"I Don't Trust Any Professional Research Tool": A Re-Imagination of Knowledge Production Workflows by, with, and for Blind and Low-Vision Researchers
Omar Khan, JooYoung Seo · 2026 · Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI '26) · doi:10.1145/3772318.3791242
Summary
This CHI 2026 paper is an autoethnographically-grounded, mixed-methods study of how blind and low-vision (BLV) researchers actually do research inside an ecosystem of tools built with sighted workflows in mind. Written by two BLV researchers (one totally blind, one low vision), it uses Creswell's explanatory sequential mixed-methods design: a cross-sectional survey of 57 BLV researchers recruited through NFB, AFB, ACB, UW DOT-IT, Blind Academics, ICEVI, and the World Blind Union, followed by 15 semi-structured interviews. Survey items were grounded in Engeström's third-generation activity theory (subject, tool, object, rules, community, division of labor), the Technology Acceptance Model, and the User Experience Questionnaire. Open-ended responses and interview transcripts were coded in ATLAS.ti using reflexive thematic analysis plus open/axial/selective coding, then mapped back to activity-theory constructs to reframe barriers as systemic contradictions rather than individual limitations. Participants spanned 14 disciplines — social sciences, law, public health, physics, music education, library sciences, queer and gender studies, meteorology, and more — with 2 to 40+ years of research experience. The paper foregrounds researcher positionality, community accountability (partnership with advocacy organizations, BLV peer review throughout), and a deliberate refusal to frame accessibility failures as individual accommodation problems.
Key findings
BLV researchers systematically substitute general-purpose tools (Word, Excel, Google Docs) for specialized research software (NVivo, ATLAS.ti, MAXQDA, SPSS) because commercial research tools fail basic accessibility: NVivo was formally declared unfixable by its developer when P3 asked for accessibility support. Nearly one-fifth of participants (17%) delegate literature review or visual output evaluation to sighted colleagues or use AI retrieval to avoid fatigue. Data analysis and visualization ranked the most inaccessible stage (mean 2.58/5), followed by literature access (2.80/5). 'Update anxiety' is pervasive — a working screen-reader configuration can break within an hour of a vendor update, driving P1's declaration: 'I don't trust any technology or any professional research tool at all.' BLV researchers shoulder documented hidden labor: maintaining trust networks to monitor which tools still work, verification labor to fact-check LLM outputs (ChatGPT used as a PDF reader, with hallucination a major concern), and redistributing collaborative tasks around inaccessibility rather than expertise. Professional consequences include lost networking at conferences, credibility challenges (having to 'prove qualifications twice or thrice' per P9), a denied best-presentation award attributed to visually unappealing slides, and — most starkly — P13 abandoning a dissertation two classes short. Less than 3% of scholarly PDFs meet basic accessibility standards.
Relevance
This paper belongs on the desk of anyone building research software, running a graduate program, organizing an academic conference, or setting promotion criteria. Its central argument — that accessibility must be re-conceptualized as fundamental research infrastructure rather than individual accommodation — carries concrete recommendations for four audiences: tool developers (accessibility-first design, stability guarantees, transparent accessibility roadmaps), institutions (treat accessibility labor as hidden labor worth funding and counting toward promotion, plan for dual-career BLV hires), conferences and publishers (separate content from aesthetic in evaluation; refuse inaccessible figures the way we refuse methodological flaws), and the broader community (dismantle ableist assumptions about who can contribute). Limitations include a predominantly U.S. sample recruited through advocacy channels (likely over-representing resource-aware researchers), a cross-sectional snapshot, focus on digital desk-based research rather than lab or fieldwork, and deliberate scope around BLV rather than intersecting disabilities. The activity-theory framing is particularly useful for practitioners who want a vocabulary beyond 'inaccessible tool' to explain why BLV colleagues are structurally disadvantaged even when each individual barrier has a workaround.
Tags: meta-research · blind and low vision · research workflows · academic accessibility · activity theory · mixed methods · reflexive thematic analysis · research tools · screen readers · ableism · epistemic injustice · qualitative analysis software · AI accessibility · scholarly publishing · disability studies
Standards referenced: WCAG · Universal Design