What We Share Here¶
This wiki describes GiveCare's public model: the caregiver needs it is organized around, the evidence it relies on, the product decisions it can responsibly describe, and the limits of the work so far.
A reader should come away knowing what GiveCare claims and what still needs validation.
What belongs here¶
We publish what helps a caregiver, partner, clinician, funder, or researcher understand:
- what GiveCare helps caregivers do
- why a construct, score, safety layer, or workflow exists
- which evidence or public source supports a claim
- what is externally validated, adapted, original, or provisional
- what the known limits are
- how to use a public resource safely
Examples that belong on the public wiki:
- why GiveCare uses a six-domain caregiver model
- why caregiver social needs differ from patient social needs
- why an assessment was chosen or adapted
- which parts of a score are externally validated and which are GiveCare's own
- why crisis routing is layered
- why SMS is the main channel
- what InvisibleBench evaluates
- what a construct should and should not be used for
What usually stays out¶
Some information matters for running the product and still stays out. We leave out anything that would:
- disclose private partner or caregiver information
- expose live operating performance before it is ready to share
- make a safety or eligibility system easier to game
- suggest the methodology pages are the full operating record
- add implementation detail a reader does not need to understand the claim
Examples that usually stay out:
- live usage, retention, response, or conversion metrics
- partner-specific results without approval
- private incident reviews
- exact prompt text
- routing logic detailed enough to reproduce or game production behavior
- tuned thresholds or coefficients while they are still changing or sensitive
Publish the rationale¶
When a topic matters to the method and its live setting is sensitive, the public page explains:
- the construct
- the reasoning behind it
- the direction of the decision
- the evidence and caveats
- the limits of the current implementation
A page can explain why caregiver stress or social support carries extra weight and leave out the live coefficients. The same goes for follow-up after a sharp decline in caregiver state: the page gives the reason and omits the production threshold.
When in doubt, explain the idea and withhold only the sensitive setting.
Disclosure labels¶
Pages that document an assessment, score, classifier, or framework label the kind of evidence behind it:
| Label | Meaning |
|---|---|
| Externally validated | Supported by published validation outside GiveCare |
| Adapted from validated frameworks | Built from established instruments or frameworks, but modified for GiveCare's context |
| GiveCare original | Designed by GiveCare, with no published instrument behind it |
| Operationally tested | Used in practice or internal evaluation, but not externally validated |
| Not yet externally validated | The construct is still provisional |
The labels let a reader tell a known instrument from an adaptation, a product heuristic, or a hypothesis still under study.
Page pattern for core constructs¶
A page about a score, assessment, framework, or safety layer follows this outline:
- What this is
- Why it exists
- Evidence base
- What is original here
- Validation status
- Known limits
- Public implementation note, if needed
Editorial rubric¶
Before publishing a claim or detail, ask:
- Does this help an outside reader judge whether the system rests on evidence and is built responsibly?
- Would publishing it expose private information, sensitive operations, or logic that could be gamed?
- Can the underlying idea be explained at a safer level of abstraction?
Leave the detail out if the first answer is no. When the second answer is yes, publish the rationale and withhold the sensitive part.
Decision table¶
| Question | Public wiki | Controlled sharing | Not public |
|---|---|---|---|
| Why this construct exists | Yes | - | - |
| Theory and reference research | Yes | - | - |
| Validation posture and caveats | Yes | - | - |
| Public safety architecture | Yes | - | - |
| Production usage metrics | No | Sometimes | Yes |
| Partner-specific outcomes | No | Sometimes | Yes |
| Exact prompts and routing details | No | Rarely | Yes |
| Exact tuned thresholds or coefficients | Sometimes, if stable and non-sensitive | Sometimes | Yes |
| Incident review details | No | Sometimes | Yes |
What this means for measurement pages¶
A page about the GiveCare Score, the caregiver social-needs assessment, or a similar construct publishes four things: the construct definition, the research and design rationale, the validation status, and the limits with the intended interpretation. It does not claim to hold every operating metric.