Open infrastructure for the caregiver research gap
Frameworks, benchmarks, and longitudinal data on a population that existing tools were not built to see.
Family caregivers are 63 million people in the United States. They are tracked inconsistently, measured late or not at all, and largely absent from the longitudinal data that shapes care policy and workforce programs.
GiveCare Labs exists to change that. Half of what we build is open source. The rest runs a live service designed to generate the kind of forward-looking, caregiver-centered data that does not yet exist at scale.
What the caregiver experiences
GiveCare is SMS-based support infrastructure for family caregivers. No app. No portal. Just text.
Chief of staff, not companion
GiveCare is not a chatbot. It does not ask how your day is going. It tracks where the caregiver is across six dimensions of their situation, connects dots over time, and surfaces what they may be eligible for before they think to ask.
Closing the action gap
Awareness is not enough. Caregivers often know they need something but cannot act on it. GiveCare moves from identification to action: eligibility screening, benefit navigation, application preparation, and community resource connection — specific to where the caregiver is in that moment.
Built for equity
SMS is a deliberate choice. Most caregivers do not have access to sophisticated AI tools. The channel is designed to reach the people who need support most, not the people most comfortable with technology.
Extends to professional caregivers
The framework maps directly to paid caregivers — home health aides, CNAs, direct support professionals — who face the same compounding pressures with less infrastructure around them.
What the data layer produces
The GiveCare Score is a single number that decomposes into six zones. What makes it useful for research is how it is collected and tracked over time.
Unit of measurement
Caregiver, not patient
Collection method
Daily EMA via SMS — ecological momentary assessment, forward-looking
Dimensions tracked
Six SDOH zones, rolling GiveCare Score, benefits eligibility flags
Population view
Anonymized aggregate — no PII at the employer or research level
Longitudinal
Continuous, not cross-sectional — score changes are traceable to specific zone shifts
Applicability
Unpaid family caregivers and professional caregivers
At the population level, employers and researchers see aggregate caregiver burden trends, zone-level shifts, and engagement patterns without access to individual data. The population dashboard was designed to give organizations something actionable — not just utilization numbers, but a window into where their caregiving population is under strain and what kind of support is being accessed.
Open projects
The frameworks and benchmarks are public. They are designed to be used, cited, and built on.
GC-SDOH-30
A validated 30-item framework that maps caregiver burden across six zones: social support, physical health and energy, housing and environment, financial strain, system navigation, and emotional wellbeing. Unlike generic SDOH tools, GC-SDOH-30 is built around the lived structure of caregiving, not the patient experience.
Read the frameworkInvisibleBench
The first open benchmark designed to evaluate AI safety across long-term caregiving relationships. Current benchmarks test single conversations. InvisibleBench tests across 3–20+ turn scenarios, identifying failure modes like crisis calibration drift, cultural othering, and attachment engineering that only emerge over time.
View benchmark resultsInterested in collaborating?
We are looking for research partners, pilot programs, and policy collaborators who want to work with live longitudinal caregiver data. If the frameworks are relevant to what you are building or studying, reach out.
Start a conversation