppli=App CurMenu=AppMenu ConcurTaskTrees Focus on Activities Hierarchical Structure Graphical Syntax Rich set of temporal operators Task allocation Objects and task attributes

9 Task Models vs Scenarios Scenarios are informal descriptions of a specific use in a specific context Task models describe the possible activities and their relationships Scenarios can support task development Task models can support scenarios identification Moving from scenarios to tasks

10 Temporal operators Enabling Disabling Interruption Choice Iteration T1 >> T2 or T1 [ ]>> T2 T1 [> T2 T1 > T2 T1 [ ] T2 T1* or T1 {n} Concurrency T1 T2 T1 [] T2 Optionality [T] Task Types Overview Comparison Locate Grouping Design and Evaluation... of Interaction tasks Selection Edit Control Application task

11 Inheritance of relationships Relationships task/subtasks

12 Optional tasks Tool Support in CTTE Flexible editing of the task model Using informal descriptions in modelling Checking completeness of the specification Saving the specification in various formats Simulating the task model Comparing task models Running scenarios

13 The CTTE editor Simulating dynamic behaviour

14 Comparison of Task Models Modelling Multi-User Applications t1 User1 t2 User2 User3 t3 t1(user1) t3 (User3) Cooperative part

15 Cooperative aspects Exercise Develop a task model in ConcurTaskTrees with at least three levels and ten tasks using the CTTE tool

16 Limitations of current approaches in UI design Visual tools do not support mapping between logical activities and UI elements UML is oriented to the system design No available tools for task-based design (when existing, not publicly available and limited functionality) Task model - based design 1. Analysis of operators amongst tasks 2. Analysis of each task (objects, attributes,...) Identify structure of dialogues Choose a suitable presentation

17 Task-based design Temporal relationships among tasks Frequency of single task and task patterns Interaction and presentations techniques suitable for task types considering type and cardinality of data Multiple consistent representations when an object supports multiple tasks Automatic support to cognitive intense problem solving user tasks Criteria to identify presentations One single presentation supporting all possible tasks As many presentations as the number of enabled task sets Intermediate solution

18 Mapping task-interaction technique: single choice exam. Known domain: Number of possible values [2,3] Number of possible values [4,7] Number of possible values [8,50] Number of possible values ]50,+ [ Control tasks Disabling tasks ([>) should be located so as to highlight their control function

19 The task model structure Grouping tasks that share the same parent task Communicating concurrent tasks ( [] ) are presented close to each other Supporting enabling operator Both sets are shown in the same presentation unit (when they exchange information, []>>) The sets are presented at different times The sets are presented in different presentation unit at the same time Situation 1 Situations 2 and 3 Name: Jo* Results: Name Age Johansen 52 Jones 27 Joxibon 18 Submit Name: Age: Jo* >20 Submit Results: Name Age Johansen 52 Jones 21

20 Deriving information from operators Task-interface relationships

21 Tasks-interface relationships Tasks-interface relationships

22 The Aerodrome case study An excerpt of task model

23 The resulting dialogues -1 The resulting dialogues -2

24 The resulting dialogues -3 Mobi-D: Example of toolsupport for model-based design

25 Task Patterns Identification of solutions for problems recurring in different applications Reuse of both design and software solutions Possibility of obtaining more compact specifications Patterns P1 Name P2 Problem P3 Task specification P4 Objects specification P5 Scenario of use P6 Possible subpatters P7 Aspects which can vary P8 Applications that can use it

26 Adaptation Adaptibility vs Adaptivity Dimensions: User, Device, Context,... Usability issues Different initial accesses Student Tourist Expert

27 Different presentations What about adaptive support? What is it? Techniques supported (direct guidance, adaptive ordering, hiding, annotation) Advantages Disadvantages

28 Example of adaptive support Exercise Design a user interface corresponding to the task model developed in the previous exercise

29 Usability Evaluation Heuristic evaluation Cognitive Walkthrough Guidelines Empirical testing Model-based evaluation Inspection-based Evaluation Use of CTT models in the evaluation phase Use of a the set of deviation types To evaluate a prototype

