Research Report on User Modeling for Accessibility

Editors' Draft 27 August 2014

This version:
Latest published version:
Latest internal version:
Previous published version:
Previous internal version:
Yehya Mohamad, Fraunhofer FIT, Germany
Christos Kouroupetroglou, CNT, Greece
Yehya Mohamad, Christos Kouroupetroglou eds. Research Report on User Modeling for Accessibility. W3C WAI Research and Development Working Group (RDWG) Notes 2014. (Available at:

A BibTex file is provided; see also information on citing and referencing this document.

Copyright © 2014 W3C® (MIT, ERCIM, Keio, Beihang), All Rights Reserved. W3C liability, trademark and document use rules apply.


User models are abstract representations of user properties including their needs, preferences, knowledge, as well as physical, cognitive, and behavioral characteristics [23]. The characteristics are represented by variables. User models are instantiated by the declaration of these variables for a particular user or group of users. Such instances of user models are called user profiles. A user profile captures the kind of information about an individual user that are essential for an adaptive system to behave differently to different users.

User modeling can be used to enhance the accessibility of user interfaces by generating and adapting them according to the particular user needs and preferences represented as user profiles. Different approaches have been taken in this field but further research and development is necessary, in particular to address the need for standards to support the interoperability and portability of user models across implementations.

More specifically, one of the main challenges of user modeling is the absence of a common approach for integrating user profiles that support different user models within individual implementations, and for migrating profiles from one implementation to another. This can be attributed to the broad variety of user profiles and the incompatibilities that can occur among them. For example, differences in user profiles can occur due to differences in scope of the modeling, source of information for the modeling, time sensitivity and update methods of the model (static vs. dynamic model). There are many other factors that further increase the variety of models. These factors will be discussed in the following papers' descriptions presented at this symposium.

Status of this document

This section describes the status of this document at the time of its publication. Other documents may supersede this document. A list of current W3C publications and the latest revision of this report can be found in the W3C technical reports index at

This [dd month yyyy] [First Public] Draft of Research Report on User Modeling for Accessibility is intended to be published and maintained as a W3C Working Group Note after review and refinement. The report provides an initial consolidated view of the outcomes of the User Modeling for Accessibility Online Symposium held on 15 July 2013.

The Research and Development Working Group (RDWG) invites discussion and feedback on this draft document by research and practitioners interested in User Modeling for Accessibility, in particular by participants of the online symposium. Specifically, RDWG is looking for feedback on:

Please send comments on this Research Report on User Modeling for Accessibility document by [dd month yyyy] to (publicly visible mailing list archive).

Publication as a Working Draft does not imply endorsement by the W3C Membership. This is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to cite this document as other than work in progress.

This document has been produced by the Research and Development Working Group (RDWG, as part of the Web Accessibility Initiative (WAI) International Program Office.

This document was produced by a group operating under the 5 February 2004 W3C Patent Policy. W3C maintains a public list of any patent disclosures made in connection with the deliverables of the group; that page also includes instructions for disclosing a patent. An individual who has actual knowledge of a patent which the individual believes contains Essential Claim(s) must disclose the information in accordance with section 6 of the W3C Patent Policy.

Table of Contents

1 Introduction

This research report aims to present the findings of the Online symposium User Modeling for Accessibility (UM4A) 15 July 2013. It constitutes a basis from which to further explore research and development within the user modeling for accessibility domain. In this case, we first present an overview of the state of the art regarding User Modeling for Accessibility, we then introduce the papers that were presented at the symposium, and finally provide input to future research on User Modeling for Accessibility.

In our everyday life we use a plenty of gadgets, especially electronic devices offering a variety of services. We cannot imagine a single day without a mobile phone, TV or a computer. These devices have huge potential to help people engage with society and surroundings; however, while potentially enabling, the enormous number of features often turns overwhelming for some older people or with certain types of disabilities, and may make devices unusable. At present there is no way of choosing appropriate accessibility options for different users and media, except a case by case analysis, which is not a scalable approach. User Modeling provides a way of choosing an appropriate feature or service based on the user and context of use.

Such a user profile should contain information that:

Thus, interoperability in terms of user modeling refers to the ability of systems to support compliant and interoperable user models that enable the propagation of user information across different systems. A potential solution could be to describe and put in place a standard be able to map between different user models.

2 Related Work

User-adaptive systems [16] are used in different domains and contexts e.g. ergonomy, simulation, e-commerce, e-learning, tourism, cultural heritage, digital libraries, etc. A user-adaptive system adapts its contents, structure and interface according to the user features contained in the user model. The user model typically maintains user properties such as preferences, interests, behavior, knowledge, goals and other facts that are deemed relevant for a user-adaptive application [3][9][17][18]. The user model is a key component of an adaptive system. Indeed, the quality of personalized services provided to the user largely depends on the characteristics of the user model, like its accuracy, the amount of data it stores, whether such data are up to date, etc.

The proliferation of the web and mobile systems that gather personal information about the user has increased recently. The use of such websites and systems leads to the replication of user data, such as preferences, knowledge, activities, tags, etc, over many applications. This fact makes the need for user model interoperability prominent.

The ongoing research in the area of user modeling can be divided in many areas, which will be described in the following sections.

2.1 Simulation use case: Digital human models for the simulation of users

Simulations using user models can be a powerful approach to support engineers and usability experts in the product development process and in ergonomic studies. Virtual human modeling (VHM), (digital human modeling DHM) are used as synonyms to user models. User models reduce the need for the production of real prototypes and can even make it even partially obsolete [10]. During the past years, research interest in using digital human modeling for ergonomics purposes increased significantly [8][11]. Lamkull et al (2009) [12] performed a comparative analysis on digital human modeling simulation results and their outcomes in the real world. The results of the study show that ergonomic digital human modeling tools are useful for providing designs of standing and unconstrained working postures. The use of virtual humans and simulation in the automotive industry showed also great potential. Porter et al (1993) [13] presented a summary of applications of digital human models in vehicle ergonomics during the early years of personal computers.

