An Uninteresting Tour Through Why Our Research Papers Aren't Accessible
Jeffrey P. Bigham, Erin L. Brady, Cole Gleason, Anhong Guo, David A. Shamma · 2016 · Proceedings of the 2016 CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA 2016) · doi:10.1145/2851581.2892588
Summary
This paper provides a candid examination of why academic research papers, delivered almost exclusively as PDFs, remain largely inaccessible to people with disabilities. The authors — who have spent years actively trying to improve the situation — trace the problem from PDF's origins as a digital replacement for fax machines through to its current status as the default academic publication format. They explain that making a PDF accessible requires adding structural tags (a document object model for reading order, headings, and figures), alternative text for images, and language metadata — none of which the format was originally designed to support. The paper evaluates the three main authoring tools used by researchers: Word for Windows (which can produce tagged PDFs but requires Adobe Acrobat for finishing), Word for Mac (which exports none of the accessibility metadata to PDF), and LaTeX (which has only incomplete implementations for accessible output). An automated accessibility check of papers from W4A, ASSETS, and CHI conferences found that even in accessibility-focused venues, compliance was remarkably low — only 26.8% of CHI 2014 papers included any document tags, and only 10.9% specified a language. The authors describe four escalating strategies they tried: writing guidelines, volunteering to remediate papers, enforcing requirements as program chair, and being paid to do remediation.
Key findings
At CHI 2015, the authors volunteered to make papers accessible and remediated 25 of 486 PDFs, each taking 20-30 minutes primarily due to writing alternative text and dealing with Acrobat bugs. When UIST 2015 requested alt text from authors, only 53 of 70 papers responded, and roughly 60% of submitted alt text simply duplicated the caption rather than providing meaningful descriptions — less than 25% was descriptive enough to convey the image's purpose. At ACM ASSETS, where accessibility was enforced as a requirement, all 30 accepted papers were made accessible through manual checking and author revisions. Professional PDF remediation quotes ranged from - per document for full accessibility, or about per page for basic tagging without alt text. The paper concludes that the PDF format itself is fundamentally the problem — designed for printing, not reading — and proposes HTML/CSS-based publication formats as the only long-term solution.
Relevance
This paper is essential reading for anyone involved in digital publishing, academic conferences, or document accessibility. It exposes a painful irony: even the research community that studies accessibility fails to make its own output accessible. The practical insights are valuable for organizations facing similar challenges — the escalating strategies of asking, volunteering, requiring, and paying map directly to how any organization might approach document accessibility at scale. The finding that author-submitted alt text is overwhelmingly inadequate (mostly caption duplicates) highlights the need for training and guidance on writing effective image descriptions. For practitioners, the paper reinforces that PDF remediation is expensive and fragile, and that choosing accessible formats from the start (born-accessible content) is far more sustainable than retrofitting. The argument for moving academic publishing to HTML is even more compelling a decade later, as web standards have continued to mature.
Tags: PDF accessibility · document accessibility · academic publishing · screen readers · alternative text
Standards referenced: WCAG 2.0 · PDF/UA · Section 508