30 The method Development of task model of the application to evaluate Analysis of deviations related to basic tasks Analysis of deviations in high-level tasks Interdisciplinary analysis Organisation of the exercise Introduction of the method Introduction of the prototype Short use of the prototype Prototype available during exercise Task model available to evaluators Audio record

31 Organisation of the exercise Evaluators drive discussion Questions derived from task model + classes of deviations They are related to user interface elements Identification of the information necessary for each task performance Tables (useful more as documentation or design rationale) Refined guidewords (basic tasks) NONE: initial information, performance, result OTHER THAN: less, more, different ILL-TIMED: too early, too late

32 Resulting tables Task: Check Deviation Guideword: None Explanation Causes Consequences Protections Recommendations No input The system is The controller The controller Duplication of The controller broken. has not an looks at the communication system has not Perception updated view of window to information problem the air traffic check the air traffic concerning the current traffic Controller is distracted or interrupted by other activities Perception problem No Controller is performance distracted or over The controller confident has the Interpretation information problem but he doesn t check it carefully Pilot calls Provide an automatic controller to warning message only check the path when an aircraft is deviating from the assigned path. Pilot calls controller to check the path No output Controller is The controller interrupted by finds a other activities deviation but immediately forgets it Interpretation for high-level tasks NONE: initial information, performance, result OTHER THAN: less, more, different ILL-TIMED: too early, too late

33 The MIDAS User Interface Example of issue (1) Check deviation - NONE/No performance Causes: workload, memory Protection: Red line connecting two flight labels Recommendation: possibility of audio warning

34 Example of issue (2) Check deviation - NONE/No input Causes: Zoom activated Protection: Button to default view Recommendation: Two views (global+local) Example of issue (3) Send stop - Different performance Causes: action slip Protection: send immediately new clearance Recommendation: Highlight the most safetycritical element

35 Social constraints Developers tend to defend every decision taken Users tend to divagate (sometimes useful, sometimes confusing) Time pressure RemUSINE Remote USability INterface Evaluator Bridge between empirical and model-based evaluation Flexible and expressive task models Limitations of usability laboratories and need to decrease the costs of usability evaluation Application to case studies

36 The architecture of RemUSINE Designer site ConcurTaskTrees Editor Task model Designer Designer RemUSINE (Preparation) Log-task table Evaluation Method (RemUSINE) User1 site User1 logs Logging tool Application.... Usern site Usern logs Logging tool Application User1 Usern The Preparation Part

37 An example of analysis 1.JS.applet("Insertions",0).member(.. ).multiclick(124,93,16,1); 2. JS.delay(7310); 3.JS.applet("Insertions",0).button("Sculptors").typeString("Bodin"); 4.JS.applet("Insertions",0).member("java.",0).multiClick(...); 5.JS.applet(...).member(, 0).typeString("Bodini Floriano",..); 6. JS.applet("Insertions",0).button( Periods ).buttonpress(); 7.JS.applet(...).member( ).member( ).select(4,"xx sec."); General Information

38 Automatic analisys of the log Task Patterns

39 Analysis of time performance Errors Detected (example 1)

40 Errors Detected (example 2) Errors Detected (example 3)

41 How to interpret the results Task model mismatch The application is too rigid The user interface does not provide enough support to understand how to perform the tasks RemUSINE points out where the problem is! Results of RemUSINE The analysis is reliable and complete Identifying the violated precondition helps to understand the reason for the error It gives indication of tasks never performed It provides quantitative information

42 Time required by RemUSINE Developing the task model Record of logs using a logging tool Preparation part of RemUSINE Generation of the results Analysis of the results provided by RemUSINE Summary on RemUSINE Suitable for graphical applications with a well defined set of tasks to support Useful especially for medium-large applications with dynamic dialogues and many remote users It can be complemented with interviews on the parts found problematic Automatic generation of multimedia reports on the usability problems detected

43 Research agenda Models for context of use Use of task model to design adaptive interfaces supporting variety of devices Patterns in HCI Computer-aided analysis of task models Integrating task models in UML to better support design of interactive applications Conclusions and acknowledgements GUITARE Project MEFISTO Project Book on Model-Based, Springer Verlag