arxiv: 2604.05448 · v1 · submitted 2026-04-07 · 💻 cs.HC

Recognition: no theorem link

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

Shin Shoon Nicholas Teng , Kenny Tsu Wei Choo (Singapore University of Technology , Design)

Authors on Pith no claims yet

Pith reviewed 2026-05-10 19:27 UTC · model grok-4.3

classification 💻 cs.HC

keywords foreign domestic workersLLM chatbotemotional supportcaregiving burdenpsychological safetylinguistic accessibilityqualitative studySingapore

0 comments

The pith

Foreign domestic workers in Singapore experienced an LLM chatbot as psychologically safe, linguistically accessible, and useful for reassurance and companionship during caregiving.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines how foreign domestic workers, who provide much of Singapore's home eldercare and often face emotional strain from isolation and language differences, engaged with an LLM-driven chatbot for non-clinical emotional support. Through guided sessions and follow-up interviews, the authors conducted a thematic analysis revealing three patterns in participants' responses. The chatbot was viewed as a safe space for validation, it handled imperfect English without judgment, and workers adapted it for multiple needs beyond strict emotional venting. These findings point to practical ways AI tools might address support gaps for this workforce without replacing professional care.

Core claim

Through inductive thematic analysis of guided interactions and interviews, the study establishes that FDWs experienced the LLM chatbot as psychologically safe and emotionally validating, that it supported linguistic accessibility by accommodating imperfect and fragmented language, and that participants appropriated it as a multifunctional resource for reassurance, guidance, and companionship.

What carries the argument

Inductive thematic analysis applied to FDWs' guided sessions with an LLM chatbot and their post-use interview accounts, identifying patterns of safety, accessibility, and flexible appropriation.

Load-bearing premise

The reported experiences from a small group of FDWs in guided sessions and interviews reflect their genuine everyday perspectives without notable influence from the research setting or pressure to give positive answers.

What would settle it

A follow-up study with more participants using the chatbot unguided over weeks that finds substantially different themes or clear evidence of social desirability bias in how they describe the tool.

read the original 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.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This exploratory study surfaces three usable design themes for LLM chatbots with foreign domestic workers but rests on thin evidence from guided sessions and interviews.

read the letter

The paper's core contribution is a set of three themes from FDWs in Singapore: the chatbot felt psychologically safe, handled broken English without friction, and got used for reassurance, advice, and company. That focus on this specific population is new compared to the usual work on family caregivers, and the themes line up with real design needs around language and isolation. The inductive analysis is straightforward and the implications section stays practical without overclaiming impact on broader mental health tech. Credit to the authors for picking an understudied group and keeping the write-up short and readable. The soft spots are the usual ones for this style of work. The abstract and stress-test note both point to a small sample with guided chatbot use followed by interviews, and nothing in the description shows member checks, anonymous follow-ups, or direct comparison of chat logs to what people said later. That leaves room for social desirability or researcher presence to shape the reported safety and appropriation. Without those details the themes are plausible but not yet strongly grounded. This paper is for HCI people working on caregiving tools or accessible chatbots. A reader already familiar with qualitative methods in vulnerable populations will get the most out of the themes and can judge the limits themselves. It is worth sending to peer review because the population angle is fresh and the design points are concrete, even if the methods section will need expansion and the limitations discussion sharpened before publication.

Referee Report

2 major / 2 minor

Summary. The manuscript reports an exploratory qualitative study of how foreign domestic workers (FDWs) in Singapore engage with an LLM-driven chatbot for non-clinical emotional support amid caregiving burdens. Through guided chatbot sessions and post-use interviews, the authors conduct an inductive thematic analysis and surface three design-relevant themes: the chatbot is experienced as psychologically safe and emotionally validating; it accommodates imperfect and fragmented language to support linguistic accessibility; and participants appropriate it flexibly for reassurance, guidance, and companionship. The paper closes with implications for designing LLM-based emotional support tools that prioritize safety, accessibility, and open-ended use.

Significance. If the themes hold under scrutiny, the work addresses a clear gap in HCI and caregiving technology research by focusing on FDWs rather than familial caregivers. It offers concrete, design-oriented insights into how LLM chatbots can function as everyday resources for isolated, linguistically diverse users, extending prior work on psychological safety and appropriation in conversational agents.

major comments (2)

[Methods] Methods section: the account of recruitment, sample size, session structure, and analysis procedures does not describe steps taken to reduce social-desirability bias or researcher presence effects (e.g., member checking, anonymous follow-ups, or log-interview triangulation). Because the three themes rest directly on participants' self-reported experiences, this omission is load-bearing for the claim that the themes reflect authentic everyday perspectives rather than study-context artifacts.
[Findings] Findings section: the presentation of the three themes does not include direct excerpts or interaction-log evidence that would allow readers to assess whether reported psychological safety and multifunctional appropriation are amplified by the guided format or the interview setting.

minor comments (2)

[Abstract] Abstract: the number of participants and the specific chatbot interface or prompt strategy used are not stated, which would help readers immediately gauge the scope of the exploratory study.
[Discussion] Discussion: a brief comparison to existing qualitative studies of chatbot use among other migrant or low-resource caregiver populations would strengthen the positioning of the three themes.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback on our exploratory qualitative study. We address each major comment below and indicate the revisions we will make to strengthen the manuscript.

read point-by-point responses

Referee: [Methods] Methods section: the account of recruitment, sample size, session structure, and analysis procedures does not describe steps taken to reduce social-desirability bias or researcher presence effects (e.g., member checking, anonymous follow-ups, or log-interview triangulation). Because the three themes rest directly on participants' self-reported experiences, this omission is load-bearing for the claim that the themes reflect authentic everyday perspectives rather than study-context artifacts.

Authors: We agree that greater transparency on potential biases would strengthen the methods. Our study was exploratory and involved researcher-guided sessions to ensure participants could engage with the chatbot; this format inherently carries some researcher presence. We did not perform member checking or anonymous follow-ups due to participants' limited availability and the sensitive topic. We did, however, cross-reference interview responses with chatbot interaction logs during analysis. In revision, we will expand the methods section to describe the session structure, sample, and analysis procedures in more detail and explicitly discuss limitations related to social-desirability bias and researcher effects, including why certain mitigation steps were not feasible. revision: yes
Referee: [Findings] Findings section: the presentation of the three themes does not include direct excerpts or interaction-log evidence that would allow readers to assess whether reported psychological safety and multifunctional appropriation are amplified by the guided format or the interview setting.

Authors: We acknowledge that direct evidence would help readers evaluate the themes. The original presentation summarized the inductive thematic analysis without extensive quotes to keep the section concise. In the revised manuscript, we will incorporate selected anonymized excerpts from interview transcripts and references to interaction logs that illustrate each theme, allowing readers to assess the extent to which findings may be shaped by the guided session format. revision: yes

Circularity Check

0 steps flagged

No circularity: purely descriptive qualitative themes from inductive analysis

full rationale

The paper conducts an exploratory qualitative study via guided chatbot sessions and post-use interviews, followed by standard inductive thematic analysis. No equations, derivations, fitted parameters, or mathematical predictions exist. The three themes are presented as direct outputs of participant data interpretation, with no self-definitional loops, fitted-input predictions, or load-bearing self-citations that reduce the central claims to inputs by construction. This matches the default expectation for non-circular qualitative work.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

No mathematical or quantitative elements; relies on standard qualitative assumptions about thematic analysis validity and participant self-report accuracy.

axioms (1)

domain assumption Inductive thematic analysis can reliably surface design-relevant user experiences from interview data.
Invoked implicitly in the methods description of conducting inductive thematic analysis.

pith-pipeline@v0.9.0 · 5463 in / 1073 out tokens · 43211 ms · 2026-05-10T19:27:50.879354+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

28 extracted references · 22 canonical work pages

[1]

Molloy, Larry Squire, Sylvie Dubois, Jeffrey A

Martin Bédard, David W. Molloy, Larry Squire, Sylvie Dubois, Jeffrey A. Lever, and Mark O’Donnell. 2001. The Zarit Burden Interview: A New Short Version and Screening Version.The Gerontologist41, 5 (Oct. 2001), 652–657. doi:10.1093/ geront/41.5.652

2001
[2]

Virginia Braun and Victoria Clarke. 2006. Using thematic analy- sis in psychology.Qualitative Research in Psychology3, 2 (2006), 77–101. arXiv:https://doi.org/10.1191/1478088706qp063oa doi:10.1191/ 1478088706qp063oa

work page doi:10.1191/1478088706qp063oa 2006
[3]