Existing available tools and frameworks provide designers with the means for creating virtual humans with different capabilities and use them for simulation purposes. DANCE [14], for instance, is an open framework for computer animation research focusing on the development of simulations and dynamic controllers, unlike many other animation systems, which are oriented towards geometric modeling and kinematic animation. SimTk's OpenSim is also a freely available user extensible software system that lets users develop models of musculoskeletal structures and create dynamic simulations of movement. There are also many tools such as JACK from Siemens, RAMSIS , Santos , Human Builder is the virtual user model for CATIA, Enovia and Delmia from Dassault Systems , offering considerable benefits to designers looking to design for all, as they allow the evaluation of a virtual prototype using virtual users with specific abilities.

In the area of accessibility a case study was presented, the HADRIAN system based on the SAMMIE CAD [15], which tried to detect accessibility issues during the interaction between users and ATM machines.

Even though significant effort has been made in physical user modeling, and many tools use virtual humans for simulation purposes, there is no widely accepted formal way for the description of the virtual users, being able to also describe users with special needs and functional limitations, such as the elderly and people with disabilities.

2.2 Adaptation use case: Adapting user interfaces at runtime

There are as well a plethora of systems developed in human computer interaction during the last three decades that are using user models. Many of them modelled users for certain applications - most notably for online recommendation and e-learning systems [5]. There is a bunch of application-independent models which merges psychology and artificial intelligence to model human behavior in detail. In theory they are capable of modeling any behavior of users while interacting with environment or a system. This type of models is termed as cognitive architecture and has also been used to simulate human machine interaction to both explain and predict interaction behavior [6]. A simplified view of these cognitive architectures is known as the GOMS model [7] and still now is most widely used in human computer interaction though it does not consider people with disabilities or non-expert users in detail. Existing user model tools focused on inclusive interaction do not yet cover a wide range of users with perceptual, cognitive and motor disabilities. Examples of such tools like EASE [2] a simulation tool for accessible design, CogTool a general purpose UI prototyping tool or SUPPLE [21] automatic generation of personalizable user interfaces

User models are used for adapting user interfaces at runtime or to produce software with adapted user interfaces.

2.3 Formalization and standardization efforts in the user modeling domain

With the explosion of the Web, and e-commerce in particular, several commercial user modeling tools appeared in the market with the objective of adapting content to users’ features and preferences. Standards and recommendations in this area had to cope with the spread of service-oriented architectures in ubiquitous environments and to cover workflow and user interface aspects e.g. UsiXML (USer Interface eXtensible Markup Language), EMMA (Extensible Multi Modal Annotation markup language) and MARIA XML (MARIA (Model-based lAnguage foR Interactive Applications). All these frameworks contain a user model component but do not cover all user modeling aspects. Another major source for the development of user models was the e-learning sector e.g. IMS AccLIP (Access For All Personal Needs and Preferences Description for Digital Delivery Information Model) and AccMD, which have been internationalized in the ISO/IEC JTC1 “Individualized Adaptability and Accessibility for Learning, Education and specification for the User modeling software Training” (ISO/IEC 24751-1:2008). The Universal Remote Console - URC Standard (ISO/IEC 24752), the goal of URC technology is to allow any device or service to be accessed and manipulated by any controller. Users can then select a user interface that fits their needs and preferences, using input and output modalities as well as interaction mechanisms that meet their individual needs and preferences.

Part 2 of the ISO/IEC 24751 (ISO/IEC 24751 2, 2008): specifies a model dividing the personal needs and preferences of the user into three categories

Furthermore, Part 6 of the ISO/IEC 24751 standard introduces a model of accessibility as a basis for understanding access issues with the interactions between users and systems in various environments. The model shows that users and systems must share capabilities of communicating through a framework to specify a profile of common access capabilities (the so called CAP) of interactive systems, users, and their environment that are necessary for accessibility to be possible.

This latest effort encapsulates rather rich semantic and has been realized as an ontology by Sala et al. [19] in the scope of an Ambient Intelligence (AmI) framework.

ETSI TS 202 746 (ETSI ES 202 746, 2010) specifies user information and preferences that will result in driving the behavior of the system. The standard adopts the concept of a user profile as a reference to a set of preferences, information and rules that are used by a device or service to deliver a customized version of capabilities to the user.

In ETSI TS 202 746, a profile contains:

In addition, the user profile is organized into blocks, such as:

Quite interesting is the ETSI Guide on Human Factors (HF); User Profile Management (ETSI EG 202 321, 2005 10).

According to the Guide, when users wish to have the behavior of devices or services personalized to their requirements a profile will be required. A profile may apply to single simple device or service, more complex devices or services, or any combination that the user may wish to use. The entire set of saved information, preferences, rules and settings a user may want to apply is their "user profile".

In addition, a user may choose to have multiple profiles. Profiles reflect a user's unique lifestyles and situations. Profiles may be related to the various situations that a user experiences in their lives. If users view these situations in a hierarchical way, it should be possible for users to have a hierarchy of profiles that reflect the relationship between these hierarchical situations. For example, a user may have a number of profiles related to different work situations such as "Work meeting" and "In my office" with a general "At work" profile at the top of this hierarchy of work related profiles and that would apply if none of the other more specific situations applied. A profile will contain details of the user and their personal requirements in a form that can be used by the system to deliver the required behaviors.

