Files in this item



application/pdfZHANG-THESIS-2021.pdf (14MB)Restricted Access
(no description provided)PDF


Title:Exploring assistive transfer systems for people with visual impairments in addressing privacy and security tasks
Author(s):Zhang, Zhuohao
Advisor(s):Wang, Yang
Department / Program:Computer Science
Discipline:Computer Science
Degree Granting Institution:University of Illinois at Urbana-Champaign
Human-Computer Interaction
Usable Privacy and Security
Abstract:People with visual impairments (PVIs) often face challenges when dealing with day-to-day visual tasks such as solving visual CAPTCHAs or screening pictures before posting them on social media. What if PVIs could transfer these visual tasks to a helper to solve? We introduce assistive transfer systems, which allow PVIs to solicit just-in-time help where a willing helper directly solves the challenge on the PVIs' behalf. So how might PVIs want such a system configured? How could AI be used as a first-layer solution to automatically solve the task? If the AI is not good enough, could the PVIs transfer the task to another person, e.g., a friend or family member? From whom they would solicit help? Whether and how they would compensate for this help? In this thesis, we discuss two concrete use cases related to designing assistive transfer systems in order to answer these questions: solving task-based visual CAPTCHAs and posting pictures to social media. We implemented two proof-of-concept assistive transfer systems — WebAlly and ShareAlly — to help with these two scenarios accordingly. We used an exploratory, role-play study with 10 pairs of participants — a PVI and a friend or a family member — to evaluate one of the implemented systems WebAlly. We asked participants to use WebAlly in four different configurations that varied in the source of help (friend vs. stranger) and compensation (paid vs. volunteer). We found that PVIs liked having WebAlly as an additional option for solving visual CAPTCHAs, when other options that preserve their independence fail. In addition, many PVIs and their friends felt that using the system would bring their relationship closer. We discuss design implications for these human-AI hybrid assistive transfer systems more broadly, e.g., the importance of complementing rather than replacing PVIs’ existing workflows.
Issue Date:2021-04-23
Rights Information:Copyright © 2021 Zhuohao Zhang
Date Available in IDEALS:2021-09-17
Date Deposited:2021-05

This item appears in the following Collection(s)

Item Statistics