A Theoretical Framework for Specifying and Analyzing Context-Aware Adaptation Vivian Genaro Motti, Nesrine Mezhoudi, Jean Vanderdonckt LILab – Louvain Interaction Laboratory - Université catholique de Louvain Place des Doyens 1 – Louvain-la-Neuve 1348 vivian.motti@uclouvain.be ABSTRACT can be provided. However, due to the fact that the scenarios An effective adaptation of user interfaces approximates in which the interaction takes place significantly vary and technological benefits to the actual needs, wishes and that the current applications still rely on conventional requirements of end users. Today the significant contexts of use, there exists a significant gap between the heterogeneity of contexts of use, given mainly by the real users’ needs and what technology actually offers them. diversity of devices and the easier access to technology, In this sense context-aware adaptation aims at providing enables an interaction from many distinct environments and users, systems that are more suitable to their actual needs covers different application domains and users’ profiles. and also to the specific characteristics of their contexts of However the majority of the applications currently use. Due to the increasing variety of platforms, and easier available still rely in a conventional context of use, i.e. an access to technology, adaptation has been receiving a able-bodied user, with a Desktop PC in stable environment. growing attention since the early 90’s. However, even with Thus several usability issues are often found, requiring a many studies dedicated to this field, still there is no unified deeper investigation of how adaptation can be efficiently framework able to support stakeholders in the specification defined and analyzed regardless of application domains. and analysis of such applications. Moreover, the results This paper proposes a theoretical framework that considers achieved so far do not have a general-purpose that covers dimensions of context and adaptation to support multiple application domains. stakeholders in the specification and analysis of context- aware adaptation. This paper presents a theoretical framework for context- aware adaptation, composed by two modules that support Author Keywords the development, analysis, evaluation and comparison of Context-aware Adaptation, Adaptivity, Adaptability. adaptive and adaptable applications in an integrated and flexible manner. This framework includes: a reference ACM Classification Keywords framework (CARF) and a design space (CADS). To orient D.2.2 [Software Engineering]: User interfaces. H.1.2 the development phases considering the involvement of [Information Systems]: Human factors. H.5.1 developers of different profiles, we abstracted relevant [Information Interfaces and Presentation]: Multimedia concepts of adaptation, and categorized them in specific Information Systems. H5.2 [Information interfaces and dimensions of varied granularity levels, aiming to provide a presentation]: User Interfaces – User-centered design. framework that is intuitive and also easy to use. General Terms This paper is organized as follows: the Section 2 discusses Human Factors; Design. related works; Section 3 defines the requirements and the design decisions; Section 4 presents the theoretical INTRODUCTION framework; Section 5 presents and discusses the results, The universal access in a current landscape of ubiquitous, provides final remarks and future works. pervasive and mobile computing, can only be achieved by means of considering the context information and RELATED WORKS effectively adapting user interfaces. By considering several Although several works have been dedicated to the domain distinct contexts of use and by providing adaptation in an of context-aware adaptation, they focus on a specific efficient manner, higher levels of usability and accessibility dimension of adaptation at a time. Concerning Design Spaces for adaptation, a design space Permission to make digital or hard copies of all or part of this work for for context-awareness has been proposed by Vanderdonckt personal or classroom use is granted without fee provided that copies are et al. (2005). They identified the main challenges for not made or distributed for profit or commercial advantage and that copies context-aware UIs and principles that must guide its bear this notice and the full citation on the first page. To copy otherwise, implementation. They also proposed a design space or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. focusing on model generation and adaptation. Our work inherited a lot from this design space, however we separated concepts that are simply declarative (e.g. possible adaptation process needs to be performed by the system agents who trigger adaptation) in the CARF, from that stores, retrieves, instantiates and applies adaptation comparative concepts (i.e. that enable comparison among rules, techniques, methods and strategies, based on the their granularity levels) in the CADS. previous creation and selection of the developer. The inference process is based on algorithms that fill adaptation Concerning Frameworks for Adaptation, there are several rules according to the context information gathered, and proposals that target at its specific aspects. Dey et al (2001) decide the best methods and strategies. While the developer proposed a framework for facilitating the development of creates, provides and selects such algorithms based on his context-aware application, defining as main requirements: needs and interests, the system provides support to perform the separation of concerns, context acquisition, storage and such operations. To generate adapted final user interfaces, interpretation, resource discovery, and distributed the system must be able to produce them based on: the communication. Our framework, although focusing on results of the inference process and the application of theoretical aspects, not only considers context-awareness, adaptation rules according to the context gathered. Then but also completes it, including also adaptation aspects. users must be able to access and provide their feedback. ROAM is a framework that assists developers in implementing applications able to run in