The EN 1332 4 standard addresses the needs of all users, including people with special needs, for example the aged, minors, the disabled, the visually impaired, those with learning difficulties, first time users, those not conversant with the local language (EN 1332 4, 2012). The standard specifies:

The IMS Learner Information Package Accessibility for LIP Information Model (ACCLIP) provides a means to describe how learners can interact with an online learning environment based on their preferences and needs. The standard is meant to serve the needs and preferences of all users, not only those with a disability. In particular, through its information model, accessibility extends beyond disability to benefit users in learning situations that require alternative modes of use. The user preferences defined aim to aid the user in displaying learning material in the style best suited to their particular needs and in specifying an interface that they can interact with effectively which allows the accessible display and control of the learning material.

Several approaches targeting the formal modeling of user properties have been presented in the scientific literature, for a variety of application domains. For example, Heckmann et al. proposed an architecture for decentralized user modeling having its basis on the user model markup language (UserML) and the general user model ontology (GUMO) [20]. UserML is an Resource Description Framework (RDF) based exchange language for user modeling between decentralized systems, while GUMO is an ontology expressed in OWL Web Ontology Language. The study aimed to address uniform interpretation of decentralized user models, and the integration of ubiquitous applications with a user model service. Sutterer et al. proposed a user profile ontology that is dedicated to describe situation dependent sub profiles, aiming to support context aware adaptive service platforms for mobile communication and information services, so as to automatically trigger the situation dependent personalization of services [21]. The design took into consideration recommendations from the human factors engineering perspective, enabling the specification of situational conditions and situation dependent user subprofiles.

In the European funded cluster VUMS (Virtual User Modeling and Simulation Cluster) cluster a through study of modeling technologies was presented in a public deliverable User Model Interoperability Requirements. . The main aim of that document is to provide an analysis of the existing standards, techniques and methodologies.

3 Symposium Papers

Paper 1: Inclusive User Modelling and Applications Pradipta Biswas & Pat Langdon, University of Cambridge, UK,

The paper “Inclusive User Modelling and Applications” describes user needs and how users would benefit from systems using an inclusive user model. So it presents some examples of users with mild disabilities, e.g. one who has brought a new computer and started to use it. He has mild age-related visual disability and tremor in finger. For such a user it will be easier for her if the font sizes and inter-button spacing can be increased. The paper elaborates on the current situation of such users, who rely on using manuals to set up the new system without that the system knows anything about the user. On the other side the paper describes the situation, where the system is using a user model and by that understands its user and automatically adjusts the settings as soon as he logs in the system. It will be further beneficial if these settings can be seamlessly applied to all other electronic devices like his TV and smart-phone, too. Change of these settings often needs slight tweaking of the design like changing color contrast, increasing font size, changing layouts of buttons and can make them far more usable as well as increase the market coverage of the products.

Paper 2: Validating User Modelling Web Service Pradipta Biswas & Pat Langdon, University of Cambridge, UK,

The paper “Validating User Modelling Web Service” presents a study on making an icon searching interface accessible through a user modeling web service. The paper offers the results to other designers either to improve designs of iconic interfaces for elderly users or developing similar user profile based adaptations for their own applications. The paper describes a set of user modeling web services developed by the authors, that can dynamically adjust font size, cursor size, color contrast, audio volume (for DTV systems) and spacing between interface elements. Detail of the user model can be found in the paper Inclusive User Modelling and Applications. This paper elaborates on the validation of the user modeling web service through a simple point and click task. It reports that the task is kept simple to ensure the statistical effect we observed in the trial is only due to the experimental conditions and not due to difficulty in learning the task. Paper 1 and Paper 2 are a very good example of a working user model with the an operational use case scenario as shown in paper2

Paper 3: Application of abstract user models as customer involvement in product development Markus Modzelewski. University of Bremen, Michael Lawo. University of Bremen, Yehya Mohamad, Fraunhofer FIT, Svetlana Matiouk, Fraunhofer FIT

The paper "Application of abstract user models as customer involvement in product development" reports about the results of a research project in the area of virtual user models and the simulation use case it describes some of the challenges of recent product development as e.g. the inclusion of customer-oriented needs in product design addressing an as wide group of population as possible. It elaborates on existing methods of user involvement that range between „Design for-” and “Design by-” approaches. This paper focuses on the modification of Quality Function Deployment by including recommendations based on customer needs as a framework for designers addressing people with age related mild to moderate disabilities. The authors present the results of the tests of the above mentioned approach using an abstract user model on user interfaces for washing machines and mobile phones as end products. Their system allows designers to specify a typical scenario to obtain suggestions and recommendations that were extracted from different sources as published user studies or knowledge of designers and standards. This paper finally shows some future research directions in the area of simulation of humens based on abstract user models

Paper 4: Personalized web accessibility assessment using virtual user models Nikolaos Kaklanis, Konstantinos Votis, Dimitrios Tzovaras, Information Technologies Institute, Centre for Research and Technology Hellas, Thessaloniki, Greece

The paper "Personalized web accessibility assessment using virtual user models" talks about “Design for All” which according to the paper has become a very popular principle during the last years. The authors postulate that however, the development of products, services and applications that are accessible by all people including any possible kind of disability is a very difficult task. Thus, new “User Sensitive Inclusive Design” methodologies appeared, in order to enforce the development of accessible products with special focus on specific target groups of users. Following the “User Sensitive Inclusive Design” principles, the present paper introduces the WaaT, a software tool that performs personalized accessibility assessment of web applications using virtual user models (VUMs), according to the Web Content Accessibility Guidelines (WCAG) 2.0 as well as the WAI-ARIA guidelines. There is a large number of software tools performing accessibility evaluation of web sites based on the guidelines of popular accessibility standards. Tools, such as the WAVE, HERA, AChecker and Worldspace FireEyes have been developed based mainly on the guidelines of WCAG 1.0, WCAG 2.0 and Section 508. Although there are many existing tools performing accessibility evaluation of web applications, there is a missing point: the personalized accessibility evaluation according to the specific needs/preferences of a user. This is extremely valuable considering that people with disabilities have often special needs varying from person to person, even if they are having the same disability. This paper reports about an approach that focuses on the use of assistive technologies and maps these technologies with accessibility guidelines. It reports about the interaction limitations of the user that are not described sufficiently. The present paper proposes a personalized accessibility assessment approach using VUMs that can efficiently describe user’s needs and preferences.

