Say It All: Feedback for Improving Non-Visual Presentation Accessibility
Yi-Hao Peng, JiWoong Jang, Jeffrey P. Bigham, Amy Pavel · 2021 · Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems (CHI '21) · doi:10.1145/3411764.3445572
Summary
This paper addresses the widespread problem of inaccessible slide-based presentations for blind and visually impaired audiences. When presenters fail to verbally describe the visual content on their slides — text, images, diagrams, graphs, and videos — audience members who cannot see the slides miss essential information. The researchers first conducted a formative analysis of 90 existing presentation videos spanning TED talks, seminars, and classroom lectures across five academic domains. They coded 610 visual elements across 269 slides for how well presenters verbally described them. This analysis revealed that 72% of informative visual elements were insufficiently described, with 27% of slide elements receiving no verbal description at all. Media elements fared worst: images on slides were either not described at all or only briefly referenced over 50% of the time. Based on these findings, the researchers developed Presentation A11y, a Google Slides Chrome extension that provides both real-time and post-presentation accessibility feedback. The real-time interface augments the presenter view by transcribing speech and highlighting slide elements (text and images) as the presenter verbally covers them, allowing presenters to glance at what they have and have not described. The post-presentation interface provides an overview of coverage percentages per slide, element-level feedback on what was missed, and actionable suggestions such as removing undescribed elements or adding descriptions to speaker notes.
Key findings
In a study with 16 presenters using their own slides, Presentation A11y significantly improved every measured metric. Presenters described a higher percentage of slide words when using real-time feedback (57% vs 46% with the traditional interface, p < 0.05). Text coverage scores improved significantly (4.24 vs 4.03, p < 0.05), as did media coverage scores (3.61 vs 2.88, p < 0.05). Most strikingly, presenters identified 3.26 times more accessibility improvements to make using the post-presentation feedback (2.25 vs 0.69 changes, p < 0.001). All 16 participants reported that post-presentation feedback improved their awareness of accessibility, and all said they would use the tool in the future. Presenters rated their own presentations as significantly more non-visually accessible when using the real-time interface (5.06 vs 4.00 on a 7-point scale, p < 0.05). Common patterns of inaccessibility identified in the formative study included "showing instead of telling" (displaying images without description), missing visual context (describing graph values but not axes or trends), and using pronouns like "this" and "here" to reference visual content.
Relevance
This research provides concrete, data-driven evidence of how pervasive presentation inaccessibility is — even among educated presenters giving prepared talks. The finding that 72% of visual elements are insufficiently described quantifies a problem that accessibility professionals have long observed anecdotally. Presentation A11y demonstrates that integrating accessibility feedback directly into content creation tools — rather than relying on separate guidelines or checklists — can meaningfully change presenter behavior. This has implications for platform providers like Google, Microsoft, and Apple who could embed similar feedback into their presentation software. The tool is also notable because it benefits the presenter's audience in real-time rather than requiring post-hoc remediation. For organizations striving to make meetings, lectures, and conferences more inclusive, this research makes a compelling case for automated accessibility feedback as a standard feature of presentation tools.
Tags: presentation accessibility · slides · audio description · blind and low vision · real-time feedback · authoring tools · non-visual access
Standards referenced: WCAG 2.1