Liddy Nevile, Associate Professor, Department of Computer Science and Computer Engineering, La Trobe University, Melbourne, 3086. Email: liddy@sunriseresearch.org
In this paper, we describe a strategy for increasing accessibility by exploiting available technologies to match digital resources to users' individual accessibility needs and preferences. This is achieved just in time for the delivery of resources to users by working with descriptions of users' accessibility needs and preferences and relating them to descriptions of resource accessibility characteristics. This strategy supports cumulative and distributed authoring of accessible components for resources where these are missing, and the reconfiguration of resources with appropriate components for users.
Assembling Web resources in an integrated way for delivery to the user is defined as just-in-time accessibility and can increase the availability of accessible resources. Moreover, compared to resources that are accessible to every potential user, universally accessible resources, these resources are less expensive, easier to develop (in terms of skills required), and developed using more satisfactory practices for authors and publishers. In addition, the provision of accessible content can be improved so significantly by the use of specifications-compliant accessibility tools, adopted by moderately competent computer users with no accessibility training, that it is cheaper and more effective to rely on the technology than yet to be developed high-levels of human expertise.
The new approach involves a shift of responsibility from individual authors to technology and a supporting community. The shift means increasing responsibility in the final provision of resources, and thus, of server software. The servers need to check the resources and possibly arrange for services to manipulate and reassemble them before delivering them. The accessible components need to be suitably described to enable their discovery. The components that constitute the final resources may be distributed. This means there is a need for metadata standards that promote interoperability. Finally, there is a need for descriptions not only of resources but also of user needs and preferences.
The term "accessibility" is easily confused with "access", a term used to describe possession of facilities for accessing the Web or rights to use resources. This kind of access is, of course, crucial to any user who is dependent upon the Web. In this paper, however, we are only concerned about people who, for whatever reason, cannot access Web resources, including services, when they are in possession of facilities that should be adequate.
When the Web first became popular, there were communities of people with disabilities who had been using computers for some time. The technology of the time allowed for text activities online and these presented few problems for adaptive technologies. People with hearing disabilities often used teletype machines and other print technologies that allowed them to communicate using what were otherwise typically sound, or image and sound, technologies, such as telephones, televisions, etc. People with sight disabilities used computers to enlarge script, to have it read aloud to them, and to produce Braille. A significant aspect of the Web that made it attractive was its ability to include images, sounds, and multi-media in general.
Unfortunately, the very technology that has opened the door to unprecedented access also harbors the possibility for the very opposite. Just as there are enabling and disabling conditions in the physical environment, so are there conditions associated with digital technology that result in the inclusion or exclusion of certain people. Technology that is not universally designed, without consideration for the full spectrum of human (dis)abilities, is likely to contain access barriers for people with print disabilities. Schmetzke, A. (2001) [HREF1]
Accessibility is defined in this paper as the matching of delivery of information and services with users’ individual needs and preferences in terms of intellectual and sensory engagement with resources containing that information or service, and their control of it. Accessibility is satisfied when there is a match regardless of culture, language or disabilities (Ford, 2005). For the users, matching their needs is of primary importance and in some cases critical to their ability to function. It should be noted, however, that this does not mean some users only want resources that are dull or boring but simply that resources should be adjusted and adapted to suit the stated needs of users so everyone can have what will be best for them.
With respect to content, accessibility can be dependent upon a number of factors. A user needs to be able to use the range of sensory modalities upon which access to the content depends. Where a person at a particular time does not have the necessary sensory capability, alternative content in another modality may be required. In some cases, the modality of content can be altered and this is particularly true of text.
As the European Commission have noted,
According to the WHO International Classification of Functioning, Disability and Health (ICF- May 2001) disability is not entirely an attribute of an individual, but rather a complex social and environmental construct largely imposed by societal attitudes and the limitations of the human-made environment. Consequently, any process of amelioration and inclusion requires social action, and it is the collective responsibility of society at large to make the environmental and attitudinal changes necessary for their full participation in all areas of life. [HREF2]”
Our growing dependence on electronic resources is increasing the inaccessibility of public information, and it is this human construction that needs to be controlled in order to support those with digital disabilities. Just how many people are affected in this way is not known, but recent research in the US suggests that as many as 62% of working adults between the ages of 18 and 64 are affected by digital disabilities [HREF3].