Paper 5: Profiling users from users’ behavior Silvia Mirri, Catia Prandi, Paola Salomoni. Dipartimento di Scienze dell'Informazione - University of Bologna

The paper "Profiling users from users’ behavior" elaborates on the static approach of user profiling vs. the dynamic approach using machine learning to gather user profile data. The static approach has been frequently exploited in accessibility context, to transcode content according to users’ needs. Usually, users have to explicitly declare their preferences (font size, luminance contrasts, media alternatives) and needs (assistive technologies they use), while repositories can be used to recover devices capabilities. Then transcoding and adaptation are driven by categorizing device capabilities and users’ needs. The authors describe a dynamic approach of user profile gathering through the usage of machine learning concepts to learn users’ preferences, understanding users’ experience and predicting users’ needs. The paper reports about a system which takes into account users’ behavior and automatically adapt Web pages (or just paragraphs). The described system profiles users by considering characteristics they have discarded and the ones they have preferred, modeling users with different needs: aging people, people with low vision, people with dyslexia, people with color blindness, etc. The more the user discards a characteristic the more the system learns to automatically adapt it, by substituting with the preferred one. This approach could be used preferably in the adaptation use case

Paper 6: Increasing the Flexibility of Accessibility Modelling Through the Use of Semantic Relationships Matthew J. Bell, Colin H. C. Machin. Loughborough University,

The paper "Increasing the Flexibility of Accessibility Modelling Through the Use of Semantic Relationships" discusses the appropriateness of accessibility solutions for an individual in a given situation, which may or may not be a dedicated assistive technology (AT). Adaptations can work as micro-ATs, providing personalization that increases the range of users who are able to access content. As they are intended for customization, many adaptations are not labelled as accessibility options. As they form a crucial role in ensuring accessibility, adaptations and device-specific settings are increasingly being included in profiles describing accessibility needs. The variety of interaction paradigms results in the need for specificity in technology profiles to capture the nuances of each particular device, interface or control. As users are frequently interacting with multiple technologies there is also a need for a user profile that is generic enough to be transportable, whilst specific enough to respond to technology nuances. This paper presents then a possible solution to this need. It proposes the use of Semantic Web to model adaptations (and other accessibility solutions) in terms of human capabilities. This paper focuses on providing a vocabulary that moves from device-specific to device-agnostic profiling through the logical structuring of profiles. The approach improves flexibility, during both profile acquisition/maintenance and the matching process. Dynamic comparison of profiles at varying (appropriate) levels of granularity, allows discovery and remedy of accessibility issues to be performed efficiently.

Paper 7: Personas can tell the story behind the model Whitney Quesenbery, WQusability and Usability in Civic Life

The paper "Personas can tell the story behind the model" elaborates on the approaches of user modeling. It describes the ways of representing users as a technical model, characterizing them through a set of properties and variables that can be collected in a ‘user profile’ and used by websites and applications to provide meet user needs and preferences through personalization in a consistent way (W3C, 2013) as described by this symposium. This paper describe the results of other research in the area of Health Level Seven International (HL7) standard in which the addition of narrative scenarios helped nurses be more effective in shaping the standard to the context in which it would be ultimately used. This paper describes the advantages of using personas as a technique, which is already an established instrument in user experience (UX) research and design used to present a fictional portrait of a type of user, based on research. This paper goes on and presents persons taht offer a way to communicate both quantitative and qualitative data behind the model, giving everyone from users to product teams a better understanding of the goal of the models. The more complex the situations in which the model will be used, the more important it is that everyone have a good understanding of what the models represent. This is critical for today (and tomorrow’s) Web, in which a single activity often spans time, devices, and people.

Paper 8: Developing a Semantic User Preferences and Device Modeling Framework that supports Adaptability of Web Applications for People with Special Needs Philip Ackermann, Carlos A Velasco, Evangelos Vlachogiannis, Fraunhofer Institute for Applied Information Technology FIT,

The paper "Developing a Semantic User Preferences and Device Modeling Framework that supports Adaptability of Web Applications for People with Special Needs" presents research results of a modeling framework that supports dynamic adaptations of the user interface of web applications. This work builds upon previous efforts of the authors leveraged with the use of the semantic framework Composite Capability/Preference Profiles (CC/PP, (CC/PP, 2007)[1][4], which allows the matching of device capabilities and user preferences. The combination of these models with those of the corresponding web applications, enables an adaptive transformation process that facilitates access to users with special needs, derived of their functional restrictions or because of context-related handicapping situations. The authors report about their approach that differs from traditional user models based upon cognitive or physiological characteristics of the users.

4 Discussion of state of the art challenges

User modeling as a technology shows a lot of potential for providing accessibility solutions either in design of products and services or in real-time adaptation of services and interfaces so that they are accessible. The state of the art analysis, the papers and the discussions on the symposium indicate a number of emerging themes, where more effort might be placed into research and development in order to push forward in that direction.

4.1 User needs research

A large proportion of the presented research is targeted towards modeling persons with physical and sensory disabilities such as hearing, motor and visual impairments. Some of the research efforts are also covering partially cognitive disabilities. Concerning the accessibility of user interfaces, user models need to cover all aspects of disability. Some user models for example, consider visual and motor disabilities and do not involve other disabilities such as hearing or cognitive. The extension of user models to become complete models covering all disabilities implies a lot of work in user studies and statistical analysis. In order for user modeling to be used for a specific target groups this needs to be thoroughly researched in terms of their needs and characteristics in order to be able to model as much information as possible and as accurately too. In terms of physical and sensory disabilities research seems to be in a good progress since needs and user characteristics are easier to identify and model. However, user needs research concerning cognitive disabilities is harder due to the complexity of the topic..