multiple devices, To handle such concepts, requirements from a system and and that enables users to migrate their applications across developer perspective must be fulfilled (see Table 1). devices without much efforts. It follows as adaptation strategies: transformations, dynamic instantiation and Table 1. Requirements for Defining an Adaptation Process offloading computation. Agents support the migration The  System  must  support:   The  Developer  must:   across devices. This framework considers as context information only the device properties, including: display Gathering,   Processing   and   Identify   and   Select   relevant   size, input method and user interface library [Chu04]. Using  Context  Information   context   information,   and   PersonisAD is an architectural framework to model and to application  aspects   use context. Its key concern is scrutability, i.e. the users can Storing   and   Querying   Identify   and   Select   adaptation   access and understand their models by using simple Adaptation  Rules   rules   operations like access, tell, and ask. Their main contribution is a generalised framework to simplify the creation of Executing   Inferences   and   Identify   and   Select   Adaptation   ubiquitous computing applications; they focused on Reasoning   Methods,   Techniques   and   modelling the environment and on the distributed and active Strategies   nature of the models [Ass07]. According to Ardissono et al. Generating   the   Adapted   Analyze   and   evaluate   the   (2008) to enhance the flexibility of the workflow in web Final  User  Interface   adaptation  levels   service composition systems, the context information and the adaptation rules must be explicitly represented in the To meet these requirements, a theoretical framework adaptation logic. They propose CAWE, a framework that supporting context-aware adaptation is proposed, being manages context-aware applications with a hierarchical composed by two theoretical modules: CARF and CADS. representation of the workflow, thus supporting the execution of alternative courses of actions and the context- THEORETICAL FRAMEWORK MODULES aware invocation of web services. It considers the The theoretical modules abstract adaptation concepts in a adaptation of the UI and the workflow execution. way that developers with different profiles are able to develop applications that execute context-aware adaptation. It is important to investigate specific dimensions of context- The Theoretical framework proposed comprises the CARF aware adaptation in depth; but today there is no and the CADS. The CARF is a context-aware reference methodology that supports CAA broadly and flexibly, framework that lists seven adaptation dimensions and their covering both its specification and evaluation. To contribute possible instances. The CADS is a design space for the in this sense, we identified related requirements, proposed, analysis, evaluation and comparison of coverage levels of and implemented a theoretical framework for CAA. adaptation. While the CARF specifies dimensions and their possible instances for implementing adaptation, the CADS REQUIREMENTS FOR ADAPTATION specifies analytical dimensions and their respective Context-aware adaptation involves four main concepts: the coverage levels for performing adaptation. context information, the adaptation process (e.g. techniques and methods), the inference phase (taking optimal The CARF is a graphical representation of relevant decisions, reasoning) and the production of the adapted concepts for performing context-aware adaptation, thus final user interface. These concepts are defined as follows: providing stakeholders an extensive list of possibilities to be considered while creating adaptive or adaptable The context information needs to be gathered, processed applications. The CARF, as Figure 3 illustrates, comprises and used in a dynamic manner by the system, and needs to 7 dimensions defined in clockwise sense as: be selected according to its relevancy by the developer. The • What: the resources subject to adaptation, i.e. the box, a window, or the complete application; navigational flow, the contents (of any type, like audio, • Modality: refers to the modality types involved in the text, or images) or the presentation; adaptation process, i.e., intra when the same modality • Why: regarding the software qualities, as to improve type is considered, inter when the type changes, and the performance or the accessibility level; multi when multiple types are available; • How: which are the techniques, methods and strategies • State Recovery Granularity: refers to the impact in applied to adapt (e.g., improve the contrast level, by the user interaction, i.e. if the user needs to start a new changing the colors of the background and text, with a session, if the user re-starts from the task level, or if the smooth transition); user re-starts just from his or her previous action; • To What: defines the context information that is taken • UI Deployment: defines whether it is dynamically into account to perform adaptation, mainly regarding executed, or statically executed; User, Platform and Environment; • User Feedback: if users can accept or reject the • Who: the agents responsible for triggering an adaptation after (pos) or before (pre) it is performed, or adaptation process, such as the end user, a third-party if the users can evaluate it numerically or literally; or a developer; • Technological Space Coverage: if the technologies • When: if the adaptation occurs at run-time, design- involved in the adaptation are all of the same (intra), time, compilation time; switch types (inter), or if multiple technologies are involved (multi); • Where: adaptation can be performed at the client, server, proxy, or with a mixed approach. • Existence of a Meta-UI: the Meta-UI is an abstract model able to govern the adaptation process; it can be The CARF can be used before the implementation phase, to absent, or a meta-UI without negotiation (i.e., pre- inform stakeholders about possible alternatives for deciding defined), with negotiation (i.e., able to evolve), or a the application requirements, but also after the plastic meta-UI (capable of automatically