Towards useful and usable interaction design tools: CanonSketchIntroductionState of the art in UI toolsAnalysis, modeling and design toolsModel-based user interface design toolsInformal toolsCollaborative design toolsA model of the designers ' styles of workDimensions of the modelNotation style dimensionsTool-usage style dimensionsCollaboration-style dimensionsEvaluating tools using the modelRelating UCD projects with the workstyle modelCanonSketch: designing for workstyle transitionsThe Wisdom UML profile for UI designCanonical abstract prototypesLinking Wisdom UML to canonical abstract prototypesExpressing UI patterns in CanonSketchEvaluating CanonSketchStudy designUsability results: Study AUsability results: Study BDesign findingsConclusionsAcknowledgementsReferencesTowards useful and usable interaction design tools: CanonSketchPedro Campos*, Nuno NunesLabUSE – Laboratory for Usage-centered Software Engineering, Department of Mathematics and Engineering, University of Madeira,Campus Universitario da Penteada, 9000-390 Funchal, Madeira, PortugalReceived 21 April 2006; received in revised form 19 January 2007; accepted 29 May 2007Available online 27 June 2007AbstractDespite all the effort dedicated to bringing better User-Centered Design (UCD) tools to market, current studies show that the industryis still dominated by tools that do not support the activities and workstyles of designers. Also, there is a growing need for interactiondesign tools aimed at software engineers, a problem related to bringing usability into the software engineering processes.We propose a new workstyle model that can be effectively used to envision, design and evaluate a new generation of innovative inter-action and software design tools, aimed at integrating usability and software engineering.We illustrate the effectiveness of our model by describing a new tool, called CanonSketch, that was built in order to support UCD interms of the dimensions in our workstyle model. We also describe an evaluation study aimed at contrasting paper prototyping with ourtool as well as the level of workstyle support. 2007 Elsevier B.V. All rights reserved.Keywords: User interface development environments; User-centered design; Graphical user interfaces; Human–computer interaction; OO&HCI1. IntroductionUser-Centered Design (UCD) is a process that fostersan early and continuous focus on users in designing andevaluating a system, in order to obtain products that arebetter suited to the users’ expectations (Gould and Lewis,1985). However, and after almost two decades of researc h,the adoption and usage of UCD tools and techniques stillremains limited to large organizations and a limited num-ber of practitioners who recognize its value. Despite allthe research efforts dedicated to bringing better tools tothe industry, designers have frequently considered thattools do not meet their needs (Myers and Rosson, 1992;Iivari, 199 6; Myers et al., 2000).In this paper, we analyze the problems behind the weakadoption and usage levels of modeling and design tools,and propose a new workstyle model to support a new gen-eration of usable tools. Some of the requirements for suchtools were discussed in a recent workshop about usabilityof model-based tools (Campos and Nunes, 2004). Amongother issues, the participants at the workshop highlightedthe following requirements as paramount to promoteusability in tools: traceability (switching back and forthbetween models, knowing which parts can be affected bychanges), support for partial designs, knowledge manage-ment and smooth progression from abstract to concretemodels and back.It is a generally accepted notion that software engineersneed better interaction design tools: tools that will helpthem create higher quality user interfaces. Besides the factthat software development companies usually do not havethe budget for having dedicated interaction designers, it isalso a fact of life that software engineers – who are notexperts at interaction design – will create UIs.What differentiates our design strategy from others –besides this line of argumentation focused on a softwareengineering perspective – is the fact that our approach isspecifically aimed at designing in order to support work-style transitions.Workstyle modeling (Wu and Graham, 2004) has beenproposed as a technique to record the interaction style of0953-5438/$ - see front matter  2007 Elsevier B.V. All rights reserved.doi:10.1016/j.intcom.2007.05.006*Corresponding author. Tel.: +351 96 2305353; fax: +351 291 705199.E-mail address: [email protected] (P. Campos).www.elsevier.com/locate/intcomInteracting with Computers 19 (2007) 597–613a group of collaborators during any software developmentactivity. In our own research, we observed that the UIaspects of current software developm ent could be relievedif both designers and developers used tools that couldtransparently adapt to an y given workstyle: a workstyle isa description of the most important aspects of the wayusers work in order to achieve their tasks. We have found– through empirical observation in small so ftware develop-ment settings (Nunes, 2003), informal and formal usabilitystudies (Cam pos and Nunes, 2005a) and data obtainedfrom surveys and questionnaires (Campos and Nunes,2007), as well as from current research literature – thatboth designers and software engineers often engage intodifferent workstyles and move between workstyles very fre-quently (what we call a workstyle transition).UCD is an iterative process, in whi ch designers oftenengage in different workstyles as they iterate towards thefinal design. Thus we claim that modeling the workstylescan be particularly useful in UCD.In related research, Constantine and Lockwood (1999)described their ideas of ‘‘galactic dimensions’’ as a meta-phor change towards fully interconnected and synchro-nized visual development tools. Traditional CASE toolswere based on a metaphor referred to as the ‘‘glass drawingboard’’, since they merely represented two-dimensionalpaper models on the glass surfa ce of a monitor. The ‘‘glassgalaxy’’ was then proposed as a multidimensional problem-solving space in which developers could drill down intoobjects in one-dimension, and be taken via software ‘‘wormholes’’ to another. Clicking on a use case could take thedeveloper to its definition in a glossary. Selecting that usecase could also show the abstract components that supportit,