Therefore, an emerging challenge identified in the symposium is the need for more intensified research efforts in discovering, documenting and modeling user needs for persons with cognitive and sensory disabilities. This also includes persons with Mental Health Disorders such as autism, ADHD, etc. This gap should be addressed and standardized soon in order get mutual benefit between various research groups, developer and end users.

4.2 Privacy concerns

The majority of recent models are based on totally different technologies posing issues with regard to interoperability and to their utilization in different environments and contexts. Due to these interoperability issues privacy concerns tend to be disregarded. Given however that some time in the future such issues could be tackled, the increase in interoperability will also increase the need for privacy policies, as data is being exchnaged between different systems and may allow access to a larger group of users. Privacy as a crucial aspect is not well considered in the majority of existing systems. There is the opinion that privacy issues do not need to be considered, if the system uses stereotypes and not specific preferences of the user. The utilization of such stereotypes or general data can be used as implicit hints about the specific user preferences or abilities/disabilities. Therefore, privacy in terms of user modeling is one of the very important issues to be tackled especially to help adoption of such systems from wider audiences. Privacy issues include the following aspects:

4.3 Type of modeling

One of the biggest, most interesting discussions and emerging themes in the area of user modeling is the type of modeling to be followed in the area. In general there are two schools of thought.

The first one is basing user modeling on user characteristics. This way a system knowing about the characteristics of a person and other contextual information such as the device used and other environmental variables can reason what kind of adaptations are needed in order to make the product accessible to the user.

The second approach is basing its modeling approach on user preferences. In this type of systems instead of user characteristics the system keeps information about the user preferences. These preferences could in some cases be chosen by the user himself over a short training/question answering session or could be recorded on the background by keeping track of users’ actions and reactions on various adaptations. When a new user starts using the system is mapped with similar users in terms of choices and preferences and the system is then able to suggest further adaptations.

Both approaches have strengths and weaknesses. The first user characteristics based approach is more flexible in terms of taking contextual parameters under account. However, if user characteristics are not described accurately for a person might mean adaptations that will not work for the users. Therefore, such solutions are quite common in systems aiming for developers and product designers that can describe with such systems their potential users and experiment with techniques such as virtual user models in order to find out accessibility problems on their designs.

On the other hand the second preferences based approach is easier for user to train and can provide easier and more accurate adaptations of interfaces in real-time. However, such systems are usually more difficult to be ported on various devices and software making them quite specific.

The aforementioned observations lead researchers currently to believe that there is a need for more research on how to bridge these two approaches under schemes that could benefit from each one's strengths and eliminate their weaknesses. Knowing, which groups of people have specific preferences and under what circumstances could lead to systems being able to reason based on both user characteristics and preferences. In addition, systems aiming at designers would be better backed up on their suggestions by real user’s feedback and preferences being taken under account.

4.4 The relation between user models and personas

User models are a technology used quite often to help designers and developers to produce accessible and inclusive products and services. A number of research programs presented in the symposium showcase such examples. However, designers and developers are also equipped with a variety of other tools on their research toolbox to apply wherever is needed.

One of the tools being used in design and user experience studies is personas. Personas are a very interesting tool in terms of user modeling because it aims to do a similar job. Personas are aiming to model specific target groups so that designers can better understand users and design for them. Personas incorporate a number of user characteristics under an umbrella easy to be presented on designers and users and provide an interface for communication between them.

User models on the other hand do a quite similar job by incorporating specific characteristics in a way that it is easier for machines to understand them, reason on them and provide appropriate solutions/adaptations. Therefore, a persona can be seen as a way of communicating to real users what a user model is communicating to machines. Future research in the area could exploit even more such tools in order to help the process of validation of user models and provide more accurate data for them

4.5 Standards, ontologies and interoperability

One of the biggest challenges faced today in the user modeling domain is the fragmentation of user model definitions. Projects employing user modeling techniques for accessibility often produce new definitions of user models depending on their needs, user group focus and targets. This makes quite difficult for models to be used across different projects and research efforts and hinders their progress. Differences in research aims and user group focus justifies to an extent why a new project employing user modeling to provide an accessibility solution would start building its own models. However, now that user modeling appears as a technology used on an increasing number of research and development projects it is crucial to start working on building standards for user modeling in that domain.

Some of the papers of this symposium presented the usage of the declaritive approach based on XML for the description of the user models e.g. paper 4. Other papers presented the usage of semantic web approach utilising so called ontologies for the description of user models e.g. paper 3 and paper 8. The usage of both approaches creates the need for developing of converters to make the exchange of data between user models of both approaches a necessity. Sergey Sosnovsky et al. has published a paper on ontological technologies for user modeling [22]. This paper brings together research from two different fields – user modeling and web ontologies – in attempt to demonstrate how recent semantic trends in web development can be combined with the modern technologies of user modeling. Over the last several years, a number of user-adaptive systems have been exploiting ontologies for the purposes of semantics representation, automatic knowledge acquisition, domain and user model visualisation and creation of interoperable and reusable architectural solutions.

The discussions in the symposium revealed that many researchers are actually supporting the idea of having standards for describing users based on models such as the medical model. On the other hand there seems to be an agreement that models describing user preferences are also quite useful and need to be worked on towards standardization.

Given the discussions about the user characteristics based (medical) models and the user preferences based models and the emerging need for both of them, it is obvious that both approaches should also be supported by standardization actions. This way the emerging need for research on the connections between them will be easier to fulfill. The medical model has some limitation eg. addressing cross-disabilities and rare impairments etc., and how this approach is not widely accepted in the accessibility community, the social model is much more preferred nowadays.

4.6 The importance of contextual information

User interface adaptation can be addressed at design time, run time or both. Utilizing user model technology to support designers to create inclusive user interfaces means incorporating it throughout the entire design process. The designer should be aware of the user requirements already at the prototyping phase, when she designs the interface either as a paper prototype or as a software one. In later stages there should be tools e.g. in a CAD environment to support the designer in creating inclusive designs. In this area we may find two different types of user interfaces, the pure software interface and / or the hardware interface e.g. of a washing machine with a hardware interface or a mobile phone with a pure software interface. Hardware interfaces with physical handles can generally be adapted to user profile only at design time and are fixed for the entire life time of the product.

This type of physical interfaces however, is being with time replaced by software display systems, where the interface are mor flexible and could be adpated to the user profile during the life time of the product. Therefore there is an increas in systems using user modeling technologies to know which user or user type is using the system and adapt their interface accordingly at runtime. This increase of adaptive user interfaces leads to an increased need for more contextual infromation

As discussed and pointed out throughout the symposium and the papers presented context of usage for an interface is a very crucial parameter that user models cannot provide themselves. Adaptation based solely on user models cannot provide actual benefits to users if there is no connection to a specific context. A user model describing users and their preferences for a TV on the person’s living room could be totally different when the same person is using a tablet device in a work environment. Such contextual parameters can include devices (tablet, mobile phone, TV, PC etc.), environmental conditions (lightning, noise etc.), intent of use / task (work, leisure, education etc.) or even social and cultural background (being on a meeting at work, being alone in my living room, being on my car driving to work or on a holiday trip etc).

Context plays a significant role in reasoning what is the best adaptation to choose for a user under specific circumstances. One of the biggest issues already presented is the connection between user characteristics and preferences. Contextual information makes this connection even more difficult to build. Therefore, although user modeling might focus on user characteristics and user preferences it is necessary to connect with contextual information in order to provide the best solutions for the user.

This means that apart from the standardization and integration efforts that need to be done on user models research on the domain should also be connected with standards and models related with sensor and sensor network technologies. This way the link between user characteristics, preferences and context of usage will become even easier to build. Therefore, in order to achieve the goal of developing inclusive user interfaces in the majority of cases there is the need for further models rather than the user model e.g. application model, environment model and task model. The mapping between the user characteristics and / or preferences and components of these models may require special knowledge rules of the domain and specialized algorithms.

4.7 Dynamic user models and methods for learning / updating user models

User models are are usually not static nature but subject to continual updating e.g. older people ofen observe continuous decline of capabilities over time. And as long as they use a device, they are going to need to update their profile as capabilities will change, not only on a day-to-day basis, but potentially more often than that, given the equipment that they are using at specific times, moving from one device to another. This continuous update require the usage of a management system to cope with this aspect.

The aspect of continuous update should be under the control of the user, as it may mean that the user interfaces changes as well continuously, which may confuse the user and jarpodize the acceptance of the system. An example of such a system was the Microsoft office assistant, which has been removed in later versions, because of complains of users.

In case users' data are not gathered before hand or users are not asked directly for their personal data and preferences, but this information is derived from their behavior while interacting with the system. The ways they choose to accomplish a tasks, the combination of things they takes interest in, these observations allow inferences about a specific user. The application dynamically learns from observing these interactions.

To cope with multifcetted requirements of different users on dynamic user models and the danger of drawbacks in acceptance of such systems the technology of machine learning algorithms may provide appropriate solutions.

Paper 5 elaborted on the differences between static and dynamic models and benefits and methodologies of dynamic user model approach

5 Conclusions and future directions

Already existing literature on user modeling for accessibility shows that user modeling is a technology that could benefit people with disabilities in their daily usage of devices and software. However, user modeling for accessibility purposes is still in its initial steps and needs to answer to a number of issues that lie ahead. The symposium papers and the discussions that followed can be summarized into four main areas that future research and development need to pay attention to. Each of the areas is also accompanied with a set of related future directions and challenges.

5.1 Standardization and adoption of standards

One of the biggest problems in all modeling technologies is the language (vocabulary) and the technologies used for describing those models. Although most of the solutions today are based on XML or Semantic Web technologies there is still a lot of discussion of what kind of information should a user model keep, how this will be organized, what are the relationships formed within this information etc. Up to now it seems as though each research team and project follows a separate way to model information needed. This is justified to an extent by the variety of problems, users and needs that each solution tries to tackle. However, this fragmentation hinders the faster evolution of the technology that could come as a result of the cooperation of research teams and developed products. To achieve a holistic approach to user modeling for accessibility there are already a number of standardization actions in place. What is missing though is the critical mass of researchers, projects and products supporting, adopting and improving those standards. Therefore one of the main areas that future research and developments should look into is the adoption and improvement of existing standards and the formation of new ones wherever this is needed.

One of the areas that possibly needs more efforts in the direction of standardization is also privacy issues and mechanisms and rules for exchanging information between sources, users’ control over their information privacy and similar issues. Tackling such issues will also provide a stable and trustable ecosystem in which user modeling based solutions will be developed ensuring easier and quicker adoption from users.

Therefore, future directions and challenges in that area include the following:

Support and extension of current user model standards such as ISO

As already pointed out the fragmentation of solutions and models in the area for user modeling for accessibility hinders evolution of research and development because it is making more difficult for research and development teams to cooperate and exchange information, knowledge and build on each others work. Although there are already actions for standardization of user models there is not yet enough support and a critical mass of researchers and practitioners behind them to support them. Therefore, there is a need for encouragement and increasing the support on such standards in order to increase adoption and further improve and extend them.

