Aleksandra Kukawka
Hello everyone!
I completed my Master's Thesis on exploring the potential of LLM-generated responses for blind and visually impaired people under the supervision of Sarah Sahabi in November 2024.
Thesis
Abstract:
Blind and visually impaired (BVI) people face many challenges in various areas of life, particularly in accessing digital content. They rely on assistive technologies like screen readers and voice assistants, but these tools have limitations. With the growth of artificial intelligence (AI), there is an opportunity to enhance the existing assistive technologies and develop new ones. My thesis investigates the potential of AI and responses generated by large language models (LLMs) to help BVI people in their daily lives. Using a combination of prototype testing, user interviews, and online surveys, we employ the grounded theory approach to understand the needs and preferences of BVI people regarding AI-based systems. We answer the following research question: What are the advantages and disadvantages of AI voice assistants with LLM-generated responses for BVI users, and how can they be improved? The findings suggest that AI can help BVI people in their daily lives. They value the personalisation and natural interactions such a system can provide. However, due to existing limitations of AI technologies, such as privacy concerns, efficiency, or inaccuracy, they still prefer to use traditional tools like screen readers over entirely replacing them with AI. Our work contributes to understanding the potential of AI for BVI users and provides insights into how AI-based systems can be designed to meet their needs better.
I'm also a former student assistant working under Adrian Wagner's supervision on the IDE research and plugin code.