Caregivers Report High Usability and Trust When Using Social Support Chatbot

In a cross-sectional mixed-methods evaluation, caregivers facing unmet health-related social needs reported high usability and trust when interacting with Daphne, an artificial intelligence chatbot engineered to identify social needs and recommend resources. The study demonstrated that digital social care tools can successfully engage vulnerable groups experiencing substantial socioeconomic strain. Across quantitative metrics and qualitative conversation transcripts, participants found the conversational platform straightforward to navigate, with the system delivering personalized assistance during brief sessions without imposing significant cognitive burden on users seeking community services.

Unmet health-related social needs significantly impact overall health outcomes, yet connecting individuals to existing community resources remains difficult. While screening tools exist across healthcare environments, providing real-time, personalized navigation remains challenging. Chatbots offer a potential solution for personalized navigation; however, their practical usability and trustworthiness, particularly among populations with high needs, require further investigation to determine whether automated systems can engage vulnerable individuals.

To evaluate Daphne, researchers conducted a mixed-methods study involving 128 caregivers. Nearly all participants (98 percent) reported unmet health-related social needs, including food insecurity (76 percent) and financial limitations (75 percent). Demographically, 68 percent of participants were female, 71 percent were aged between 30 and 49 years, 44 percent were White, and 36 percent were Black or African American. In addition, 83 percent of participants relied on Medicaid or Medicare coverage, and 78 percent cared for a child with special healthcare needs.

Interactions with the chatbot averaged three minutes and 20 conversational turns per session. Across 1,523 total dialogue turns, Daphne achieved a 99 percent intent accuracy rate, focusing primarily on financial, housing, and food needs. The research team collected data on usability, task load, satisfaction, and trust alongside qualitative feedback and conversation transcripts. Regression and pairwise analyses were then conducted to explore associations between user demographics, specific health-related social needs, and overall user experience metrics.

Caregivers reported high usability, recording a mean System Usability Scale score of 84.7, along with low task load, reflected by a NASA Task Load Index score of 6.8. User satisfaction was high, yielding a Net Promoter Score of 8.0, and trust was similarly high, with a mean score of 4.1. Usability and trust remained generally consistent across demographic groups, though researchers detected slight usability decreases among respondents with higher education levels and those in lower income categories.

Notably, certain health-related social needs, specifically transportation barriers and utility disruptions, correlated with higher trust and satisfaction scores, suggesting that the chatbot provided particular value for users confronting structural barriers. The authors concluded that Daphne demonstrated strong usability and trust across diverse groups, supporting digital social care practices and providing empirical evidence that chatbots can engage vulnerable populations. Future work, they emphasized, should focus on evaluating longitudinal impact and practical effectiveness to improve accessibility and real-world implementation.

Subject of Research: Usability and user experience of an AI chatbot for health-related social needs

Article Title: Evaluating user experiences with an AI chatbot for health-related social needs: A cross-sectional mixed methods study

Article References: Sezgin, E., Jackson, D. I., Hussain, S.-A., Kocaballi, A. B., Richardson, C., Skeens, M., Fosler-Lussier, E., Ewing, A. P., Donneyong, M., & Pai, A. (2026). Evaluating user experiences with an AI chatbot for health-related social needs: A cross-sectional mixed methods study. PLOS Digital Health, 5(10), e0001789. https://doi.org/10.1371/journal.pdig.0001789

Image Credits: AI Generated

DOI: 10.1371/journal.pdig.0001789

Keywords: artificial intelligence, chatbots, health-related social needs, caregiver support, digital health, social care, system usability scale, public health, Evaluating, user, experiences, chatbot