Accessibility Heuristics Utilizing Learnability Characteristics of Synthesized Speech Applications

by

Ted L. Wattenberg

A Preliminary Proposal Paper submitted in partial fulfillment of the requirements

for the degree of Doctor of Philosophy

December 27, 2004

Graduate School of Computer and Information Sciences

Nova Southeastern University

Table of Contents

Chapters

1. Introduction 1

Accessibility Usability Methods Need to Include the Learnability of Assistive Technologies 1

Adapting Evaluation Methods for Assistive Technology 1

Defining Learnability in Terms of Assistive Technology 3

2. Problem Statement and Goals6

Goals 8

3. Relevance, Significance, and Literature Review 12

Text-to-Speech 13

Models of reading Comprehension 16

The Physiological and Neurobiological Mechanisms of Reading 16

The Incrementality of Reading 18

The Causal Effects of Reading 19

Landscape Model of Reading 19

Prior Knowledge, Reading Strategies, and Reading Comprehension 21

Models of Screen Reading Comprehension 22

Linear vs. Nonlinear Reading 22

Cognitive Problems Associated with Screen Readers 23

Learnability Characteristics Attributed to the Use of Text-to-Speech 27

Instructional Strategies Associated with Learnability 30

At-risk Readers 31

4. Barriers and Issues 33

5. Approach 36

Research Design 40

Online Focus Groups 42

Planning 44

Pre-testing 49

Sample 50

Analysis 53

Findings: Accessibility Heuristics 56

Human Subject and Ethical Concerns 57

6. Milestones 57

7. Resources 58

8. Appendixes

Appendix A – Home Page for the Accessible Learning Through Text-to-Speech Project 60

Appendix B – Gatekeeper Online Form 61

Appendix C – Participant Online Form 64

9. Annotated Bibliography 68

10. References 165

1

1

Introduction

Accessibility Usability Methods Need to Include the Learnability of Assistive Technologies

Adapting Evaluation Methods for Assistive Technologies

Americans with disabilities are less than half as likely to own a computer and only one quarter as likely to use the Internet as someone without a disability (Kaye, 2000; Wattenberg, 2004b). Estimates of World Wide Web (Web) site inaccessibility to users with disabilities range from eighty percent up to ninety-five percent (Shneiderman, 2003). Studies by Nielsen (2001) established baseline usability of people without disabilities working on the Web to be three times greater than people with vision impairments using screen readers or screen magnifiers. These low rates of usage are due partly because the use of assistive technologies involve the integration of hardware and software that result in usability characteristics that are more complex than merely the summation of the problems associated with individual devices and computer applications (Arnheim, 1999; Scherer, 2004).

It is difficult to adapt traditional evaluation methods of guidelines, usability testing, and heuristic evaluations for the purpose of finding and solving accessibility usability problems (Dicks, 2002; Scherer, 2004). There have been nearly two hundred guidelines produced by varying governmental and professional organizations trying to capture the multitude of criteria involved in universal usability product design (Vanderheiden, 2000). One of the most used sets of guidelines, adopted by the World Wide Web Consortium (W3C) as the Web Accessibility Initiative, are the Web Content Accessibility Guidelines 1.0 (Chisholm, Vanderheiden & Jacobs, 2001; Wattenberg, 2004b). While the use of these and other guidelines have been shown to be beneficial in the development of accessible product prototypes they have not resulted in many marketable products that are both usable and accessible by people with disabilities (Redish, Molich, Bias, Dumas, Bailey & Spool, 2002). Some of the problems that have been identified with the use of Web accessibility guidelines are that the complexity of the guidelines make it difficult for multiple design teams to coordinate effectively; that even after the use of guidelines, significant accessibility problems can be found due to the vast differences between learning styles of people with disabilities; and that many accessibility errors are missed because of the rapid introduction of new products which the guidelines did not originally consider (Jackson, 2003). In previous studies, Nielsen and Molich (1990) found that experienced programmers and product designers, when solely using guidelines, could only prevent thirty percent of identified usability problems.

A number of automatic validation tools are available to help evaluate the accessibility levels of a completed Web site including Bobby, A-Prompt, W3C HTML Validation Tool, AccVerify, and Lift (Slatin & Rush, 2003, chap. 6; Wattenberg, 2004b). All of these tools are beneficial aids but assume that developers have the time and motivation to comprehend the complex and often lengthy recommendations that the validation tools produce. Studies indicate that most designers are not adequately trained in their use or fully understand the need to comply with accessibility requirements. These tools have also been found to bypass emerging technologies and miss important usability problems.

Nielsen and Mack first introduced heuristic evaluations in the early 1990’s utilizing novice to expert evaluators to test for usability problems by following sets of known heuristics (Nielsen & Molich, 1990). Heuristics are similar to design principles and guidelines, except that in usability evaluations they are used to discover problems rather than to remedy product development requirements. Originally, Nielsen created ten heuristics that were found to be pertinent in usability evaluations of Web sites (Nielsen, 1994; Preece, Rogers & Sharp, 2002, chap. 13). Current research indicates that expert evaluators are more effective in finding usability problems than novice users of technology or those not trained in heuristic evaluation methods (Dicks, 2002; Molich & Jeffries, 2003; Nielsen, 2001; Paddison & Englefield, 2003; Redish et al., 2002). Additionally, double experts, those evaluators that are both experienced in usability methods and the specific technology being studied, are the most effective usability evaluators.

Modification of Nielsen’s (2001) original 10 heuristics for finding Web-based usability problems using discount evaluation methods is necessary in order to successfully identify accessibility usability problems. Paddison and Englefield (2003) utilized two constructs, technical accessibility and usable accessibility, to measure how people with disabilities use technology. These constructs differentiate between the engineering practices of ensuring that all technical components of a computer environment meet current accessibility legislative requirements, technical accessibility, or whether a product actually accentuates practices of user-centered design, usable accessibility. A key concern highlighted by Paddison and Englefield is the need for usability evaluators to have knowledge in accessibility, the use of assistive technology, and other issues of usability affecting people with disabilities. In contradiction to their findings, Paddison and Englefield’s heuristics were not developed with significant input from actual users of assistive technology or experts in their use. The researchers expressed doubts about knowing how to contact or effectively involve actual users of assistive technology within a usability methodology.

Defining Learnability in Terms of Assistive Technology

As the options of assistive technology and their availability have increased, so have the recommendations by rehabilitation and educational specialists increased for their use by people with disabilities (King, 2003; Scherer, 2004, 2003; Seelman, 2003, 2001). Complicating these recommendations is that the evaluation process of determining what technology is appropriate for each person is a difficult task and has only just begun to be studied. Studies by Riemer-Reiss and Wacker (2000) found that people with disabilities abandon as high as seventy-five percent of assistive technologies given to them. Many users become dissatisfied with their devices and services, usually resulting in the discontinuance of the assistive devices. Factors that affect user choices are relative advantage, compatibility, trialability, re-invention, and training.

The usefulness of assistive technology is dependent on whether the person using the technology can utilize it to feel more connected to their environment, develop a more positive self-identity, enjoy associations with other individuals or groups of people, or work collaboratively with others (Scherer & Parette, 2004). How assistive technology is delivered is as important as whether the technology was originally designed to be accessible. Studies by Scherer (2004) indicate that it is not usually the inability of a technology to work or to perform the needed adaptive action, but the emotional acceptance and usage of the technology by the person with the disability that prevents assistive technology from being more widely accepted and used. The following research areas, considered essential for the success of a good match between a person with disabilities and an assistive technology, have been identified as needing further study (Scherer & Parette, 2004; Scherer, 2003):

  1. Learner characteristics and preferences: Technology should be used to enhance a person’s ability to learn and be independent. Information is needed about how to affectively integrate technology with each person’s unique learning styles.
  2. Environments for learning: Lifelong learning involves diverse learning centers such as libraries, laboratories, community-based centers, the home, collaborative and team work groups, and new distance learning environments. Information is needed in understanding how assistive technology is utilized in all types of learning environments.
  3. Access to information and instruction: Technology can diversify the availability of information by allowing content to be presented in alternate formats; Braille, e-text, captioning, or enlarged text. Effective methods of delivering the content and emotional meaning of information to diverse learning styles need to be studied.

4. Systems and methods for instructional delivery: Learner-centered instructional strategies are needed to help people with disabilities to use new technologies and gain access to digital information.

Within the fields of speech pathology, linguistics, and developmental psychology learnability is a specific construct of language acquisition that measures a person’s ability to acquire and use language communication skills (Haramundanis, 2001; Mirenda, 2003). Learnability theory encompasses areas of grammar and syntax, cognitive science, developmental psychology, and educational theory. Measurement criteria of a person’s level of learnability is based on educational theories that describe learning as an active process involving social communication and the ability to utilize knowledge beyond the immediate moment or situation (Dewey, 1938; Bredo, 1998, Haramundanis, 2001; Mirenda, 2003). Learning implies a continuance of experience, linking a previous activity to future activities.

Learnability constructs have been utilized within theories of artificial intelligence and computational learning for many years (Beimel, Bergadano, Bshouty, Kushilevitz & Varricchio, 2000), but have only recently begun to be adapted for use within the field of Human Computer Interaction (Haramundanis, 2001). Theories of artificial intelligence are concerned with how people understand and learn complex mathematical concepts so that once a concept is understood within one context the knowledge can be transferred to other activities (Beimel et al., 2004). Computational learning algorithms consider the development of systems where computers automate functional tasks without human assistance. Designers of computer technology products are becoming interested in the learnability of computer games that include the use of new applications of voice recognition and text-to-speech technologies (Egloff, 2004; Keller & Stevens, 2004; Paymans, Lindenberg &Neerincx, 2004).

Learnability theories differentiate between the usability of a product and its learnability (Egloff, 2004; Paymans et al., 2004). Key constructs of learnability are the ease of use of a product, concepts of gender preferences, and the level of cognitive development needed by the user. The usability of a product relates to the enjoyment derived by the user and leads to the overall effectiveness of the product’s use. The ability of a user to have a continuity of usability leads to that product’s level of learnability. Paymans et al. (2004) found that a product can be found to be easy to use but at the same time not easy to learn. Even though a user understands the content of material or can perform an independent activity, they still may be unable to reapply the skills or knowledge from one activity to new activities later on. Additionally, the level of usability or learnability for an individual activity may be different than levels found when using the same technology within a collaborative or group environment.

Problem Statement and Goals

Assistive technologies have the potential of helping and enabling people with disabilities, but often result in opposite outcomes of limiting, isolating, or making people feel more dependent on others’ help to survive (Scherer, 2004). An assistive technology must be both usable and learnable for it to be accepted and retained by a person with a disability (Egloff, 2004; Paymans et al., 2004). Unfortunately, there is not always a positive correlation between the usability of an assistive technology and its learnability. It has been found to be difficult to adapt usability evaluation methods for the purpose of finding and solving accessibility usability problems primarily because of the diversity of user learnability characteristics (King, 2003; Rowan, Gregor, Sloan, & Booth, 2000; Scherer, 2004; Shneiderman, 2003; Wattenberg, 2004a).

The specific problem that will be addressed in this dissertation is the inability to perform adequate accessibility usability and learnability evaluations of assistive technology products utilizing text-to-speech applications of synthesized speech (Walth & Wattenberg, 2004; Wattenberg, 2004a; Wattenberg, 2004b). Previous attempts of developing usability evaluative methods or instructional strategies that are effective for all users of assistive technology as a homogeneous population have not been successful. Synthesized speech applications include voice recognition and text-to-speech. Voice recognition applications use voice commands as a method of inputting information or controlling computer-based technologies. Text-to-speech applications convert printed information to computer synthesized speech that can be controlled by the user.

The concentration of synthesized speech applications of text-to-speech is supported by previous work linking concepts of learnability within the fields of speech pathology, linguistics, developmental psychology, artificial intelligence and computational learning and a reader’s comprehension of e-text (Egloff, 2004; Keller & Stevens, 2004; Paymans et al., 2004). Users of synthesized voice applications are the fastest growing populations of users of assistive technology and include people with vision impairments, learning disabilities, language deficiencies, and people with physical and mobility impairments (Walth & Wattenberg, 2004; Wattenberg, 2004 b). While the population of users of synthesized speech applications as an assistive technology is diverse, their learnability characteristics are similar, requiring the ability to form mental images through audio cues that facilitate higher cognition.

Limiting this study’s focus to text-to-speech applications is warranted because of the increased research complexity that would be required of separating and independently observing the sub-population groups for all users of synthesized speech applications. This complexity would extend the study beyond a reasonable length of time and require excessive numbers of observed participants to measure adequately. Additionally, national surveys of special education teacher programs regarding the integration of assistive technology have identified text-to-speech applications as the most useful assistive technologies for students (Michaels & McDermott, 2003; Michaels, Prezant, Morabito, & Jackson, 2003). The reports also identified the need for effective methods of preparing teachers how to deliver these technologies to students with disabilities.

