Reframing multilingual accessibility — from a language problem into a systems design problem.
How fragmented workflows and broken trust disrupt comprehension on digital libraries — a UX research project on where multilingual readers actually lose understanding, and what could realistically change.
Role
UX Researcher
Timeline
3 months
Platform
Open Library
Methods
Interviews · Survey design · Affinity mapping
The reading workspace multilingual students actually live in: the book, plus every tool they open just to understand it.
Quick facts
Role: UX Researcher — interviews, survey design, synthesis, recommendations
Methods: 8 semi-structured interviews, a survey across the UMSI community, affinity mapping
Deliverables: Workflow analysis, three key insights, three design recommendations evaluated against legal and technical constraints
The reframe
The problem
Research
Insights
Recommendations
The reframe
It wasn't a language problem. It was a systems problem.
At the start, the assumption was simple: multilingual readers struggle on digital libraries because translation tools aren't good enough. But interviews with international students and academic staff told a different story. Readers weren't blocked by language alone — they were blocked by the experience of reading itself. Understanding a single paragraph often meant switching between four or five tools, losing context with every lookup, and questioning whether any translation could be trusted.
Fragmented workflows, invisible tools, and broken trust disrupt comprehension far more than vocabulary ever could.
Goal: identify where multilingual readers lose comprehension during digital reading — and which design changes could reduce friction without violating legal or technical constraints.
The problem
Most digital libraries assume English fluency by default.
Translation tools exist, but they're hidden, fragmented, or disconnected from the reading experience. So multilingual readers rely on external workarounds that interrupt flow, increase cognitive load, and undermine trust. Here's what reading one difficult paragraph actually looks like:
1 — Encounter unfamiliar text. Reader finds a word or phrase they don't understand.
2 — Switch to a translation tool. Google Translate, a dictionary app, or browser search.
3 — Lose context. Forget where they were in the original text.
4 — Question accuracy. Check multiple sources to verify the translation.
5 — Repeat. The cycle continues for every unfamiliar element.
The result: comprehension breaks down not from lack of language knowledge, but from cognitive overload and workflow fragmentation.
Research approach
I studied workflows, not feature usage.
Eight semi-structured interviews with multilingual students and academic support staff, a survey across the UMSI community to capture broader patterns, and affinity mapping to identify themes across experiences and workflows.
Research participants: multilingual students and the staff who support them.
Key insights
Three insights reframed the work.
01 — Comprehension breaks when the workflow breaks.
"I switch between Google Translate, my dictionary app, and Wikipedia just to understand one paragraph."
Why it matters: fragmentation actively disrupts understanding — even when readers are fluent.
02 — Users want unity, not more tools.
"I just want everything in one place."
Why it matters: accessibility is shaped by integration, not feature count.
03 — Unreliable translation erodes trust.
"I don't trust the translation — I double-check everywhere."
Why it matters: when users distrust the system, they slow down, reread, and reconstruct meaning manually.
Design recommendations
Three changes that could realistically ship.
01 — Improve visibility of existing translation tools.
Move language support into primary navigation and reduce the steps required to access it.
Expected impact: lower cognitive load, higher discoverability.
02 — Introduce contextual AI as a reading assistant.
Support comprehension through optional, privacy-preserving AI explanations — without generating or altering original content.
Expected impact: increased trust without legal or ethical risk.
03 — Strengthen feedback loops with users.
Enable structured, lightweight reporting so multilingual readers can surface issues and influence improvements.
Expected impact: shift users from workaround builders to co-improvers.
Constraints and ethics
Designed within real constraints, not ideal ones.
All recommendations were evaluated within real-world constraints — copyright (respect for original content and attribution rights), privacy (user data protection and consent mechanisms), and academic integrity (maintaining standards and transparency). Rather than proposing idealized solutions, this project focused on what could realistically scale within existing legal and technical frameworks.
Reflection
This project reframed accessibility as a systems design problem rather than a language problem — a perspective that applies far beyond digital libraries.
Thank you for reading. For questions or collaboration opportunities, feel free to reach out.