2000 Conference Proceedings

Go to previous article 
Go to next article 
Return to 2000 Table of Contents


Pier Luigi Emiliani
Institute of Research on Electromagnetic Waves "Nello Carrara" (IROE)
National Research Council
Via Panciatichi 64, Firenze, I-50127 ITALY
Email: ple@iroe.fi.cnr.it 
Website: http://www.iroe.fi.cnr.it/index-e.html

Constantine Stephanidis
Institute of Computer Science
Foundation for Research and Technology - Hellas
Science and Technology Park of Crete
Heraklion, Crete, GR-71110 GREECE
Email: cs@ics.forth.gr 
Website: http://www.ics.forth.gr/proj/at-hci/cs.html

The World Wide Web (Web) is today the primary medium for remote electronic communication and information exchange. With the progressive establishment of the Information Society and the gradual advances in related technologies, the role of the Web as a "tool" in everyday social, work, or other activities is expected to increase dramatically. Due to these developments and expected evolution, there has been increased attention lately to the accessibility of Web-based applications and services by different categories of users, including disabled and elderly people.

The increased awareness of Web-related accessibility issues has given rise to several research efforts in the area, including technical approaches to achieving accessibility for different user categories, guidelines for the development and presentation of accessible Web content, etc. At the same time, it is becoming evident that, to avoid information overload and to maximize the user's experience with the vast amounts of information available on the Web, it is necessary to tailor the structure and actual content of information provided to the user, based on their individual (or group) characteristics, preferences, etc. [Stephanidis et al., 1998] Although research efforts in the field offer substantial insight both into the requirements of Web users in general, and those of disabled users in particular, they lack in that they do not address the involved accessibility issues as a whole, but rather as distinct problems. Furthermore, in many cases, the solutions offered are ad-hoc and do not offer themselves to reusability across application domains, disability groups, etc.

It is argued that, in contrast to the above view of Web accessibility, we should focus on the development of methodological and technical frameworks that will allow us to take a more comprehensive and cohesive approach to Web accessibility. This need is even more pressing when one considers the advent of Web-based information systems, which are progressively replacing or complementing many traditional information outlets, such as directories, libraries, newspapers, etc. The goal, then, can be formulated as arriving at solutions that will enable us to design, implement and maintain systems that are inherently accessible by the broadest possible user population (i.e., universal accessibility [Stephanidis & Emiliani, 1999, Stephanidis et al., 1998]), while, at the same time, being open, extensible and reusable across different user-, usage context-, and application domain- characteristics.

This paper will provide a brief overview the ACTS-AC042 AVANTI project (see Acknowledgments), which explored the employment of self-adaptation techniques for the development of Web-based information systems that are manually and automatically tailored to the abilities, skills, requirements and preferences of end users, including disabled people.

The AVANTI project

The AVANTI project (September 1995 - August 1998), aimed to address the interaction requirements of individuals with diverse abilities, skills, requirements and preferences (including disabled and elderly people), using Web-based information systems. In particular, the project set out to investigate the possibilities offered by the utilization of existing and new adaptation techniques in achieving accessibility and high-quality of interaction and information access.

1 System Architecture

The AVANTI project advocated a new approach to the development of Web-based information systems, putting forward a conceptual framework for the construction of systems that support adaptability and adaptivity at both the content and the user interface levels. The AVANTI framework comprises five main components (see also Figure 1):

  1. A collection of multimedia databases, which contain the actual information and are accessed through a common communication interface (Multimedia Database Interface - MDI).
  2. The User Modeling Server (UMS) [Schreck & Nill, 1998; Kobsa & Pohl, 1995], which maintains and updates individual user profiles, as well as user stereotypes.
  3. The Content Model (CM) [Fink, 1997], which retains a meta-description of the information available in the system.
  4. The Hyper-Structure Adaptor (HSA) [Nill, 1998; Fink, Kobsa & Nill, 1997], which adapts the information content, according to user characteristics, preferences and interests.
  5. The User Interface (UI) component [Stephanidis et al., 2000; Stephanidis et al., 1998], which is also capable of adapting itself to the users' abilities, skills and preferences, as well as to the current context of use.
Figure 1: The AVANTI system architecture.

The cooperation between the main architectural components of the AVANTI system is presented in Figure 1. The following short scenario (also depicted in Figure 1) presents the typical "route" of a request for a hypermedia document in the system:

  1. The user requests a hypermedia document. The user interface forwards this request to the content adaptation component (HSA).
  2. The HSA matches the request to an appropriate hypermedia document "template", assembles the (adapted) document taking into account user- and content-related information that is provided by the user modeling component (UMS), and propagates the adapted document to the UI.
  3. The UI interprets the hypermedia document, transparently retrieves multimedia objects from the AVANTI databases via the MDI, and finally presents the requested hypermedia page to the user, employing appropriate (accessible) interaction and presentation facilities.

The above conceptual framework has been applied in the development of three information systems, in the context of AVANTI: the Siena information system [Del Bianco, 1998], offering touristic and mobility information to residents and visitors of the city of Siena (Italy); the Kuusamo information system [Penttila & Suihko, 1998], providing information on traveling and accommodation in Kuusamo (Finland) and its surroundings; and (c) the Rome information system [Ghetti & Bellini, 1998], aimed at providing cultural and administrative information for the city of Rome (Italy).

2 Adaptations at the Information Content and User Interface Levels

Adaptations at the information content level are supported in AVANTI through the Hyperstructure Adaptor (HSA) [Fink, Kobsa & Nill, 1997; Nill, 1998] which dynamically constructs adapted hypermedia documents for each particular user, based on assumptions about the user characteristics and the interaction situation provided by the User Model Server [Schreck & Nill, 1998; Kobsa & Pohl, 1995]. The documents are constructed from static elements, and alternative hypermedia objects. The HSA assembles the adapted documents employing a set of adaptation rules and information acquired from the user model (i.e., assumptions about user-relevant characteristics such as knowledge, interests, preferences), and / or content-related information about multimedia objects from the Content Model and Multimedia Database Interface components.

The user characteristics that trigger appropriate adaptation types at the content level, mainly concern the type of disability, the expertise and the interests of the user, while the resulting adaptations mostly concern [Stephanidis et al., 1998]: (i) alternative presentation using different media (e.g., text vs. graphics, alternative color schemes); (ii) additional functionality (e.g., adaptive "shortcut" links to frequently visited portions of the system, conditional presentation of technical details, and "role-taking" facilities allowing users to identify themselves as having a particular disability, or active interest in one); and, (iii) different structure and different levels of detail in the provided information. It is interesting to note that the knowledge about the user and the interaction session is mostly based on information acquired dynamically during run-time (e.g., navigation monitoring, user selection, explicit user invocation), with the exception of the initial profile of the user, which can be retrieved from the UMS, acquired through a short questionnaire session during the initiation of the interaction, or retrieved from a smart card if one is available.

The design and development of the AVANTI browser's user interface (which acted as the front-end to the AVANTI information systems) have followed the Unified User Interface Design Methodology (U2ID) [Savidis, Akoumianakis & Stephanidis, 2000], developed in the context of the EC TIDE ACCESS TP1001 project. The resulting unified interface is a single artifact, in which adaptability and adaptivity techniques are employed, in order to suit the requirements of three user categories: able bodied, blind and motor impaired.

Adaptations at the user interface are supported through the cooperation of the browser and the User Model Server (see also previous section); the former monitors user interaction and notifies accordingly the latter, which, in turn, draws inferences on the state of the interaction (i.e., detects interaction situations), and successively updates its knowledge. The updated knowledge is used by the user interface to decide upon and carry out adaptations.