Goals

The general goal of the proposed study is to support the increased effectiveness and use of technology by people with disabilities by the identification of learnability problems of users of text-to-speech applications. The identification of learnability attributes of synthesized speech is necessary for the affective development of instructional strategies. Accessibility usability heuristics will provide educators with a tool to gain a deeper understanding in how people with disabilities use technology. It is essential that users and experts in the use of assistive technologies are incorporated into these methodologies because of the complexity and diversity of learnability characteristics.

This student proposes to accomplish this goal by the completion of specific goals within two areas: 1) the discovery of what has already been studied and shown to be valid accessibility learnability and usability criteria of text-to-speech applications; and 2) the development of a set of accessibility heuristics for the purpose of identifying learnability problems of text-to-speech applications that utilizes evaluators that are users or experts of assistive technologies.

The student proposes to complete a review of literature concerning three areas of the usability of assistive technologies. The first concerns what is known about learnability characteristics of people with disabilities that utilize text-to-speech applications (King, 2003; Scherer, 2004; Vitense, Jacko, & Emery, 2002). The second area is about discovering what is known of usability evaluation methods that have been successful finding accessibility problems for text-to-speech applications (Nielsen, 2001; Paddison & Englefield, 2003; Redish et al., 2002; Rowan et al., 2000; Slatin & Rush, 2003; Vanderheiden, 2000). The third area of the review of literature will examine the affectiveness of delivery models of text-to-speech applications used as an assistive technology (Scherer, 2004).

Additionally, the student proposes to complete a qualitative analysis of the experiences of users and experts of assistive technologies concerning learnability, usability, and other delivery problems that affect the usage of text-to-speech applications (Paddison & Englefield, 2003; Scherer, 2004). The combination of reviewing quantitative studies along with a qualitative analysis of the ideas of actual users and experts is necessary because of the diversity of learnability patterns of people with disability and the relative decentralized nature of the delivery of assistive technology by rehabilitation specialists.

The student proposes the development of a set of accessibility heuristics based on the review of literature and the qualitative analysis of users and experts of text-to-speech applications used as an assistive technology. The ability to measure both the usability and the learnability of assistive technologies will support future studies concerning the use of assistive technologies (Dicks, 2002; Paddison & Englefield, 2003). Some of these areas are the correct matching of a technology to a person with a disability; the effective use of assistive technologies within the work place; the use of assistive technologies while performing specific learning tasks of reading, writing, or mathematics; and the problems associated with using assistive technology while working collaboratively.

The proposed research concerning the usability, learnability, and delivery of text-to-speech applications suggest the following research questions:

  1. How does the usability of text-to-speech applications affect reading performance?
  2. What is the link between usability and performance?
  3. What usability or performance based attributes, characteristics, or heuristics concerning the use of text-to-speech technologies have already been identified?
  4. What modifications of usability evaluation methodologies have already been tried in order to understand accessibility?
  5. What are the requirements for the successful participation of users or experts of assistive technology within usability evaluative methodologies?
  6. How can the basics of learnability theory be used to better understand how people learn to use text-to-speech technologies?
  7. How are learnability characteristics of readers using text-to-speech technology different than those readers of traditional text formats?
  8. What are the sub-populations of users of text-to-speech technologies and how are their learning modalities differentiated?
  9. How do differences in speech voices, the rate of speech, the gender of the voice, errors in the text, and syntax affect users of text-to-speech technologies?
  10. How do hypertext and other computer-based text characteristics of text-to-speech technologies affect learnability?
  11. How does the use of text-to-speech technologies affect factors of social learning, such as group participation, collaborative learning, and team projects?
  12. How are text-to-speech technologies currently delivered to people with disabilities?
  13. How are assistive technologies matched and procured by people with disabilities?
  14. What instructional strategies have been used to help people learn with text-to-speech technologies?
  15. How has text-to-speech technologies been integrated within employment and workplace environments?

The student proposes to name this study the Accessible Learning Through Text-to-Speech (Alt-Learning) Project. The Alt-Learning Project, through the use of online focus groups, will try to identify usability and learnability attributes of people using text-to-speech technology as an assistive technology. Two primary target groups of users of text-to-speech technology will be observed: people that are considered blind and people with vision but cannot read normal text materials due to learning disabilities or mobility impairments. Additionally, focus groups consisting of professionals responsible for the delivery of assistive technology will explore the concepts of usability and learnability from the perspective of instructional strategies and the matching of assistive technology with a person with a disability.