It is also important to remember that many people with disabilities do not want to be treated as such: they want to be able to participate in the world equally with others so they want the same resources as other people get. They strongly rejected the discriminating practice of providing a text-only version for them in the 1990’s and preferred the inclusive idea of universally accessible resources - a one size fits all solution that includes them. These people can now be included by flexible, accessible information delivery systems.
The usual model for accessibility is of resources that are self-contained collections of all the components that may be required by any user. ‘Universal accessibility’ is established by testing conformance of the resource to the best guidelines for accessible authoring, currently those produced by the World Wide Web Consortium [HREF4]. The Web Content Accessibility Guidelines (WCAG) [HREF5] are supported by their authors with documentation that provides techniques for compliance, and also by the many online resources provided by users of the WCAG. The WCAG themselves do not specify how the goals are to be implemented as they aim to provide normative guidelines that can survive the development of technology. Indeed, they are often incorporated into other technologies by those associated with their recommending body, the World Wide Web Consortium, and others.
The WCAG can be best thought of as a set of requirements for resource developers or publishers. They aim to identify the most important barriers for the greatest number of people with special needs and to provide guidelines for their elimination. The problem is that sometimes there is a conflict between such a requirement and one for a less significant number of people, and sometimes it is a matter of judgment what will be best as a general requirement.
Given the absence of implementation rules, and that many of the decisions that have to be made, at the time of authoring or evaluation of the accessibility of the resource, are subjective, it is not possible for an author to know with certainty what is required in any given situation. This means that authors need to be very expert in matters to do with accessibility to make authoring decisions. There is ample evidence available to show that such expertise is not common: even among those who are interested and try to make their resources accessible, there are many perceived difficulties and such people represent only a very small proportion of the people who author digital content for the use of others. The Australian-based Web Standards Group technical supportlist is busy on a daily basis. [HREF6]
In addition, expertise is often used to design resources so they perform well on a given set of devices, such as specified software and hardware, when neither of those selected are conformant to the recognised accessibility standards and so what suits the proprietary devices is frequently unsuitable for users with other software or devices.
Currently an author who attempts to make accessible resources almost always uses software that enables them to examine and alter the encoding to increase the accessibility of the resource. To make universally accessible resources, they have to evaluate their resource and, where necessary, augment it with alternatives for any aural, visual or tactile content or where controls are dependent on any of these sensory capabilities. In addition, if the resource is to be accessible to people with disabilities such as dyslexia, they may need to provide alternative versions of text and augment the text with more than the usual amount of visual material while reducing the content’s complexity.
Usually an author has an audience in mind, and it is not uncommon for such an author to consider that their audience will not include anyone with disabilities, or that they know the specific range of disabilities of their users. They are often wrong, as many digital disabilities are not visible and not accurately anticipated by authors, or later their resources are put to different uses. On the other hand, there are few users who need the full range of possibilities demanded by the WCAG for universally accessible resources. This means there is a great deal of redundancy implicit in the authoring of universally accessible resources. It is not easy to convince authors to do all the work that is required to make resources universally accessible and where they are convinced, to enable them to do this without great difficulty where the resources are interactive and interesting.
Publishers of resources are responsible for the publication of resources developed by their employees. This means they have a significant interest in the production of accessible resources. Publishers’ efforts to improve the situation have taken many forms.
Typically, responsible institutional or corporate units start by evaluating the (universal) accessibility of resources and simply requiring authors or others to increase their accessibility. This is not an effective strategy, given the difficulties for the authors.
The difficulty of rendering existing resources universally accessible is well known. In fact, the difficulties have proven so great and impractical that institutional developers try to insist upon universally accessible design ab initio.
