Toward a Better Understanding of Accessibility Adoption: Developer Perceptions and Challenges

Abstract

The primary aim of this dissertation is to explore the challenges developers face in interpreting and implementing accessibility in web applications. We analyze developers’ discussions on web accessibility to gain a comprehensive understanding of the challenges, misconceptions, and best practices prevalent within the development community. As part of this analysis, we built a taxonomy of accessibility aspects discussed by developers on Stack Overflow, identifying recurring trends, common obstacles, and the types of disabilities associated with the features addressed by developers in their posts. This dissertation also evaluates the extent to which developers on online platforms engage with and deliberate upon accessibility issues, assessing their awareness and implementation of accessibility standards throughout the web application development process. Given the volume and variety of these discussions, manual analysis alone would be insufficient to capture the full scope of accessibility challenges. Therefore, we employed supervised machine learning techniques to classify these posts based on their relevance to different aspects of the WCAG 2.2 guidelines principle. By training our models on labeled data, we were able to automatically detect patterns and keywords that indicate specific accessibility issues, even when the language used by developers is not directly aligned with the official guidelines. The results emphasize developers’ struggles with complex accessibility issues, such as time-based media customization and screen reader configuration. The findings indicate that machine learning holds significant potential for enhancing compliance with accessibility standards, providing a pathway for more efficient and accurate adherence to these guidelines.