Foreign Domestic Workers’ Perspectives on an LLM-Based Emotional Support tool for Caregiving Burden | Shin Shoon Nicholas Teng, Kenny Tsu Wei Choo | ResearchPod

Transcript: Foreign Domestic Workers’ Perspectives on an LLM-Based Emotional Support tool for Caregiving Burden

Alex: Welcome to another episode of ResearchPod.

Sam: Today we're looking at a study on how some caregivers use AI chat tools to handle the emotional side of their work. The paper is called "Foreign Domestic Workers’ Perspectives on an LLM-Based Emotional Support Tool for Caregiving Burden," by researchers at the Singapore University of Technology and Design.

Alex: So this is about women who live with families to care for elderly people, right? And the big question is whether talking to an AI chatbot helps them deal with the tough feelings that come up?

Sam: Yes, exactly. These workers, known as foreign domestic workers or FDWs, often feel a heavy emotional load from tasks like helping frail older adults with daily needs. Things get harder because they're far from home, speak limited English, work long hours with few days off, and can't easily share frustrations due to boss-employee dynamics.

Alex: That paints a clear picture of isolation. Imagine being in a stranger's home all day, managing someone with memory issues like dementia, no close friends nearby to talk to, and worrying that complaining might cost your job.

Sam: Precisely. The emotional caregiving burden here isn't just the physical work; it's the stress from feeling alone with those feelings, shaped by language gaps and power imbalances in their employment. Traditional studies focus on family caregivers, but this group—FDWs in places like Singapore, where aging populations rely on them heavily—has been overlooked.

Alex: Human support sounds risky for them. Does the paper suggest why an AI chatbot might step in where people can't?

Sam: Chatbots like Sophia use advanced language systems to respond in a private, always-available way, without the stigma or hierarchy of talking to employers, family back home, or even friends. They're non-clinical—just a conversational partner that listens empathetically and reflects back what you say, helping you feel heard. For FDWs, this could lower barriers to opening up about daily stresses.

Alex: Okay, so it's like a digital ear that doesn't judge or get tired. What makes the emotional part so burdensome specifically for these caregivers?

Sam: Caregiving burden builds when the demands—like unpredictable behaviors from elders with health issues—mix with no outlet to process them emotionally. For FDWs, long hours in a non-native language and employment rules amplify that, leaving them with uncertainty and bottled-up frustration.

Alex: That tracks with the isolation you mentioned. So the researchers talked to actual FDWs about using this tool?

Sam: They did a small exploratory study with seven FDWs in Singapore, all with eldercare experience and some mild burden from a quick survey. Each had a one-hour session: they recalled a tough caregiving incident, chatted with Sophia while thinking aloud, and reflected after. From transcripts, they pulled themes on how it felt to use it.

Foreign Domestic Workers’ Perspectives on an LLM-Based Emotional Support tool for Caregiving Burden

Shin Shoon Nicholas Teng, Kenny Tsu Wei Choo

Apr 7, 2026·5:06·arxiv·Original Paper

Foreign Domestic Workers (FDWs)LLM-Driven ChatbotPsychological SafetyLinguistic AccessibilityMultifunctional Appropriation

Listen

5 min

Briefing

Abstract

Foreign Domestic Workers (FDWs) play a central role in home-based eldercare yet often experience substantial emotional caregiving burden shaped by linguistic barriers, social isolation, and limited access to support. While caregiving burden has been extensively studied among familial caregivers, little is known about how FDWs engage with emotional support technologies. We present an exploratory qualitative study of how FDWs in Singapore interact with a Large Language Model (LLM)-driven chatbot as an everyday, non-clinical form of emotional support. Through interviews and guided chatbot interactions, we conducted an inductive thematic analysis of participants' experiences. We identify three design-relevant themes: chatbots were experienced as psychologically safe and emotionally validating; they supported linguistic accessibility by accommodating imperfect and fragmented language; and they were appropriated as multifunctional resources for reassurance, guidance, and companionship. We discuss implications for designing LLM-driven emotional support tools that foreground psychological safety, accessibility, and flexible appropriation.

Overview

This paper explores how Foreign Domestic Workers (FDWs) in Singapore—migrant caregivers who handle much of home-based eldercare—experience an LLM-powered chatbot named Sophia as a tool for managing emotional caregiving burden. FDWs face unique stressors: linguistic barriers (many have limited English), social isolation from being far from home, and employment constraints that limit access to formal support. Unlike studies on family caregivers, this qualitative research with 7 FDWs uses interviews and guided chatbot sessions to uncover how non-clinical AI support fills gaps in their lives. The findings highlight why LLMs are promising for vulnerable groups: they're private, always available, and non-judgmental. Key themes reveal design principles for emotional support tools that prioritize safety, accessibility, and flexibility.

Psychological Safety as a Non-Judgmental Space

FDWs described the chatbot as a 'safe space' where they could vent frustrations without fear of judgment, retaliation, or professional repercussions. In their constrained lives—living with employers, facing potential deportation for complaints—human support often carries risks. Sophia provided emotional validation through empathetic, reflective responses, making FDWs feel 'heard' and reducing isolation. This aligns with HCI principles where AI's lack of social baggage creates psychological safety, enabling free disclosure of raw emotions like stress from erratic elder behaviors or family separation. Intuition: Humans hesitate due to stigma; AI doesn't gossip or report back.

Linguistic Accessibility for Non-Native Speakers

A standout strength was the chatbot's tolerance for imperfect language—broken English, fragments, or non-native phrasing. FDWs with limited proficiency still received meaningful support, as the LLM parsed intent over grammar. This overcomes a major barrier in Singapore's multilingual context, where formal services demand fluent English. The paper shows how LLMs' robustness to 'messy' input democratizes access, turning potential frustration into relief. Build intuition: Traditional chatbots fail on accents/errors; LLMs excel by modeling diverse language patterns from vast training data.

Multifunctional Appropriation Beyond Emotions

FDWs didn't just seek therapy—they repurposed Sophia for reassurance (e.g., 'Am I doing this right?'), companionship (daily chats combating loneliness), guidance on caregiving dilemmas, and resolving uncertainties like medical decisions. This 'appropriation' shows users adapt tools to holistic needs, not rigid categories. Designers should build flexibility, avoiding over-specialization. Key insight: In resource-scarce lives, one tool serves many roles, amplifying value.

Design Implications and Broader Impact

The study urges LLM tools foregrounding safety (neutral tone, privacy), accessibility (multilingual, error-tolerant), and multifunctionality (open-ended prompts). It bridges HCI gaps for non-familial caregivers, informing ageing societies reliant on migrant labor. Why it matters: As populations age, FDWs are invisible pillars; AI can humanely lighten their load without replacing human connection. Limitations include small sample (n=7), but rich themes guide future work like scaling to other migrant groups.

Foreign Domestic Workers’ Perspectives on an LLM-Based Emotional Support tool for Caregiving Burden: Research Paper Podcast