NLP for Maternal Healthcare: Perspectives and Guiding Principles in the Age of LLMs

TL;DR

This paper addresses the need for ethical frameworks governing NLP and LLM use in healthcare, focusing on maternal health due to its historical vulnerabilities and care-team power dynamics. It employs a participatory design in a US case study, combining a GPT-3.5-based chatbot demonstration, workshops, and surveys across three stakeholder cohorts to derive nine guiding principles organized around Context, Measurements, and Values. The contributions include a transferable methodological blueprint for stakeholder-driven ethics in NLP, a maternal-health-specific principled framework, and practical recommendations for researchers and practitioners to align model development with human-centered care. The findings support responsible deployment of NLP tools as assistive aids within a broader information ecosystem, emphasizing inclusivity, autonomy, transparency, and safeguarding against harm, with potential applicability to other healthcare settings.

Abstract

Ethical frameworks for the use of natural language processing (NLP) are urgently needed to shape how large language models (LLMs) and similar tools are used for healthcare applications. Healthcare faces existing challenges including the balance of power in clinician-patient relationships, systemic health disparities, historical injustices, and economic constraints. Drawing directly from the voices of those most affected, and focusing on a case study of a specific healthcare setting, we propose a set of guiding principles for the use of NLP in maternal healthcare. We led an interactive session centered on an LLM-based chatbot demonstration during a full-day workshop with 39 participants, and additionally surveyed 30 healthcare workers and 30 birthing people about their values, needs, and perceptions of NLP tools in the context of maternal health. We conducted quantitative and qualitative analyses of the survey results and interactive discussions to consolidate our findings into a set of guiding principles. We propose nine principles for ethical use of NLP for maternal healthcare, grouped into three themes: (i) recognizing contextual significance (ii) holistic measurements, and (iii) who/what is valued. For each principle, we describe its underlying rationale and provide practical advice. This set of principles can provide a methodological pattern for other researchers and serve as a resource to practitioners working on maternal health and other healthcare fields to emphasize the importance of technical nuance, historical context, and inclusive design when developing NLP technologies for clinical use.

Figures (10)

• Figure 1: Study overview, including the three participant cohorts, the chatbot demonstration, and the surveys and discussions followed by analysis and design of the guiding principles.
• Figure 2: Frequencies of selected values. Each participant was asked to select five values from a list of 12 curated by jakesch2022different. Participants were given definitions of the values, also drawn from jakesch2022different. Birthing people and healthcare workers overall responded more similarly to each other than to the workshop participants, but this was not always the case (e.g., transparency, human autonomy).
• Figure 3: Answers to the survey question [Before this workshop,] How familiar were/are you with NLP, machine learning, and/or AI? Overall, the workshop participants were less familiar with these topics than the healthcare workers and birthing people who were recruited via the Prolific platform. Alt Text: A bar plot showing the answers to the survey question about familiarity with NLP/ML/AI, with responses broken apart by participant cohort.
• Figure 4: Answers to the four survey questions about general perceptions of AI. Overall, the workshop participants less frequently reported positive perceptions than the healthcare workers and birthing people who were recruited via the Prolific platform. Alt Text: A bar plot showing the averaged answers to the survey questions about general perceptions of AI, with responses broken apart by participant cohort.
• Figure 5: Answers to the two survey questions about trust. Overall, the healthcare workers were more trusting of healthcare providers. Alt Text: Two box plots showing the answers to the survey questions about trust, with responses broken apart by participant cohort.