Beyond Judgment: Exploring LLM as a Support System for Maternal Mental Health

TL;DR

This study addresses how mothers use Large Language Models (LLMs) to avoid social judgment and obtain non-medical emotional support. Using a cross-sectional online survey with $N=107$, the authors show that mothers view LLMs as judgment-free, readily available resources that provide quick information and emotional reassurance, especially when real-world support is limited or faces scrutiny. While many participants still value human warmth and long-term connection, a substantial minority—notably in joint-family settings—rely on LLMs to bypass judgment and manage moments of guilt, anger, or loneliness. The findings highlight LLMs as low-risk, situational support tools that complement human networks and underscore the influence of social context on emotional technology use, informing design considerations for empathetic yet boundary-aware AI companions for caregivers.

Abstract

In the age of Large Language Models (LLMs), much work has already been done on how LLMs support medication advice and serve as information providers; however, how mothers use these tools for emotional and informational support to avoid social judgment remains underexplored. In this study, we have conducted a 10-day mixed-methods exploratory survey (N=107) to investigate how mothers use LLMs as a non-judgmental resource for emotional support and regulation, as well as situational reassurance. Our findings show that mothers are asking LLMs various questions about childcare to reassure themselves and avoid judgment, particularly around childcare decisions, maternal guilt, and late-night caregiving. Open-ended responses also show that mothers are comfortable with LLMs because they do not have to think about social consequences or judgment. Although they use LLMs for quick information or reassurance to avoid judgment, the results also show that more than half of the participants value human warmth over LLMs; however, a significant minority, especially those who live in a joint family, consider LLMs to avoid human judgment. These findings help us understand how we can frame LLMs as low-risk interaction support rather than as a replacement for human support, and highlight the role of social context in shaping emotional technology use.

Figures (5)

• Figure 1: Participant perceptions of LLMs’ non-judgmental nature and the trade-off between accuracy and emotional warmth ($N=107$). Left: frequency of preferring LLMs due to perceived absence of judgment. Right: relative importance of accuracy, empathy, lack of judgment, and human emotional warmth.
• Figure 2: Distribution of preferences regarding accuracy and emotional warmth across family types. While respondents from both nuclear and joint families predominantly reported missing human warmth, participants from joint families more frequently emphasized the absence of human judgment and perceived LLMs' empathy, whereas those from nuclear families placed relatively greater emphasis on accuracy and data-driven responses.
• Figure 3: Participant reports of using LLMs for emotional and practical support (Left), maternal guilt reassurance (Right). Responses indicate selective use of LLMs, with greater engagement for reassurance and reflection than for managing guilt or intense emotional states.
• Figure 4: Use of LLMs for anger management (Left) and late-night loneliness (Right) among mothers ($N=107$). Results show limited use of LLMs for managing anger and selective use during late-night caregiving for reassurance or reduced isolation.
• Figure 5: Associations between employment/education context and late-night LLMs use. Left: distribution of midnight_g responses across employment status groups. Right: distribution of midnight_g across student level groups. Both associations were statistically significant in bivariate $\chi^2$ tests (employment status: $\chi^2=16.58$, $df=6$, $p=.011$, Cramér’s $V=.278$; education level: $\chi^2=8.20$, $df=3$, $p=.042$, Cramér’s $V=.277$).