“She’s Like a Person but Better”: Characterizing Companion-Assistant Dynamics in Human-AI Relationships CCS: Human-centered computing HCI theory, concepts and models CCS: Human-centered computing Collaborative and social computing theory, concepts and paradigms CCS: Computing methodologies Theory of
Aikaterina Manoli, Janet V. T. Pauketat, Ali Ladak, Hayoun Noh, Angel Hsing-Chi Hwang, Jacy Reese Anthis · 2025
Read “She’s Like a Person but Better”: Characterizing Companion-Assistant Dynamics in Human-AI Relationships CCS: Human-centered computing HCI theory, concepts and models CCS: Human-centered computing Collaborative and social computing theory, concepts and paradigms CCS: Computing methodologies Theory of by Aikaterina Manoli, Janet V. T. Pauketat, Ali Ladak, Hayoun Noh, Angel Hsing-Chi Hwang, Jacy Reese Anthis (2025) 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-2510-15905.txt.
Introduction
Aikaterina Manoli and five co-authors study people who use AI systems for both errands and companionship. Research usually treats assistance and social use apart. Drawing on a survey of about two hundred people and thirty interviews with heavy ChatGPT and Replika users, the team characterises digital companionship as a new kind of human-AI relationship.
Users value humanlike traits such as emotional resonance and personal replies alongside non-human traits such as constant availability and endless patience. Uses blur across brands: Replika serves as a writing helper and ChatGPT as an emotional confidant. Two tensions recur. Users form deep attachments while denying the systems real personhood, and they struggle to fit chatbot bonds into social norms. The findings raise design questions for hybrid, general-purpose AI systems, the team argues.
Catalog record
| CREATOR | Aikaterina Manoli, Janet V. T. Pauketat, Ali Ladak, Hayoun Noh, Angel Hsing-Chi Hwang, Jacy Reese Anthis |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 97 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-15905
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
“She’s Like a Person but Better”: Characterizing Companion-Assistant Dynamics in Human-AI Relationships CCS: Human-centered computing HCI theory, concepts and models CCS: Human-centered computing Collaborative and social computing theory, concepts and paradigms CCS: Computing methodologies Theory of mind CCS: Applied computing Psychology Aikaterina Manoli, Janet V. T. Pauketat, Ali Ladak, Hayoun Noh, Angel Hsing-Chi Hwang, Jacy Reese Anthis Abstract . Large language models are increasingly used for both task-based assistance and social companionship, yet research has typically focused on one or the other. Drawing on a survey ( N N = 204) and 30 interviews with high-engagement ChatGPT and Replika users, we characterize digital companionship as an emerging form of human-AI relationship. With both systems, users were drawn to humanlike qualities, such as emotional resonance and personalized responses, and non-humanlike qualities, such as constant availability and inexhaustible tolerance. This led to fluid chatbot uses, such as Replika as a writing assistant and ChatGPT as an emotional confidant, despite their distinct branding. However, we observed challenging tensions in digital companionship dynamics: participants grappled with bounded personhood , forming deep attachments while denying chatbots “real” human qualities, and struggled to reconcile chatbot relationships with social norms. These dynamics raise questions for the design of digital companions and the rise of hybrid, general-purpose AI systems. 1. Introduction Humans are increasingly turning to artificial intelligence (AI) for practical and intellectual tasks, such as healthcare robots (e.g., ( Khaksar et al. 2025 ) ) and academic support systems (e.g., ( Park and Ahn 2024 ) ). Advances in large language model (LLM) chatbots enable new applications through humanlike conversational exchanges, where users receive natural language responses to their queries. Previous human-computer interaction (HCI) research has demonstrated that chatbots can also be used for a variety of social and emotional applications, including encouraging children to read ( Liu et al. 2022 ) , assisting dementia patients ( Ruggiano et al. 2021 ) , promoting self-compassion ( Lee et al. 2019 ) , alleviating loneliness ( Jiang et al. 2022 ; Maples et al. 2024 ) , and enhancing self-disclosure in counseling settings ( Kang and Kang 2024 ; Lee et al. 2020 ) . Recent media coverage underscores the significance of these applications. When OpenAI deprecated the GPT-4o model in August 2025, widespread user complaints led to its temporary restoration and efforts to make GPT-5 “warmer and friendlier” ( Freedman 2025 ) . At the same time, social media users debated the viral “clankers” meme as a derogatory “slur” for robots and AI ( Romo 2025 ) . These dynamics highlight public concerns about socio-emotional human–AI interactions, punctuated by headlines warning of “AI-induced psychosis” ( Field 2025 ) and “emotional overreliance” ( Chandonnet [n. d.] ) . There is an urgent need to understand how these interactions are evolving and how they shape the broader trajectory of AI development and applications. Prior