As continuous advocates of the ‘get it right the first time approach’, we consider that this is not really making enough difference: millions of pages are still being created in inaccessible formats and there are not enough competent authors to change this. What was manageable in the early days of Web content development has become too complex and, like the author, many have had to recognize they can no longer be sure they know how to make universally accessible content.
To raise the standard of accessible authoring is a significant task then. Nevertheless, to achieve at least awareness of the problems, on a wide scale, the typical institution initiates training of potential authors. Soon, this proves too difficult: even without hard data to support this claim it can be assumed from the lack of accessibility of resources in general and the knowledge that most institutions have been trying to improve their accessibility record through training for a number of years now.
An alternative strategy is for the institution to impose templates on authors that limit their flexibility in order to enforce accessibility. This is usually done for what are considered ‘corporate pages’ in a main Website but it is not usually done at the level of the individual author who is simply communicating with others. It has the risk attached that a simple mistake in accessibility in a template can be proliferated across a Website with incredible ease. It is also a limiting practice. Institutions would not impose templates on their staff for normal writing and yet they do this as if the Web is not a natural medium for communication.
Indeed, the W3C Authoring Tools Accessibility Guidelines (ATAG) [HREF7] have helped authoring tool developers make better tools for authors, but these tools are only just emerging on the market and many authors will or can not pay for new tools. We, as authors of the ATAG, consider that they have a significant role in the strategy institutions should adopt. Instead of requiring an author to make good their resources, or others to repair them, we propose that publishers should supply their authors with the latest ATAG compliant tools that will do most of this work for the authors.
In a 2003 audit of the La Trobe University Website (typical of Australian university Websites) [HREF8], a section containing more than 14,432 pages was surveyed and tested for conformance with the WCAG. When detailed audits of sub-sets of pages were undertaken, very few pages were found to be conformant. In fact the site was not particularly bad but most pages contained at least one tiny inaccessible feature so they were not conformant. The WCAG test of universal accessibility of resources is absolute and can lead to a misleading description of an almost perfect resource.
Further, as the Disability Rights Commission (DRC) in the United Kingdom showed in their 2004 report entitled, “Formal Investigation: The Web: Access and Inclusion for Disabled People” [HREF9], the current W3C Web Content Accessibility Guidelines (WCAG 1.0) which are, it is generally agreed, the best in the world, do not cover all situations and resources can be inaccessible even when they conform fully to the guidelines. It is not conceivable that even the new guidelines [HREF10] due for release in the next few years, will solve all the problems. Perhaps it is just not sensible to expect such guidelines to anticipate every situation and provide for it. In addition, it may not be realistic to expect every author to make every page fully conformant even though many of them can produce very useful and popular information and services for the Web.
In the DRC project in the UK, where requirements for accessibility are similar to those in Australia, the results were distressing. In that project, extensive testing was done of the accessibility (as defined by WCAG) and usability of 1000 websites. Many pages failed the basic elements of the WCAG test. Further, it was found that in some cases, pages that did pass that test were nevertheless inaccessible to live subjects and almost certainly would have failed usability tests.
In general, research and experience have shown that very few interactive resources, a necessary part of today’s educational materials, are universally accessible. Repairing inaccessible resources is rarely just a matter of providing extra code but usually involves having to re-work completely the structure of the resource. It can be a very challenging task. Often the best result is only an approximation of the original resource, and the author may not be pleased about its quality or use to replace the original. This is particularly true when the author has taken advantage of proprietary features of a user access agent (e.g. browser or a particular multimedia player) to achieve some special effect that cannot be got using fully conforming accessibility specifications.
Typically, a resource is considered accessible if it satisfies the guidelines and inaccessible if it fails them. Success and failure in terms of the accessibility of resources is relative in real life. A rich, complex, generally accessible resource may be good and useful but described as ‘inaccessible’.
