"Ignorance is Not Bliss": Designing Personalized Moderation to Address Ableist Hate on Social Media

TL;DR

This work addresses the persistence of ableist hate on social media and the inadequacy of platform-wide moderation to protect disabled users. Through 23 interviews and eight focus groups employing design probes, it explores how personalized moderation can filter and present ableist content, with a preference for filtering by ableist types and using content warnings over full removal. Key findings reveal distrust in AI accuracy, contextual challenges of identifying ableism, and the importance of agency, reversibility, and transparency in design. The study offers concrete design recommendations to enhance safety and usability of personalized moderation for ableism, with implications for broader identity-based harms and platform responsibility.

Abstract

Disabled people on social media often experience ableist hate and microaggressions. Prior work has shown that platform moderation often fails to remove ableist hate leaving disabled users exposed to harmful content. This paper examines how personalized moderation can safeguard users from viewing ableist comments. During interviews and focus groups with 23 disabled social media users, we presented design probes to elicit perceptions on configuring their filters of ableist speech (e.g. intensity of ableism and types of ableism) and customizing the presentation of the ableist speech to mitigate the harm (e.g. AI rephrasing the comment and content warnings). We found that participants preferred configuring their filters through types of ableist speech and favored content warnings. We surface participants distrust in AI-based moderation, skepticism in AI's accuracy, and varied tolerances in viewing ableist hate. Finally we share design recommendations to support users' agency, mitigate harm from hate, and promote safety.

Figures (4)

• Figure 1: Screenshot of Intel's Bleep interface.
• Figure 2: Diagram of Design Probe 1, configuring filters based on ableism. This includes 3 different designs: 1A (toggling ableist content), 1B (slider based on quantity of ableist posts), and 1C (slider based on intensity of ableism).
• Figure 3: Diagram of Design Probe 2, configuring filters based on ableist types of hate. This includes 3 different designs: 2A (toggling each ableist type), 2B (slider for quantity of each ableist type), and 2C (slider for intensity of each ableist type). To reduce cognitive load, we asked participants to imagine design 2B and 2C to be applied to all the ableist types.
• Figure 4: Diagram of Design Probe 3, customizing the presentation of the filtered hate. The original hateful comment contained slurs and an accusation that the user was faking their disability. We presented three designs to obscure or describe the hate, in contrast to current filters which completely remove moderated comments: 3A (AI rephrasing of the hate), 3B (content warning categorizing the type of ableist hate), and 3C (general content warning of "ableism"). For each design, users have an option to click and view the original comment.