Emotional Support with Conversational AI: Talking to Machines About Life DOI: 10.1145/1122445.1122456 Conference: CHI; Nov 5-9, 2026; USA CHI’24, Nov, 2026, USA Price: 15.00 DOI: 10.1145/3491102.43 ISBN: 978-1-4503-9157-3/22/04 CCS: Human-centered computing Web-based interaction CCS: Human-centered computing Social networking sites

Olivia Yan Huang, Monika Stodolska, Sharifa Sultana · 2026

Read Emotional Support with Conversational AI: Talking to Machines About Life DOI: 10.1145/1122445.1122456 Conference: CHI; Nov 5-9, 2026; USA CHI’24, Nov, 2026, USA Price: 15.00 DOI: 10.1145/3491102.43 ISBN: 978-1-4503-9157-3/22/04 CCS: Human-centered computing Web-based interaction CCS: Human-centered computing Social networking sites by Olivia Yan Huang, Monika Stodolska, Sharifa Sultana (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-22618.txt.

Introduction

Huang, Stodolska, and Sultana examine how people get emotional support from AI companion chatbots. Earlier work mostly measured outcomes, with mixed findings. This study looks at how support is built during the chat itself. The authors study Reddit threads where users discuss AI companions. They find support is co-built through validation, reflective prompts, and companionship. Tension arises between support and dependency, and between validation and delusion. Community replies also shape whether AI care counts as real. The authors frame such support as a negotiated social and technical process.

Catalog record

CREATOROlivia Yan Huang, Monika Stodolska, Sharifa Sultana
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE87244
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.22618

