Didactopus/docs/learner-accessibility.md

Learner Accessibility

Didactopus should make the learner loop usable without assuming visual graph navigation or silent waiting on slow local models.

The current accessibility baseline is built on the graph-grounded learner session backend.

Current Outputs

Running:

python -m didactopus.learner_session_demo

now writes:

• examples/ocw-information-entropy-session.json
• examples/ocw-information-entropy-session.html
• examples/ocw-information-entropy-session.txt

What The Accessible Outputs Do

The HTML output is meant to be screen-reader-friendly and keyboard-friendly:

• skip link to the main content
• semantic headings
• reading-order sections for study plan, conversation, and evaluation
• grounded source fragments rendered as ordinary text instead of only visual diagrams
• deterministic learner-facing labels localized for supported output languages

The plain-text output is a linearized learner-session transcript that is suitable for:

• terminal reading
• screen-reader reading
• low-bandwidth sharing
• future text-to-speech pipelines

Why This Matters

Didactopus should help learners work with structure, not just with pictures and dashboards.

This is especially important for:

• blind learners
• screen-reader users
• learners on low-power hardware
• situations where audio or text needs to be generated locally

Relationship To The Roadmap

This is the accessibility baseline, not the endpoint.

Likely next steps:

• local text-to-speech for mentor, practice, and evaluator turns
• speech-to-text for learner answers
• explicit spoken structural cues
• text-first alternatives for more generated visualizations