A new mixed-methods study of 128 caregivers found that an AI chatbot called Daphne, designed to flag unmet health-related social needs and point users toward community resources, earned high marks for ease of use and trust. The findings, published in PLOS Digital Health, suggest automated tools can meaningfully engage caregivers facing serious financial and material hardship rather than just healthcare professionals or tech-savvy users.
Who participated and what they needed
Nearly all participants in the study, 98 percent, reported at least one unmet health-related social need. Food insecurity affected 76 percent and financial limitations affected 75 percent of respondents. The group was predominantly female (68 percent), largely aged 30 to 49 (71 percent), and included both White (44 percent) and Black or African American (36 percent) caregivers. Most relied on public insurance, with 83 percent covered by Medicaid or Medicare, and 78 percent were caring for a child with special healthcare needs. This profile reflects a population researchers describe as facing substantial socioeconomic strain, making the chatbot's performance with this group especially relevant to public health practice.
How Daphne performed
Conversations with Daphne averaged three minutes and about 20 turns per session. Across 1,523 total dialogue exchanges, the chatbot correctly identified user intent 99 percent of the time, concentrating on financial, housing, and food-related needs. Caregivers rated the tool highly on the System Usability Scale, with a mean score of 84.7, while reporting low mental effort on the NASA Task Load Index (6.8). Satisfaction was strong, reflected in a Net Promoter Score of 8.0, and trust registered a mean of 4.1 on the study's scale. Researchers noted only slight usability dips among participants with higher education levels or lower incomes, indicating the tool performed consistently across most demographic groups.
Where the tool added the most value
Caregivers dealing with transportation barriers or utility disruptions reported even higher trust and satisfaction scores than the overall sample, suggesting the chatbot was particularly useful for people confronting concrete structural obstacles to care. The study's authors frame this as evidence that conversational AI can supplement, rather than replace, existing social care screening conducted in clinical settings.
Limits and next steps
This research was cross-sectional, capturing a single snapshot of user experience rather than tracking outcomes over time. The authors say future work should examine longitudinal impact, such as whether caregivers who use Daphne actually connect with resources and experience improved circumstances, and assess real-world implementation challenges across different healthcare systems. The study offers encouraging usability data but does not yet demonstrate long-term effectiveness.