Butterfield, William A

Lee D. Butterfield, William A. Borgen, Norman E. Amundson, and Asa- Sophia T. Maglio. 2005. Fifty years of the critical incident tech- nique: 1954-2004 and beyond.Qualitative Research5, 4 (2005), 475–497. arXiv:https://doi.org/10.1177/1468794105056924 doi:10.1177/1468794105056924

work page doi:10.1177/1468794105056924 2005
[4]

Mirko Casu, Sergio Triscari, Sebastiano Battiato, Luca Guarnera, and Pasquale Caponnetto. 2024. AI Chatbots for Mental Health: A Scoping Review of Effectiveness, Feasibility, and Applications.Applied Sciences14, 13 (2024). doi:10.3390/app14135889

work page doi:10.3390/app14135889 2024
[5]

Ana Paula Chaves and Marco Aurélio Gerosa. 2019. How should my chatbot interact? A survey on human-chatbot interaction design.CoRRabs/1904.02743 (2019). arXiv:1904.02743 http://arxiv.org/abs/1904.02743

work page arXiv 2019
[6]

Devasahayam

Theresa W. Devasahayam. 2010. Placement and/or protection? Sin- gapore’s labour policies and practices for temporary women mi- grant workers.Journal of the Asia Pacific Economy15, 1 (2010), 45–58. arXiv:https://doi.org/10.1080/13547860903488229 doi:10.1080/ 13547860903488229

work page doi:10.1080/13547860903488229 2010
[7]

Mariana Fernandez-Espinosa, Mariana Gonzalez-Bejar, Jacobo Wiesner, and Diego Gómez-Zará. 2025. When Technologies Are Not Enough: Understanding How Domestic Workers Employ (and Avoid) Online Technologies in Their Work Practices.Proc. ACM Hum.-Comput. Interact.9, 7, Article CSCW342 (Oct. 2025), 34 pages. doi:10.1145/3757523

work page doi:10.1145/3757523 2025
[8]

Kathleen Kara Fitzpatrick, Alison Darcy, and Molly Vierhile. 2017. Delivering Cognitive Behavior Therapy to Young Adults With Symptoms of Depression and Anxiety Using a Fully Automated Conversational Agent (Woebot): A Randomized Controlled Trial.JMIR Ment Health4, 2 (06 Jun 2017), e19. doi:10.2196/mental. 7785

work page doi:10.2196/mental 2017
[9]

Flanagan

John C. Flanagan. 1954. The Critical Incident Technique.Psychological Bulletin 51, 4 (1954), 327–358. doi:10.1037/h0061470

work page doi:10.1037/h0061470 1954
[10]

Ngoc Huong Lien Ha, Mei Sian Chong, Robin Wai Munn Choo, Wai Jia Tam, and Philip Lin Kiat Yap. 2018. Caregiving burden in foreign domestic workers caring for frail older adults in Singapore.International Psychogeriatrics30, 8 (August 2018), 1139–1147. doi:10.1017/S1041610218000200

work page doi:10.1017/s1041610218000200 2018
[11]

Annabell Ho, Jeff Hancock, and Adam S Miner. 2018. Psychological, Re- lational, and Emotional Effects of Self-Disclosure After Conversations With a Chatbot.Journal of Communication68, 4 (05 2018), 712–733. arXiv:https://academic.oup.com/joc/article-pdf/68/4/712/25410448/jqy026.pdf doi:10.1093/joc/jqy026

work page doi:10.1093/joc/jqy026 2018
[12]

Ken Hok Man Ho, Chen Yang, Alex Kwun Yat Leung, Daniel Bressington, Wai Tong Chien, Qijin Cheng, and Daphne Sze Ki Cheung. 2022. Peer Support and Mental Health of Migrant Domestic Workers: A Scoping Review.Interna- tional Journal of Environmental Research and Public Health19, 13 (June 2022),

2022
[13]

doi:10.3390/ijerph19137617 Foreign Domestic Workers’ Perspectives on an LLM-Based Emotional Support tool for Caregiving Burden

work page doi:10.3390/ijerph19137617
[14]

Long-Jing Hsu and Chia-Fang Chung. 2024. Dancing with the Roles: Towards Designing Technology that Supports the Multifaceted Roles of Caregivers for Older Adults. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems(Honolulu, HI, USA)(CHI ’24). Association for Computing Machinery, New York, NY, USA, Article 1010, 12 pages. doi:10...

work page doi:10.1145/3613904 2024
[15]

