Using Social Media Websites to Support Scenario-Based Design of Assistive Technology

Abstract

Having representative users, who have the targeted disability, in accessibility studies is vital to the validity of research findings. Although it is a widely accepted tenet in the HCI community, many barriers and difficulties make it very resource-demanding for accessibility researchers to recruit representative users. As a result, researchers recruit non-representative users, who do not have the targeted disability, instead of representative users in accessibility studies. Although such an approach has been widely justified, evidence showed that findings derived from non-representative users could be biased and even misleading. To address this problem, researchers have come up with different solutions such as building pools of users to recruit from. But still, the data is not widely available and needs a lot of effort and resource to build and maintain. On the other hand, online social media websites have become popular in the last decade. Many online communities have emerged that allow online users to discuss health-related subjects, exchange useful information, or provide emotional support. A large amount of data accumulated in such online communities have gained attention from researchers in the healthcare domain. And many researches have been done based on data from social media websites to better understand health problems to improve the wellbeing of people. Despite the increasing popularity, the value of data from social media websites for accessibility research remains untapped. Hence, my work aims to create methods that could extract valuable information from data collected on social media websites for accessibility practitioners to support their design process. First, I investigate methods that enable researchers to effectively collect representative data from social media websites. More specifically, I look into machine learning approaches that could allow researchers to automatically identify online users who have disabilities (representative users). Second, I investigate methods that could extract useful information from user-generated free-text using techniques drawn from the information extraction domain. Last, I explore how such information should be visualized and presented for designers to support the scenario-based design process in accessibility studies.