Provision for privacy issues

Given the discussions on privacy there is a need for research and development of appropriate mechanisms to address privacy concerns for the following issues:

5.2 Contextual information

The developments in sensor technologies and in the area of the Internet of Things (see W3C Workshop on the Web of Things) provide user modeling with an opportunity to know and evaluate the context under which a user is using a device/software/service. Such information is invaluable for adapting the interface to accommodate user needs. However, models for keeping and using such information are still in initial steps of evolution and face similar problems with user models. Therefore, user modeling technology for accessibility should keep an eye on the developments in the area and possibly play a significant role in forming such models for contextual information. Given that, the future research direction and challenge in that direction is:

Research and standardization on contextual information models

Although standards on user models are being developed, there is an obvious need to connect such models with preferences and needs of users. However, such a connection is difficult and impractical to happen without knowing contextual information of usage of a device/software/service. Therefore, it is important that more effort is needed on designing models for such contextual information, especially nowadays that sensor packed devices (mobile phones, wearables, etc.) can provide related data in abundance.

5.3 User needs studies

Symposium discussions also lead to the conclusion that there is a lack of research covering specific user group needs where there is a significant amount of research related to other user groups. This imbalance in covering and discovering user needs of specific user groups provides an opportunity for future research to focus more on them. Such groups usually include people with cognitive disabilities, behavioral disorders and age related disabilities. In this process of user needs discovering user modeling could greatly benefit from techniques used in user experience studies and especially from personas that can provide a friendly interface to describe and get feedback about user models. Moreover this process would also benefit from a formal definition of the different levels of user needs according to necessity (eg. what is considered to be a need, how necessary is a requirement and how necessary is a preference). Clarifying such issues will provide a common understanding and a basic foundation for future research in the area.

Future directions and challenges related to that aspect include the following:

Improvement and extension of user models for more user groups

It is evident from research so far that there is a lack of research for user needs and requirements for specific groups of people with disabilities. Therefore, there is an obvious need for more research to identify needs and requirements for users (especially with cognitive and age related disabilities) to improve and extend user models for disabilities to cover a bigger spectrum of disabilities

Defining levels of needs

Discussions in the area of user modeling for accessibility often refer to user needs, requirements and preferences. However there is still no formal definition of how important and necessary is each level of need. In terms of user interface adaptation, such a definition could help determining the importance and necessity for each possible adaptation and their prioritization.

Exploiting user experience techniques such as personas

In order for the connection between user models and preferences to be more successful, a big number of user-centered research is needed in order to discover connections beyond the obvious ones. This research can benefit a lot from the techniques used in user experience studies and especially from personas that can provide a link between user models and actual users.

5.4 Connecting characteristics to preferences

One of the most important and fascinating areas of research related to user modeling for accessibility is the connection between user models describing user characteristics and user models describing preferences. Currently, most of the research projects follow one or another way to model their users and provide the appropriate adaptations. However, the development of standards for user models and the development of technologies providing with more and better structured contextual information could help in connecting those two approaches. Reasoning mechanisms provided by Semantic Web technologies and machine learning technologies can greatly help in forming and discovering such connections between user characteristics and preferences. Moreover, cloud computing and big data technologies are also expected to boost the discovering of such connections beyond the obvious ones that user studies can bring forward.

So, future research directions and challenges in the area include the following:

Connecting user models with user needs and preferences

Having modeled contextual information and being able to connect a user model with a specific context, research on discovering connections between user models/characteristics and user preferences can move forward easier and more successful in providing help to people with disabilities

Research on dynamic user models using machine learning technology

A very useful technology for this process of connections discovery between user models and preferences could be machine learning technology and algorithms. Machine learning technology combined with user models and the abundance of contextual and usage information that today’s devices and systems can provide can help in discovering new connections between users and preferences and also make the training process of such adaptation systems quicker and easier for the end-user.

In conclusion, putting more effort in the aforementioned directions is what user modeling for accessibility needs in order to provide useful solutions for accessibility either by helping the design of more accessible products and services or by providing adaptive interfaces for devices, software and services to better accommodate all users’ needs.

6 References

[1] G. Klyne, F. Reynolds, C. Woodrow, H. Ohto and M. H. Butler (eds.) (2002). Composite Capability/Preference Profiles (CC/PP): Structure and Vocabularies, W3C Working Draft 08 November 2002. World Wide Web Consortium. Available at: , Accessed on 12th December, 2012

[2] Fait, H., Mankoff, J., EASE: A Simulation Tool for Accessible Design. Technical Report UCB-CS-03-1260, Computer Science Division, UC Berkeley, July, 2003.

[3] Kobsa, A., Koenemann, J., & Pohl, W. (2001). Personalized hypermedia presentation techniques for improving online customer relationships. The Knowledge Engineering Review, Vol. 16:2, S. 111–155. Cambridge University Press.

[4] G. Klyne, F. Reynolds, C. Woodrow, H. Ohto and M. H. Butler (eds.) (2002). Composite Capability/Preference Profiles (CC/PP): Structure and Vocabularies, W3C Working Draft 08 November 2002. World Wide Web Consortium. Available at:, Accessed on 12th December, 2012

[5] Peissner, M., Schuller, A., & Spath, D. (2011). A Design Patterns Approach to Adaptive User Interfaces for Users with Special Needs. . In: J. Jacko (Ed.) Human-Computer Interaction. Design and Development Approaches, LNCS 6761, Berlin: Springer-Verlag. S. 268–277.

[6] Biswas P., Langdon P. & Robinson P. (2012) Designing inclusive interfaces through user modelling and simulation, International Journal of Human Computer Interaction, Taylor & Francis, Vol 28, Issue 1 DOI:10.1080/10447318.2011.565718

[7] John B. E. and Kieras D. "The GOMS Family of User Interface Analysis Techniques: Comparison And Contrast." ACM Transactions on Computer Human Interaction 3 (1996): 320-351.