Testing against the guidelines is not time-wasting: it is the best-known way of getting an indication of the universal accessibility of a resource, and for that it is a useful thing to do. At this point, however, it is important to ask what is the point? Is not the real issue whether it is accessible to the user?
Instead of the universally accessible resource approach, we focus on what is delivered to a user in the ‘AccessForAll’ approach. In an electronic and information technology context, this means matching users’ capabilities, at the time, and the resources they need to use. The definition’s authority exists in the field of accessibility of electronic resources [HREF11] as well as in the more traditional area of physical access to the built environment. AccessForAll is about the relationship between a resource and its user. It is not a fixed quality of a resource. Resources can be adapted right up until the moment of delivery.
Disability is no more an absolute quality of a person than accessibility is an absolute quality of a resource. Users, including those with permanent disabilities, have requirements that often differ from place to place, time to time, and according to context and purpose. For instance, a user in safety clothing using dangerous equipment may not be able to use a computer mouse and keyboard and so may need voiced instructions. The same user may return to a company computer laboratory and welcome the use of the mouse and keyboard. The same user again, this time in a lecture theatre, will need large print projected on a corporate display screen.
Accessibility needs and preferences can apply to classes of users or to individual users. User accessibility needs are extended to include preferences as many users request support in ways that are appropriate at some times but not at others, as in the case of the use of a dictionary or calculator. Many users cannot operate without certain needs fulfilled but find life easier with other preferences accommodated.
Widespread implementation of the AccessForAll model will require Web services that support profiles of users’ needs and preferences; descriptions of resource accessibility characteristics; evaluations of matches of user profiles to available resources; post-hoc authoring of accessible components for resources, and reassembly of resources with appropriate components for delivery to users. We are confident that, given such services, the network effects gained by widespread use will further increase the accessibility of resources as the authoring and availability of accessible components become increasingly distributed.
Based on research and experience, we believe trying to make all authors produce universally accessible resources has failed because it is too expensive, too demanding in terms of skills, and that it is unlikely to achieve its goals in the future.
The common response to accessibility requirements is a significant commitment to training.
The Authoring Tools Accessibility Guidelines of W3C can be used as functional requirements for the accessibility of authoring tools of all kinds. The goal is accessibility-promoting tools so authors of resources, simply by using the tools, will make their resources accessible. We, as authors of the Guidelines, believe that authoring accessible content can be improved so significantly by the use of accessibility specifications conformant tools that the latest tools should be bought for authors instead of intensive training. In fact, we suggest that those who do not want to buy such tools should use the available high-quality open source ATAG-compliant tools.
There is no escaping the need for accessible content components, of course, but they do not need to be part of the original resource for the resource to be accessible at the time of presentation to the user. Sometimes the offending content can be transformed by a service into something that is accessible; sometimes acceptable alternative content can be found, and at other times new alternative content can be authored, as when closed captions or transcriptions are produced for video objects. Where there is a need for the production of new alternative content, we consider that often it can be created by untrained, willing authors using good authoring tools just in time.
The range of ATAG compliant tools support accessible authoring practices by providing:
Currently, there are very few people capable of knowing for sure if a resource is, in fact, universally accessible, even where this is tested according to the guidelines. This cannot be tested automatically and depends upon extensive knowledge of WCAG techniques and expert judgment for testing. There is software to assist in this process but it requires human effort and knowledge with respect to the guidelines. Good evaluation tools help testers work their way through the testing process, keeping track of what is tested and the results, and can produce interoperable, machine-readable reports.
Given that few resources are universally accessible, however, we can assume that most resources will need attention if they are to be rendered accessible for a particular user.
As a strong motivation for accessibility often arises in a community of users rather than authors, it is not uncommon to find a third party creating an accessible component for an existing resource or part of a resource. Usually closed captions for films, for example, are produced by a party specializing in captions . So are the foreign language versions of the spoken sound tracks. ubAccess [HREF12] has a service that transforms content for people with dyslexia; a number of Braille translation services operate in different countries to cater for different Braille dialects, and online systems such as Babelfish [HREF13] help with translation services.