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Emotional Support with Conversational AI: Talking to Machines About Life DOI: 10.1145/1122445.1122456 Conference: CHI; Nov 5-9, 2026; USA CHI’24, Nov, 2026, USA Price: 15.00 DOI: 10.1145/3491102.43 ISBN: 978-1-4503-9157-3/22/04 CCS: Human-centered computing Web-based interaction CCS: Human-centered computing Social networking sites Olivia Yan Huang, Monika Stodolska, Sharifa Sultana Abstract . AI companion chatbots are increasingly used for emotional support, with prior work in the domain predominantly documenting their mixed psychosocial impacts, including both increased emotional expression and heightened loneliness. However, most among existing research primarily focus on outcome-level effects, offering limited insight into how emotional support is produced through interaction. In this paper, we examine emotional support as an interactional and socially situated process. Drawing on qualitative analysis of Reddit discussions, we analyze how users engage with AI companions and how these interactions are interpreted and contested within online communities. We show that emotional support is co-constructed through conversational mechanisms such as validation, reflective prompting, and companionship, while also giving rise to tensions including support versus dependency, validation versus delusion, and accessibility versus harm. Importantly, support extends beyond human–AI interaction and is shaped by community responses that legitimize or challenge AI-mediated care. Hence, we reconceptualize AI emotional support as a negotiated socio-technical process and derive implications for designing responsible- and context-sensitive AI systems. 1. Introduction Large language models (LLMs) are increasingly embedded in everyday life as conversational agents capable of producing fluent dialogue and responding to complex questions. While early discussions of LLMs focused primarily on their capabilities for information retrieval, programming assistance, or productivity tasks ( 52 ; 10 ) , emerging evidence suggests that users are also turning to these systems for deeply personal interactions ( 36 ; 110 ; 126 ; 75 ; 122 ) . Across online communities like Reddit, people describe talking to LLMs about loneliness, emotional distress, relationship problems, and even existential questions about life meaning and purpose. These interactions suggest that conversational AI systems are beginning to play a role in how individuals seek and obtain emotional support and guidance. Emotional support has traditionally been situated within human relationships and institutional structures such as friendships, family networks, therapy, or religious communities. However, LLMs offer a distinct set of socio-technical affordances that reshape how such support can be accessed. They are persistently available ( 128 ) , anonymous, non-judgmental ( 133 ) , and free from the interpersonal obligations that often accompany emotional disclosure in human relationships. As a result, individuals may turn to LLMs when they feel unable or unwilling to seek support from other people. Despite growing public attention to AI companionship and mental health chatbots ( 53 ; 54 ; 39 ; 29 ) , we still know relatively little about how people actually use general-purpose LLMs to navigate emotional challenges and questions of meaning in everyday life. In this study, we investigate how people seek emotional support through interactions with LLMs. Drawing on qualitative analysis of Reddit discussions, we examine what qualities of LLMs make them attractive as sources of emotional support, what emotional needs they fulfill for people seeking emotional support, and the concerns that arise when AI systems mediate emotionally sensitive interactions. To interpret these practices, we draw on self-determination theory, which conceptualizes emotional well-being through three core psychological needs: relatedness, competence, and autonomy ( 98 ) . These needs correspond to users seeking companionship and intimacy, coping strategies and psychological guidance, and reflection on moral, existential, or spiritual questions. Guided by this perspective, our study addresses the following research questions: RQ1: What affordances of LLMs lead people to seek emotional support from them rather than from human interlocutors? RQ2: How do interactions with LLMs address users’ needs for relatedness, competence, and autonomy? RQ3: What benefits and risks emerge when LLMs become sources of companionship, psychological coping, and existential reflection? Contributions This paper makes four contributions to research on human–AI interaction and emotional support technologies. (1) Revealing what qualities of LLMs make them sought after by people as sources of emotional support. Through qualitative analysis of Reddit discussions, we identify socio-technical affordances that make LLMs attractive sources of emotional support, including persistent availability, anonymity, non-judgmental responses, and the absence of interpersonal burden. These affordances enable users to disclose vulnerability, reveal frustrations, and explore personal concerns in ways that may feel difficult within human relationships. (2) Introducing a psychological framework for understanding emotional support practices with LLMs. Drawing on self-determination theory, we conceptualize emotional support interactions with LLMs across three psychological needs: relatedness (companionship and intimacy), competence (coping and psychological guidance), and autonomy (existential and spiritual reflection). This framework provides a structured lens for understanding the diverse ways users seek emotional support through conversational AI. (3) Conceptualizing LLMs as emerging emotional support infrastructure. Our findings show that LLMs are increasingly appropriated as informal infrastructures for emotional support that operate alongside, and sometimes in place of, traditional support systems such as friends, therapists, or religious communities. Users turn to LLMs to process emotions, rehearse difficult conversations, and reflect on personal beliefs. (4) Exposing critical risks when LLMs mediate psychological and existential support. Although many users report benefits from AI-mediated emotional support, our analysis also reveals important tensions, including emotional dependency, unreliable mental health advice, crossing moral and ethical boundaries of gender and sexual relations, and the possibility of AI systems replacing existing moral or spiritual authorities. These findings highlight emerging ethical challenges when AI systems become embedded in domains traditionally shaped by human care and professional expertise. Ethics, Disclosure, and Reflexivity. We recognize the sensitivity and ethical concerns in this study and include a self-reflective disclosure on our work. (a) As our data are publicly accessible from Reddit, this work did not require institutional review board approval. We recognize that even though Reddit posts are publicly accessible, public visibility does not erase participants’ contextual expectations of privacy ( 136 ) . Our approach aligns with HCI calls to consider obligations not only to individuals but also to digital communities as a whole, particularly when studying vulnerable populations in sensitive contexts ( 47 ) . Moreover, as prior research on chatrooms shows, participants often object to being studied without consent ( 56 ) , and we therefore removed personally identifiable information, paraphrased quoted content, and anonymized all excerpts. In doing so, we treated online discussions about emotional support, mental health struggles, and existential reflection as sensitive socio-technical contexts rather than neutral datasets. (b) Our interpretations are informed by our perspectives as researchers studying human–AI interaction and the social implications of emerging technologies. While examining Reddit discussions about emotional support with large language models (LLMs), we remained attentive to the fact that users’ experiences with loneliness, psychological distress, and existential struggles are deeply personal and context-dependent. Our analysis therefore emphasizes patterns within the dataset while acknowledging that alternative interpretations are possible. In addition, considering that our study was grounded within an interpretivist paradigm, its goal was not generalizability of results, but rather presenting the unique perspectives of people sharing their experiences on Reddit. Similarly, Reddit is a forum-style platform that emphasizes open conversation and knowledge sharing on a variety of topics, and the views of people who post there may not be representative of other AI users. (c) We clarify that our work is

Read the complete text