The categories of interface adaptation supported by the AVANTI UI include [Stephanidis et al., 1998]: (i) support for different interaction modalities and input / output devices; (ii) automatic adaptation of the presentation of interaction elements; (iii) task-based adaptive assistance; (iv) awareness prompting; (v) limited support for error prevention; (vi) limited support for metaphor-level adaptation. In the context of the user interface, adaptability is mainly targeted at ensuring accessible interaction by appropriately modifying the interactive components so that end-users have (at least partial) access to them, while adaptivity aims at increasing interaction quality by continuously adapting the interface to better fit the needs and preferences of users.

Additional features that have been included in the AVANTI browser, in order to meet the requirements of the target user categories, include adaptive support of multiple interaction metaphors, special I/O devices and extended navigation functionality. Alternative metaphors have been designed for the different usage contexts of the AVANTI system. A desktop application and an information kiosk metaphor have been designed, to be used for stand-alone access to the Web, and as a front-end to the information systems available from public kiosks, respectively. Furthermore, special purpose input / output devices have been integrated into the system to support blind and motor-impaired individuals: binary switches, joysticks, touch screens and touch tablets, speech input and output, and Braille output.

3 Evaluation Results

The AVANTI system has been subject to extensive summative evaluations at three different cities in three different scenarios with a total of 180 subjects [Andreadis et al., 1998]. Users were tourists, business travelers, citizens, travel agency clerks, and users with vision and motor impairments. The common question shared by all experiments was whether the developed system is beneficial for users, and specifically whether adaptability and adaptivity in the information content and user interfaces of hypermedia information systems was useful for end-users.

Users were subject to usability tests to determine the system’s learnability, efficiency of use and memorizability, users’ error-proneness, satisfaction and overall attitude, and the specific contribution of adaptability and adaptivity. Data were collected through observation, interviews, questionnaires and the analysis of log files. The results allow one to conclude that AVANTI’s adaptation features were generally well understood, used, and appreciated (some of them were not self-explanatory though) and that the benefit for information systems probably lies more in user satisfaction, than in efficiency gain or error reduction [Kobsa, 1999].

As far as disabled users are concerned, adaptations were found to be a successful means of attaining accessibility, both at the user interface and the information content levels. Specifically, blind users and users with various degrees of motor-impairments were able to both interact with the system in an effective and efficient manner, and locate information that was of particular interest to them due to their disability. In the case of the Siena information system, in particular, where explicit mobility information was present, users found the system properly designed for them and the information on accessibility very clear and useful for mobility. In general, it can be concluded from the evaluation that disability-oriented adaptations in the three information systems resulted in the users perceiving the overall system (both in terms of interactivity and information provision) as having been specifically designed to meet their particular needs. The latter is perhaps the most important point to note, since it indicates the feasibility of developing information systems on the Web, which are inherently and concurrently accessible by multiple user categories. More information about the AVANTI project's evaluation results can be found in [Andreadis et al., 1998].


The normative perspective of the paper is that Web accessibility should be recognized as a crucial quality requirement in the emerging Information Society. In this context, and in view of the fact that the Web is becoming a pervasive medium that is rapidly penetrating an increasing range of human activities, it is absolutely necessary to ensure that all potential users can attain and sustain equitable and high-quality access to Web applications and services. In this respect, accessibility becomes a universal requirement for the population at large, rather a niche requirement of people with disabilities.

The AVANTI project has demonstrated the feasibility of, and proposed a methodological and technical framework for, developing Web-based information systems that are inherently accessible by multiple categories of end-users, including people with various forms and degrees of disability.


The R&D work reported in this paper has been carried out in the context of the ACTS AC042 AVANTI project "Adaptive and Adaptable Interactions to Multimedia Telecommunications Applications", partially funded by the European Commission (DG XIII). The AVANTI consortium comprised the following partners: ALCATEL Siette (Italy) - Prime contractor; CNR-IROE (Italy); ICS-FORTH (Greece); GMD (Germany); University of Sienna (Italy); MA Systems (UK); MATHEMA (Italy); VTT (Finland); ECG (Italy); University of Linz (Austria); TELECOM ITALIA (Italy); EUROGICIEL (France); TECO (Italy); Studio ADR (Italy).