Creating the accessible alternative components and making them available for use is shared by accessible content authors and repositories. Once there is an alternative for a resource component, it is a pity if a new one has to be created just because the existing alternative cannot be found. This means, of course, that repositories of accessible content should be online and their collections available and discoverable (see below). In the case of education, there should be no barriers to the development of networks of distributed accessible components.
Currently there is no standard way of providing descriptions of user’s needs and preferences in terms of accessibility. These need to be available anywhere in a common, machine and human readable language, even where the users are operating in languages other than English or users are unable to articulate their needs for themselves.
Institutions can be expected to want to have generic user or community profiles in which they can specify that certain logos appear on resources and perhaps that certain graphic styles apply. Such examples suggest that users will require not only multiple profiles, but also that profiles should ‘cascade’ at least allowing those which are of essential needs and preferences to override those that are purely commercial or decorative.
The management of user profiles will need to include then, the adjusting of values for elements within the profile according to immediate needs, the over-riding of more essential profiles over less essential ones, and the choice of profiles according to locations or circumstances.
In some cases, the requirements will be clear: a person who is ‘eyes-busy’ will need all visual modality information to be replaced by some other accessible form. It is important to note that they will also need all information relating to control of the device in non-visual form. This means that both resource presentation and device controls need to be adjusted. The profile has to contain information about how to adjust the controls. Additionally, as shown above, the profile may contain information about the substance of the resource being accessed, the form of information or the service being offered. For someone who is dyslexic, for instance, it will be necessary to substitute appropriate text if it is not already available. This amounts to a change in the content that is distinguishable from the form of its presentation and its control. There are, then, three major dimensions for the profile: presentation, control and content characteristics. In addition, all of these may change according to circumstances, so a description of the needs and preferences identified by context may be desirable.
In collaboration, the IMS Global Learning Consortium, the Dublin Core Metadata Initiative, the CEN-ISSS LTSC and others have co-authored the AccessForAll taxonomy of needs and preferences [HREF14]. It has been fully specified in the AccessForAll specifications. It is undergoing additional scrutiny as it progresses through the International Standards Organisation’s processes. The taxonomy can be used to provide what are known as personal needs profiles (PNP), according to these specifications. PNPs are easily created using Web forms and they can be stored and delivered using a Web service. Many users may share a PNP.
Just as user needs and preferences must be described in a common, machine-readable language, so must digital resource accessibility characteristics.
The difficulty with describing any resource is that it is almost always composed of a number of items and it may be only one or some of those individual items or objects that are inaccessible. This means that somehow the alternatives have to be organised in a logical fashion. They may, of course, also have to be organised and presented together, even where some of them are not co-located.
Most resource descriptions (DRDs) are done at resource level. Having a distributed model for resources where they are described at object level is an extra challenge. It seems sensible to identify the original objects, and to classify those that are needed in special circumstances (to be substituted for them, or to augment them) as alternative resources.
Following the description requirements for the PNP makes sense but it may mean having many descriptions for many items. In order to manage this, automating at least part of the process and storing the results is necessary. Determining all the characteristics of a resource or object that will be needed is a mixed activity with some being determinable automatically, such as the format of the digital file, and some being subject to human judgement, such as whether a text description of an image is really equivalent to the image or not.
As there are legal requirements that relate to the accessibility of content objects, it is not always possible to be sure that the descriptions available are reliable. For this reason, a notion of trustworthy descriptions is important. Similarly, it is important to know when the resource or object was evaluated, as it may change over time. As the content may be distributed, so may the descriptions of it, and so they should all be in a standard form and interoperable.
Alternative content components are often produced by specialist businesses such as caption centres. These components may already be profiled according to local specifications. If they have no profile records (metadata) or have them in a local form, these can be created, or otherwise they can be harvested and transformed into the common format. They can then be included in the discovery hub for distribution via a Web service.
