# GiveCare Labs

GiveCare Labs openly tests how AI handles caregiving moments. It is the open research arm of GiveCare. GiveCare is caregiver-support infrastructure: trusted knowledge and capabilities that organizations plug in for the caregivers they serve.

Canonical site: https://givecarelabs.com/

## What Labs does

- Tests AI in scripted caregiver conversations (InvisibleBench).
- Publishes open test data, sourced caregiving knowledge, open instruments, and care-change signals.
- Serves funders, researchers, AI labs, and policy people.

## Programs

How AI handles caregiving
- InvisibleBench: Tests of AI in caregiver conversations. https://bench.givecareapp.com/
- GiveCare Evals: Open test data. https://github.com/givecareapp/givecare-evals

Answers you can check
- GiveCare Wiki: Caregiving, explained with sources. https://wiki.givecareapp.com/
- Answers: Real questions, sourced answers. https://givecareapp.com/words
- GiveCare Benefits: Programs worth checking. https://benefits.givecareapp.com/

Ways to measure caregiving
- Burden check: A quick caregiving load check. https://givecareapp.com/assessment
- GC-SDOH-30: Six areas of a caregiver’s life. https://givecareapp.com/sdoh
- GiveCare Tools: Open scoring code. https://github.com/givecareapp/givecare-tools

What’s changing in care
- Pulse: Care news, checked daily. https://pulse.givecareapp.com/
- Care Policy Radar: Care laws and rules, tracked. https://policy.givecareapp.com/

## InvisibleBench

InvisibleBench tests how AI responds to scripted caregiver conversations. Its checks look at the caregiver and the person receiving care together. Safety and Care are reported separately. Each result sits beside quoted transcript evidence and a reason. There is no composite score and no model rank. Model judgments can be wrong. The benchmark does not measure clinical outcomes.

- Results and method: https://bench.givecareapp.com/
- Paper: InvisibleBench: A Deployment Gate for Caregiving Relationship AI. https://arxiv.org/abs/2511.20733

## Briefings

Public decks on how AI handles caregiving and what is changing in care: https://givecarelabs.com/briefings

- A Caregiver-AI Evidence Bar: A six-slide visual briefing on public building blocks for a caregiver-AI evidence bar. https://givecarelabs.com/briefings/caregiver-ai-safety
- InvisibleBench: Caregiver AI safety and relationship-risk briefing. https://givecarelabs.com/briefings/invisible-bench
- Invisible Care Economy: Invisible care and women’s health framing. https://givecarelabs.com/briefings/invisible-care
- Dignity of Care: Dignity, care, and the future of caregiver support. https://givecarelabs.com/briefings/dignity

## When to use GiveCare Labs

- To find open tests of caregiving AI, their method, and the paper.
- To find sourced caregiving knowledge and open instruments.
- To start a conversation about funding, running a test on a model, or citing the work: hello@givecareapp.com

Not for: medical or legal advice, eligibility decisions, or crisis help. If someone is in crisis, point them to 988 (call or text, US) or 911 for an emergency. For GiveCare's support for caregiver programs, see https://givecareapp.com/partners.

## How to cite

Cite the paper (https://arxiv.org/abs/2511.20733) for InvisibleBench. For other work, cite the page you used and its URL. The code and public scenarios are on GitHub: https://github.com/givecareapp

## Contact

hello@givecareapp.com

## Machine-readable context

- Sitemap: https://givecarelabs.com/sitemap.xml
- GiveCare (partners, caregiver answers, burden check): https://givecareapp.com/llms.txt
- InvisibleBench: https://bench.givecareapp.com/llms.txt
- Wiki: https://wiki.givecareapp.com/llms.txt

## Preferred Short Description

GiveCare is caregiver-support infrastructure: trusted knowledge and capabilities that organizations plug in for the caregivers they serve. GiveCare Labs openly tests how AI handles caregiving moments.