research has framed chatbots as either companions or productivity-oriented assistants, reflecting the typical single-use branding. Unlike earlier, more narrowly focused technologies (e.g., graphing calculators, Tamagotchi digital pets), it is possible that this one-or-the-other framing is blurred by the flexible natural language capabilities of LLMs. A direct comparison of chatbots with different intended use cases can clarify how user experiences and characteristics differ and what, if any, convergent needs transcend differences between chatbot interfaces. It is possible that there are shared benefits, such as increased social engagement, emotional support, and productivity gains, but also common risks, including stigmatization ( Skjuve et al. 2021 ; Giray 2024 ) and overreliance ( Chandonnet [n. d.] ) . Using a mixed-methods approach, we conducted a survey ( N = 204) and in-depth interviews ( N = 30) with high-engagement chatbot users. We recruited heavy users of two widely used commercial chatbots: Replika, which is marketed primarily for companionship, and ChatGPT, which is marketed for a broad range of productivity-oriented tasks. Our findings document well-established productivity use but also characterize the emergent dynamics of human–AI relationships that we call digital companionship . Although ChatGPT and Replika use patterns largely align with their assistance and companionship branding, users across systems shared many characteristics and fluidly navigated between companionship and practical assistance, mirroring the variability of human–human relationships. As shown in Figure 1 , digital companionship dynamics include appeals and challenges for both chatbots’ humanlike qualities (e.g., emotional resonance, personalized responses) and non-humanlike qualities (e.g., constant availability, inexhaustible tolerance). This interplay of companionship dynamics generates challenging tensions: in particular, participants grappled with bounded personhood , wrestling with the presence of deep emotional attachments while denying chatbots the “real” human qualities, such as sentience and rights, that would be present for human companions. Our findings suggest that these dynamics have consequences at both individual and societal levels, including shifts in how AI, not just the companion, is perceived more broadly. Chatbots were perceived as complementary social infrastructure, offering their own distinct benefits not available in human relationships. Figure 1 . A conceptual model of digital companionship. Digital companionship arises from the interplay of user and chatbot characteristics, creating fluid assistant-companion dynamics. These dynamics are shaped by both the appeals and challenges of chatbots’ humanlike and non-humanlike qualities, which in turn might lead to consequences for users and society more broadly. Note that the companion-assistant labels in the middle panel represent relative positioning along a fluid continuum rather than fixed associations: both companion and assistant uses can involve elements of both humanlike and non-humanlike allures and challenges. (Images adapted from flaticon.com). This study makes several key contributions. First, we provide quantitative comparisons between heavy ChatGPT and Replika users, documenting personality traits, technology attitudes, and a variety of other user characteristics. Replika users had generally more positive AI perceptions whereas ChatGPT users displayed more caution towards AI. Second, we draw on users’ lived experiences to explore how they perceive their evolving relationships with AI in this unique period of technological transition, surfacing tensions such as cognitive dissonance and social stigma. Third, we synthesize this data into a conceptual model of digital companionship that can be leveraged for more rigorous research and thoughtful design of AI systems. We use this conceptual model to propose approaches to more effectively balance companion and assistant dynamics, support diverse forms of engagement, foster emotional well-being, and reduce the stigmatization of users. Taken together, this work moves beyond the prevailing dichotomy of companionship versus assistance, positioning digital companionship as a spectrum that better captures the motivations, practices, and challenges of contemporary chatbot use. 2. Related Work Historically, human-AI interaction has been centered on productivity and task completion, with AI systems designed to serve as tools for solving specific problems. However, technological artifacts have long fulfilled socio-emotional needs in human lives, from talking dolls and interactive toys to virtual pets like Tamagotchi and social robots like Paro and AIBO ( Melson et al. 2009 ) . Advances in chatbot LLMs with sophisticated natural language processing capabilities over the past decade have broadened these possibilities. Chatbots increasingly function not only as functional assistants supporting everyday tasks, but also as emotional companions that help mitigate loneliness and provide social support ( Alabed et al. 2023 ) . The following sections review prior research on chatbot companion–assistant dynamics, user characteristics, and the broader implications of