adapting implementation phase to analyze further possibilities that across several contexts); were not initially considered. • Autonomy Level: designed systems are pre-defined by The CADS graphically represents a context-aware design default and no adaptation is performed, adaptable ones space, highlighting relevant dimensions for adaptation. allow users to intervene, adaptive ones are Orthogonal axes represent dimensions and their respective automatically adapted and mixed-approach ones granularity levels. Adaptive and adaptable applications can combine user and systems’ adaptations. be analyzed and compared by means of CADS. Its benefits include: extensibility (once additional dimensions and The CADS and the CARF are complementary approaches granularity levels can be incorporated), flexibility (once that provide a theoretical methodology to support the dimensions can be included or removed, aiming a more implementation and analysis of adaptive and adaptable focused analysis), exploratory (once it provides a unified view of all dimensions and their coverage levels simultaneously), comparative (once multiple applications can be consistently compared), and descriptive (once all dimensions and their granularity levels are well defined) [Lafon, 2000]. Figure 4 illustrates the CADS applied to the analysis of a given application. The black axes correspond to respective coverage levels regarding each of the CADS dimensions. In clockwise sense, they are defined as follows: • UI Component Granularity: defines the level of granularity that is subject to adaptation, e.g., one edit Figure 1. CARF Figure 2. CADS applications. They enable stakeholders of different to equally accommodate further application domains. In this technological profiles to take decisions about an adaptation sense we conclude that the application and re-use of the process and to analyze and compare multiple applications. proposed framework is feasible, and that it can be in the future applied to effectively support the development of Table 2. Adaptation Techniques detailed: Dyslexia. adaptive and adaptable applications for different domains. We believe that further evaluation efforts are needed to Name   Dyslexia   clearly identify the usage, and potential adaptations, of the Reference   http://www.studiostudio.nl/   framework according to specific stakeholder profiles. As future works we intend to provide online versions of the Description   Given  a  text  content,  its  font  type  is  replaced  by   tools in order to make them publicly available. a  specific  one  for  dyslexic  users   Rationale   The   text   content   is   defined,   its   font   type   is   ACKNOWLEDGMENTS modified   This work received funding from the European Example   The   user   reads   an   e-­‐book,   he   is   dyslexic,   its   e-­‐ Commission’s Seventh Framework Program under grant reader  automatically  changes  the  font  type   agreement number 258030 (FP7-ICT-2009-5). Context   According  to  the  user  impairments   REFERENCES 1. Ardissono, L., Goy, A. and Petrone, G. ‘‘A framework Advantages   Improve  the  accessibility  levels   for the development of distributed, context-aware Disadvantages   It   may   affect   the   performance   (due   to   the   text   adaptive hypermedia applications’’, in AH’08 Proc. of the processing  and  rendering)   5th Int. Conf. on Adaptive Hypermedia and Adaptive Web-Based Systems, Berlin/ Heidelberg: Springer, pp. Picture   259-262, 2008   2.Assad, M., Carmichael, D. J., Kay, J. and Kummerfeld, B.. 2007. PersonisAD: distributed, active, scrutable model FINAL REMARKS framework for context-aware services. In Proc. of the 5th To identify potential interaction issues concerning the int. conf. on Pervasive computing (PERVASIVE'07), theoretical framework, a well-defined evaluation plan and Springer-Verlag, Berlin, Heidelberg, 55-72. evaluation criteria must be established. Our evaluation plan 3.Chu, H., Song, H., Wong, C., Kurakake, S, and Katagiri, intends to obtain not only the end users’ perspective about M. 2004. Roam, a seamless application framework. J. the interaction with the system, but also the experts’ one. Syst. Softw. 69, 3 (January 2004), 209-226. The requirements initially identified aided to guide the DOI=10.1016/S0164-1212(03)00052-9 definition of the application. The theoretical modules http://dx.doi.org/10.1016/S0164-1212(03)00052-9 supported the requirements gathering and analysis of study 4.A. K. Dey, G. D. Abowd, and D. Salber. 2001. A cases, permitting to select the concepts of interest based on conceptual framework and a toolkit for supporting the an extensive landscape of concepts (provided by the rapid prototyping of context-aware applications. Hum.- CARF). The specific results consist in the development of Comput. Interact. 16, 2 (December 2001), 97-166. the sub-modules, namely: a context-aware reference DOI=10.1207/S15327051HCI16234_02 framework, defining and listing dimensions and possible http://dx.doi.org/10.1207/S15327051HCI16234_02 instances for adaptation (CARF), a context-aware design space that identify coverage levels of adaptation, permitting 5.Lafon, M. B. (2000) Instrumental Interaction: An multiple applications to be analyzed and compared Interaction Model for Designing PostWIMP User (CADS). As main results, this work defined requirements to Interfaces. In Proc. of the SIGCHI conference on Human perform context-aware adaptation of user interfaces. factors in computing systems (CHI 2000), ACM Press, Besides it also proposes, implements and validates a New York, NY, 2000, 446-453. theoretical framework to support the specification and 6. J. Vanderdonckt, D. Grolaux, P. Van Roy, Q. Limbourg, analysis of context-aware adaptive and adaptable B. Macq, B. Michel, A design space for context sensitive application. Such a framework includes two main user interfaces, Proc. of 14th Int. Conf. on Intelligent and theoretical modules. The design space enables many Adaptive Systems and Software Engineering IASSEi’05, applications to be analyzed and compared regarding their Toronto, Canada, 20–22 July 2005. coverage levels for adaptation. Given the modularity of the framework and its generic approach, we believe it is sufficiently extensive and flexible