"With most of it being pictures now, I rarely use it": Understanding Twitter's Evolving Accessibility to Blind Users
Meredith Ringel Morris, Annuska Zolyomi, Catherine Yao, Sina Bahram, Jeffrey P. Bigham, Shaun K. Kane · 2016 · Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems (CHI 2016) · doi:10.1145/2858036.2858116
Summary
This multi-method study examines how blind people use Twitter and the accessibility barriers they face as the platform shifts from text-based to increasingly image-heavy content. The researchers combined an online survey of 132 blind Twitter users, large-scale analysis of six months of tweets and profile data from 116 blind and 116 sighted control accounts via the Twitter firehose API, and qualitative coding of 900 randomly sampled image-containing tweets. The study was motivated by Twitter's historical reputation as the most accessible social media platform for blind users due to its originally text-based nature, and investigated whether this was changing as embedded images became more prevalent. The survey captured blind users' motivations, challenges, and workarounds, while the firehose analysis provided objective behavioral data comparing blind and sighted Twitter usage patterns. The image analysis examined what types of visual content were being shared, how important images were to understanding tweets, and whether tweet text could serve as an adequate substitute for image descriptions.
Key findings
The proportion of tweets containing embedded multimedia nearly doubled from 15.4% in January 2014 to 28.4% by June 2015, with retweets containing even more imagery (22.8% to 42.2%). Blind users posted significantly less multimedia content (4.78% of tweets vs. 23.43% for sighted users). Of 756 image-containing tweets analyzed, 55.6% had images rated as "very important" to the tweet's meaning, yet 61.8% of accompanying tweet text would serve as a "poor" caption. Nine distinct image types were identified, with photographs (64.4%) being most common, followed by images with embedded text (11.5%), pictures of text (9.0%), and screenshots (7.0%). The study found blind users were significantly less likely to customize their profile photo (57.6% vs. 98.5% of sighted users) and header image (11.4% vs. 83.3%), with many unaware these visual elements existed or how to change them. A logistic regression model could distinguish blind from sighted accounts with 90.5% accuracy using features like multimedia posting rate, default profile images, and geolocation settings — raising privacy concerns about involuntary disability disclosure. 85% of blind respondents used Twitter for blindness-related reasons (advocacy, networking, information), and the most popular hashtags among blind users were disability/accessibility-related (#a11y, #blind, #accessibility) versus meme-related for sighted users (#wcw, #tbt).
Relevance
This research provides foundational evidence for why alt text on social media matters and anticipated the platform changes that eventually followed — Twitter added alt text support later in 2016, likely influenced in part by this work. The finding that over half of embedded images were "very important" to tweet meaning while tweet text almost never served as an adequate description quantifies the accessibility gap that missing alt text creates. For practitioners, several findings remain highly relevant: the diversity of image types (photos, screenshots, text-as-images, memes) means that no single automated captioning approach will suffice; the "screenshorting" trend of posting text as images continues to erode accessibility on all platforms; and the privacy risk of involuntary disability disclosure through usage patterns is an underappreciated concern in accessibility research. The study's recommendations — platform-level alt text support, showing image prevalence statistics on profiles, and addressing the root causes of screenshorting by expanding character limits — offer a blueprint for improving social media accessibility that extends well beyond Twitter.
Tags: social media accessibility · blind users · alternative text · image description · Twitter · privacy · disability disclosure