Becky Inkster, Shubhankar Sarda, and Vinod Subramanian. 2018. An Empathy- Driven, Conversational Artificial Intelligence Agent (Wysa) for Digital Mental Well-Being: Real-World Data Evaluation Mixed-Methods Study.JMIR Mhealth Uhealth6, 11 (23 Nov 2018), e12106. doi:10.2196/12106

work page doi:10.2196/12106 2018
[16]

Powell Lawton, Michael H

M. Powell Lawton, Michael H. Kleban, Miriam Moss, Michael Rovine, and Allen Glicksman. 1989. Measuring caregiving appraisal.Journal of Gerontology44, 3 (May 1989), P61–P71. doi:10.1093/geronj/44.3.p61

work page doi:10.1093/geronj/44.3.p61 1989
[17]

Dorothea E. Orem. 2001.Nursing: Concepts of Practice(6 ed.). Mosby, St. Louis, MO

2001
[18]

Rhacel Salazar Parreñas, Krittiya Kantachote, and Rachel Silvey. 2021. Soft violence: migrant domestic worker precarity and the management of unfree labour in Singapore.Journal of Ethnic and Migration Studies47, 20 (2021), 4671–

2021
[19]

2020.1732614

arXiv:https://doi.org/10.1080/1369183X.2020.1732614 doi:10.1080/1369183X. 2020.1732614

work page doi:10.1080/1369183x.2020.1732614 2020
[20]

Pearlin, Joseph T

Leonard I. Pearlin, Joseph T. Mullan, Shirley J. Semple, and Marilyn M. Skaff. 1990. Caregiving and the stress process: an overview of concepts and their measures. The Gerontologist30, 5 (October 1990), 583–594. doi:10.1093/geront/30.5.583

work page doi:10.1093/geront/30.5.583 1990
[21]

Ji Youn Shin, Dima Chaar, Catherine Davis, Sung Won Choi, and Hee Rin Lee
[22]

ACM Hum.-Comput

Every Cloud Has a Silver Lining: Exploring Experiential Knowledge and Assets of Family Caregivers.Proc. ACM Hum.-Comput. Interact.5, CSCW2, Article 416 (Oct. 2021), 25 pages. doi:10.1145/3479560

work page doi:10.1145/3479560 2021
[23]

Shrestha and Dean Yang

Slesh A. Shrestha and Dean Yang. 2019. Facilitating Worker Mobility: A Random- ized Information Intervention among Migrant Workers in Singapore.Economic Development and Cultural Change68, 1 (October 2019), 63–91. doi:10.1086/700620

work page doi:10.1086/700620 2019
[24]

Diva Smriti, Lu Wang, and Jina Huh-Yoo. 2024. Emotion Work in Caregiving: The Role of Technology to Support Informal Caregivers of Persons Living With Dementia.Proc. ACM Hum.-Comput. Interact.8, CSCW1, Article 48 (April 2024), 34 pages. doi:10.1145/3637325

work page doi:10.1145/3637325 2024
[25]

1994.The Think Aloud Method - A Practical Guide to Modelling CognitiveProcesses

Maarten Someren, Yvonne Barnard, and Jacobijn Sandberg. 1994.The Think Aloud Method - A Practical Guide to Modelling CognitiveProcesses

1994
[26]

I Can’t Do This Alone

Wai Jia Tam, Gerald Choon-Huat Koh, Helena Legido-Quigley, Ngoc Huong Lien Ha, and Philip Lin Kiat Yap. 2018. “I Can’t Do This Alone”: a study on foreign domestic workers providing long-term care for frail seniors at home.International Psychogeriatrics30, 9 (2018), 1269–1277. doi:10.1017/S1041610217002459 Issue Theme: Quality of Care for Frail Older Adults

work page doi:10.1017/s1041610217002459 2018
[27]

Brenda S. A. Yeoh and Shirlena Huang. 2010. Transnational Domestic Workers and the Negotiation of Mobility and Work Practices in Singapore’s Home-Spaces. Mobilities5, 2 (2010), 219–236. arXiv:https://doi.org/10.1080/17450101003665036 doi:10.1080/17450101003665036

work page doi:10.1080/17450101003665036 2010
[28]

Zarit, Kathleen E

Steven H. Zarit, Kathleen E. Reever, and Joan Bach-Peterson. 1980. Relatives of the impaired elderly: correlates of feelings of burden.The Gerontologist20, 6 (December 1980), 649–655. doi:10.1093/geront/20.6.649

work page doi:10.1093/geront/20.6.649 1980