UX and accessibility for research software

Research software succeeds when researchers can actually use it; UX is not cosmetic polish but part of software quality. This skill treats usability as an engineering practice, not cosmetic polish.

Practice

• Know the disciplines: UX research (understanding users and tasks), interaction design, visual design and accessibility are different jobs - be explicit about which one a request needs.
• Start from users and tasks, not features: apply Design Thinking - empathize with the target researchers, define the task, ideate, prototype cheaply (Figma or Miro before code), test with real users.
• Scope deliberately: an MVP proves the workflow works; a Minimum Loveable Product adds the qualities that make researchers adopt it. Pick which one the project stage needs and say so.
• Evaluate against Nielsen's usability heuristics (visibility of status, match with the users' world, error prevention and recovery, consistency, minimalism); they apply to CLIs and notebooks as much as GUIs - clear --help, sane defaults and good error messages are UX.
• Accessibility is a requirement, not an extra: color-contrast, keyboard navigation and screen-reader support for web tools; readable output and terminal-friendly formats for CLIs.
• Usability findings become issues like any other defect; retest after fixing (pair with rseng-testing for automated regression of CLI/API ergonomics where possible).

Accessibility, concretely

• Web-facing research tools target WCAG (the W3C accessibility standard; level AA is the common bar and often a legal one for publicly funded services): text alternatives for non-text content, sufficient contrast, full keyboard operability, labeled form controls, and structure a screen reader can navigate (headings, landmarks, table headers).
• Test it, do not assume it: automated checkers (axe-class tools in CI) catch the mechanical failures; a keyboard-only walk-through and a screen-reader pass catch the rest. Accessibility findings are defects with severities, tracked like any other (rseng-project-tracking).
• Scientific outputs have accessibility too: colorblind-safe and perceptually honest palettes (rseng-scientific-visualization), alt text that states what the figure SHOWS ("temperature rises sharply after 1980"), not what it is ("a line chart"), and data tables as alternatives to figure-only results (rseng-science-communication).
• Documentation accessibility: plain language, defined jargon, meaningful link text, heading hierarchy - the same qualities that serve newcomers and non-native speakers (rseng-documentation, rseng-trainer's calibration).

CLI usability, concretely

Most research tools are CLIs, and CLI UX is a codified craft (the Command Line Interface Guidelines distill it):

• --help that teaches: usage, an actual example, and the most-used flags first; man-page or docs parity (rseng-documentation).
• Errors that help recovery: say what failed, on which input, and what to try - "config key 'stations' missing (stations.json)" beats a traceback (rseng-defensive-coding's fail-loud, made humane).
• Sane defaults, explicit overrides; standard conventions respected (exit codes for CI - rseng-ci-cd; stdout for data, stderr for messages so pipes work; --version; no color when not a TTY).
• Progress for long operations (rseng-hpc-computing jobs excepted - log lines there), and quiet/verbose flags because both scripting and debugging are real uses.
• Test the ergonomics: golden-file tests on --help and error messages keep promised UX from rotting (rseng-testing).

Working with this skill

The generated references.md beside this file lists the source material and pointers:

• references.md - verified Learn more pointers

Learn more (verified):

• https://www.w3.org/WAI/standards-guidelines/wcag/ - WCAG
• https://www.a11yproject.com - the A11y Project
• https://clig.dev - Command Line Interface Guidelines

Related skills

Check whether any of these applies before moving on:

• rseng-defensive-coding - humane fail-loud error messages
• rseng-documentation - docs accessibility and help parity
• rseng-science-communication - accessible slides and figures
• rseng-scientific-visualization - colorblind-safe honest palettes
• rseng-testing - golden-file tests on CLI ergonomics
• rseng-user-support - error-message questions reveal UX gaps