The AccessForAll specifications produced by the IMS-led collaboration provide a suitable taxonomy and specifications for digital resources.
When every effort has been made to make a resource technically accessible, it can still fail to be useful to a user for non-technical reasons. A user who is dyslexic or a blind may find that resources that are very good for their colleagues are not at all appropriate for them. In such cases, what are called ‘equivalent alternative resources’ should be provided. Such resources serve the same purpose, but achieve the goals in a different way. Their description is, therefore, not the same as that of the original resource, even though the purpose might be.
Identifying alternative resources with a similar purpose normally requires human judgement. There is already a suitable common language for learning resource descriptions in the IMS Global Learning Consortium’s Learning Design specifications [HREF15].
Services for automatically generating alternative content are being developed using Semantic Web technologies. Already some components can be retrieved from existing services such as the Smart Web Accessibility Platform [HREF16] and AnnoSource (a La Trobe student project for a service that helps with rendering inaccessible content accessible ).
Finally, there is a need for a service that provides the right combination of content and services for the user, where and when they need it. This means there needs to be a way of bringing together all the pieces, including the user and resource profiles, the context information, and the pieces that are to be assembled for delivery to the user as the resource they require.
For a user, or an assistant working with them, it must be possible to create the necessary profiles and to change them for the immediate circumstances. In addition, it must be possible to make formal descriptions of the resources and link all of these together for the matching process. There are several layers of discovery involved. There is more than just discovery information needed, however, and therefore a need for systems that facilitate the making of such descriptions.
The Inclusive Learning Exchange (TILE) [HREF17] is an example of a learning management system that works by checking the user’s profile and then finding objects from which to compose a resource that suits their needs. As TILE includes a tool for creating and editing the user’s profile, this can be done while the user is using the service. TILE uses the AccessForAll metadata profiles to match resources to user's needs, with the capability to provide captions, transcripts, signage, different formats and more to suit users' needs.
Fairfax, in Australia, has offered a striking economic reason for being concerned about accessibility. In 2003-4, they redeveloped their website with accessibility in mind and the result is a saving of an estimated $AUD1,000,000 per year in transmission costs. A bigger publisher would save even more. [HREF18]
Flexible assembly satisfies the requirements for the users, allows for more participation in the content production process and has the benefit that it limits the production and transfer of content that will not be of use to the recipient.
Descriptions of the accessibility of content of large collections can be done with tools designed for that purpose. Publishers can identify potential problems and gaps in their resource collections in advance.
The strategy proposed, using technology to augment, supplement and in some cases replace author expertise, is most likely to be achieved by a combination of tools than the adoption of any particular tool. Many of these are not yet available as Web services but are already available as system components. The big changes will be possible when they are made into services as this will increase the network capabilities of the systems, and thus the quantity of sharing that will be possible. The possibilities will only be realised if there is commitment to them. This is not so difficult to imagine: the achievements of normal people using word processors, electronic spreadsheets and presentation tools today are similar to what can be expected for accessibility in the future with the tools and practices proposed.
The work behind the user and resource profiles that allow for matching of resources to users' needs and preferences, was predicated upon the notion that digital resources are easily manipulated by computer applications. [HREF19]
In times of increasing complexity and reliance on technology, it is important to ensure that what is being gained is increased quality of life. Currently huge demands are made on all as a result of the move towards a more technology-dependent world, and opportunities lost by early introduction of new technologies are being repaired by careful development of newer technologies, but it is important to balance cost-effectiveness with outcomes. If we choose to support everyone in terms of greater adaptability to personal needs and preferences, and can do so by placing greater reliance on collaboratively developed technologies, we should find it easier to manage the financial and equitable burdens placed upon us.
Figure 1: Diagram showing flow of behaviours using AccessForAll metadata in TILE [HREF20].
Ford M. & Nevile L., 2004, “Issues enabling support for Multi-locational Accessibility“, IDABC: Cross-border E-Government Services for Administrations, Businesses and Citizens Conference', Brussels, February 2005. (in press)