Screen Reader Users in the Vibe Coding Era: Adaptation, Empowerment, and New Accessibility Landscape
Project Overview
This document explores the impact of generative AI, specifically advanced AI code assistants like GitHub Copilot, on education through a longitudinal study focusing on screen reader users. The research reveals that these AI tools significantly enhance programming skills and accessibility for visually impaired developers, allowing for greater efficiency and skill development. However, participants identified challenges such as difficulties in communicating with the AI, reviewing outputs, and retaining control over the coding process. To address these issues, the study offers design recommendations aimed at improving the accessibility and usability of AI-assisted coding environments. Overall, the findings underscore the potential of generative AI in education to facilitate learning and skill acquisition while also highlighting the need for continued improvements to ensure equitable access for all users.
Key Applications
GitHub Copilot
Context: Programming tasks performed by screen reader users with visual impairments.
Implementation: Participants engaged with GitHub Copilot through a structured tutorial and followed up with two weeks of regular programming tasks, documenting their experiences.
Outcomes: Participants reported enhanced programming capabilities and increased efficiency, as well as a better ability to manage tasks that were previously challenging due to accessibility issues.
Challenges: Users faced difficulties in communicating with the AI, reviewing outputs, managing multiple views, and maintaining situational awareness.
Implementation Barriers
Inaccessibility of tools
Many programming tools and environments are not designed with accessibility in mind, making it difficult for screen reader users to navigate and use them effectively. This includes issues with complex interfaces that increase cognitive load.
Proposed Solutions: Design improvements focusing on UI simplicity, predictable interactions, and comprehensive screen reader-specific documentation. Streamlining UI design to reduce complexity and improving the predictability of interactions.
Learning curve
Screen reader users often face a steep learning curve when using advanced AI tools, as they have limited exposure and available resources tailored to their needs.
Proposed Solutions: Providing inclusive onboarding resources, tutorials, and community support aimed specifically at visually impaired developers.
Transparency and control issues
Participants expressed concerns about losing control over the programming process due to the AI's automation features, which can lead to unintended changes.
Proposed Solutions: Implementing clear status notifications and providing users with options to review changes before they are applied.
Project Team
Nan Chen
Researcher
Luna K. Qiu
Researcher
Arran Zeyu Wang
Researcher
Zilong Wang
Researcher
Yuqing Yang
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Nan Chen, Luna K. Qiu, Arran Zeyu Wang, Zilong Wang, Yuqing Yang
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai