How This Wiki Works¶
This wiki is GiveCare's public reference. It holds caregiver resources and the reasoning behind GiveCare's model, so partners and researchers can check that reasoning.
A reader should be able to see what GiveCare claims, where the claim comes from, how current it is, and what it does not show.
What the wiki contains¶
There are two kinds of pages.
| Area | Purpose | Examples |
|---|---|---|
| Caregiver resources | Help a caregiver or navigator find a next step | benefits, guides, organizations, condition pages |
| Research and methodology | Show GiveCare's reasoning | six-domain framework, crisis routing, SMS accessibility, benefits discovery, InvisibleBench |
Methodology pages explain reasoning and leave operating detail out.
How pages are created¶
A new page starts from one of four things:
- a public program, benefit, organization, or guide a caregiver can use
- research or public data that changes how we understand caregiver needs
- a product or safety decision that outside readers should be able to follow
- a question from a partner, clinician, researcher, or caregiver that deserves a lasting answer
We write the smallest public version that answers the question: define the term, say why it matters, cite the strongest sources, and state what is still uncertain.
How we keep it current¶
Pages change when research, a program, a product decision, or a partner question changes what we know. A change to one page often means revising its neighbors so definitions and citations still agree. When a newer source contradicts an older claim, we qualify or replace the older claim.
Automated checks look for broken links, missing citations, stale resource pages, and orphaned source notes. These checks flag problems for review.
How we use sources¶
Every significant factual claim should trace to a source. We prefer primary sources: government program pages, statutes, peer-reviewed research, official organization pages, and original evaluation materials.
The right source depends on the claim. A clinical construct calls for a peer-reviewed paper, and a local service for the page of the agency that runs it. Each page names its source type so a reader can weigh the claim.
When evidence is incomplete, mixed, adapted, or preliminary, the page says so.
How AI is used¶
AI helps draft and maintain these pages. It summarizes sources, checks cross-references, finds stale pages, and keeps related pages consistent. A page is admitted only after automated checks compare it with its cited sources. People set the standards and review any change to safety-relevant content.
We leave out internal prompts, private partner information, live operating metrics, and any implementation detail that would make a safety system easier to game.
What readers should expect¶
- Cited. Major factual claims link to a source page or an external reference.
- Plain. No internal context is needed to follow a page.
- Dated. Resource pages show their last review, and stale pages are flagged.
- Clear about limits. Pages separate validated evidence, adapted frameworks, operating experience, and hypotheses.
- Linked. Related concepts point to each other.
A page that reads like an internal note or a runbook gets rewritten for the public or moved out.