[8] Pierre T. Kirisci, Patrick Klein, Markus Modzelewski, Michael Lawo, Yehya Mohamad, Thomas Fiddian, Chris Bowden, Antoinette Fennell, Joshue O. Connor: Supporting Inclusive Design of User Interfaces with a Virtual User Model. HCI (6) 2011: PP 69-78. The four-volume set LNCS 6765-6768

[9] Carmagnola, F., Cena, F., und Gena, C. (2007b). User Modeling in the Social Web. In: Apolloni, B., Howlett, R. J., und Jain, L. C. (Hrsg.), Knowledge-Based Intelligent Information and Engineering Systems, Band 4694 in Lecture Notes in Computer Science, S. 745–752, Berlin, Heidelberg. Springer.

[10] Cappelli, T.M. & Duffy, V.G. (2006). Motion Capture for Job Risk Classifications Incorporating Dynamic Aspects of Work. Digital Human Modeling for Design and Engineering Conference, Lyon, 4-6 July 2006. Warrendale: SAE International.

[11] Laitila, L. (2005). Datormanikinprogram om verktyg vid arbetsplatsutformning – En kritisk studie av programanvändning. Thesis. Luleå Technical University, Luleå.

[12] Lamkull, D., Hanson, L., Ortengren, R. (2009). A comparative study of digital human modelling simulation results and their outcomes in reality: A case study within manual assembly of automobiles. International Journal of Industrial Ergonomics 39 (2009) 428-441.

[13] Porter, J., Case, K., Freer, M.T., Bonney, M.C. (1993). Automotive Ergonomics, Chapter Computer-aided ergonomics design of automobiles. London: Taylor and Francis.

[14] Shapiro, A., Faloutsos, P., Ng-Thow-Hing V. (2005). Dynamic animation and control environment, In Proceedings of Graphics Interface 2005, pp. 61-70.

[15] Russell Marshall, Keith Case, Steve Summerskill, Ruth Sims, Diane Gyi, and Peter Davis 2009: Virtual Task Simulation for Inclusive Design. V.G. Duffy (Ed.): Digital Human Modelling, HCII 2009, LNCS 5620, pp. 642–652, 2009.

[16] Brusilovsky, P., Millán, E.: User models for adaptive hypermedia and adaptive educational systems. In: Brusilovsky, P., Kobsa, A., Nejdl, W. (eds.) The Adaptive Web, Methods and Strategies of Web Personalization. Lecture Notes in Computer Science, vol. 4321, pp. 3–53, ISBN 978-3-540-72078-2. Springer, New York (2007)

[17] Brusilovsky, P.: Methods and techniques of adaptive hypermedia. User Model. User-Adap. Inter. 60(2–3), 87–129 (1996)

[18] Kobsa, A.: Generic user modeling systems. User Model. User-Adap. Inter. 11, 49–63 (2001)

[19] Sala, P., Fernández, C., Mocholí, J., Presencia, P., & Naranjo, J. (2011). Implementation of the ISO/IEC 24756 for the Interaction Modeling of an AAL Space. En C. Stephanidis (Ed.), Universal Access in HCI, Part III, HCI 2011, LNCS (Vol. 6767, págs. 210‐219).

[20] Heckmann, D., Schwartz, T., Brandherm, B., & Kröner, A. (2005). Decentralized User Modeling with UserML and GUMO. In P. Dolog, & J. Vassileva (Ed.), Proc. of Decentralized, Agent Based and Social Approaches to User Modeling Workshop (DASUM), (pp. 61‐65). Edinburgh, Scotland, United Kingdom

[21] Sutterer, M., Droegehorn, O., & David, K. (2008). UPOS: User Profile Ontology with Situation‐ Dependent Preferences Support. Proc. of the 1st IEEE Int. Conf. on Advances in Computer‐Human Interaction, (pp. 230‐235). Sainte Luce, Martinique.

[22] Sosnovsky, S., Dicheva, D.: Ontological technologies for user modelling. International Journal of Metadata, Semantics and Ontologies 5(1), 32–71 (2010)

[23] Benyon, D. R. and Murray, D. M. Applying user modelling to human-computer interaction design AI Review (6) pp 43 - 69, 1993 Available at: PDF

[24] Kay J. and Kummerfeld B.. Scrutability, user control and privacy for distributed personalization. In A. Kobsa, R. Chellappa, and S. Spiekermann, editors, Online Proceedings of PEP 2006, CHI 2006 Workshop on Privacy-Enhanced Personalization, 2006.

7 Symposium Proceedings

Research Report on User Modeling for Accessibility

This document should be cited as follows:

... (2013)
Available at:

The latest version of this document is available at:

A permanent link to this version of the document is:{{date}}/

A BibTex file is provided containing:


Contributed Extended Abstract Papers

The links provided in this section, including those in the BibTex files, are permanent; see also the W3C URI Persistence Policy.

title = {W3C WAI Symposium on User Modeling for Accessibility},
year = {2012},
editor = {W3C WAI Research and Development Working Group (RDWG)},
series = {W3C WAI Research and Development Working Group (RDWG) Symposia},
publisher = {W3C Web Accessibility Initiative (WAI)},
url = {},

8 Acknowledgements

Participants of the W3C WAI Research and Development Working Group (RDWG) involved in the development of this document include: [Alphabetical List of Contributors]

RDWG would also like to thank the chairs and scientific committee members as well as the paper authors of the RDWG online symposium on User Modeling for Accessibility.

9 Appendix

9.1 Question and Answers

the symposium was divided in many topics and the questions were grouped around these topics.

The transcript of the symposium is published online

the groups were as follows: