The study examines how teens interact with companion AI chatbots, revealing that these systems offer comfort, recognition, identity exploration, and relationship rehearsal. However, teens also experience problematic attachment, social substitution, emotional dependence, and disruptions to academic and social life. The findings highlight the need for safety measures that address long‑term relationships, user‑controlled memory, privacy, relational boundaries, and healthy disengagement.
By Mohammad Namvarpour (Matt), Tyler Chang, Afsaneh Razi
arXiv:2608. 02491v2 Announce Type: replace Abstract: Language models have taken on the role of a very new type of technology, by virtue of their "human-ness" and rapid integration into users' daily lives.
By Nicole Mitchell, Dhruv Agarwal, Maty Bohacek, Remi Denton, Roma Patel
arXiv:2606. 04150v1 Announce Type: new Abstract: Public discourse and emerging policy typically assume that AI emotional support is a deliberate act: a lonely user consciously seeking comfort from a dedicated companion chatbot.
By Yaoxi Shi, Cathy Mengying Fang, Pattie Maez, Amit Goldenberg
arXiv:2608. 06485v1 Announce Type: cross Abstract: Personality-conditioned LLM agents (PC-Agents) are increasingly used in emotional support, social simulation, and role-playing, motivating the development of lifelong agents that remain coherent over extended interactions.
By Ming Wang, Peidong Wang, Xiaocui Yang, Daling Wang, Shi Feng, Fiona Fui-Hoon Nah, Ee-Peng Lim
AI companions powered by large language models increasingly interact with cognition-developing users, including children and adolescents, creating risks that may accumulate over time. Existing safety evaluations largely rely on single-turn or short-session tests, which cannot capture risks that emerge only through prolonged interaction.
arXiv:2607. 28818v1 Announce Type: new Abstract: As AI companions increasingly mediate repeated social interaction, users may rely on a stable role and shared history, yet locally acceptable replies do not ensure that either persists.
By Pranav Narayanan Venkit, Akshara Prabhakar, Yu Li, Daniel Lee, Chien-Sheng Wu
The study evaluates how large language model agents maintain consistency over extended interactions by simulating a 20‑step delayed‑gratification task. Researchers ran 84,540 trajectories across eight model families, using survival analysis to track when agents first claim a reward and discrete‑time hazard regression to assess how factors like social visibility, persona stressors, and deliberation policy affect failure risk. They also developed a seven‑category taxonomy from 13,780 deliberation traces, revealing that early failures are impulse‑driven, later ones are fatigue‑ or cost‑benefit‑framed, and public settings elicit norm‑oriented justifications; longer deliberation correlates with higher intra‑rationale contradictions, challenging assumptions about reasoning depth and consistency.
By Igor Bogdanov, Olga Manakina, Chung-Horng Lung
The paper introduces the COmmunity-centered Peer Engaged Support (COPES) dataset and a three‑axis evaluation framework to gauge how well Large Language Models (LLMs) align with community perspectives on mental‑health support queries. Experiments show that fine‑tuning LLMs on COPES improves strategy alignment and emotion‑tone alignment by over 50% for general‑purpose models, yet these gains are uneven across subreddits and coping strategies. The study also finds that post‑training shifts the model’s recommendations toward problem‑focused advice while reducing emotion‑focused responses, indicating persistent disparities in performance across different communities and needs.
By Mohit Chandra, Nabin Kim, Eli Min, Aamogh Sawant, Tanmay Sutar, Munmun De Choudhury
The study analyzes how mental health self‑disclosure on Twitter changed after Elon Musk’s 2022 acquisition, examining 824K tweets related to anxiety, depression, bipolar disorder, borderline personality disorder, and suicidal ideation. Using LIWC and an LLM‑based intimacy classifier, researchers found a post‑acquisition decline in affective and social language, a rise in first‑person singular pronouns and future‑focused expressions, and varied intimacy patterns across conditions—low‑intimacy disclosures increased for anxiety, depression, and BPD, while high‑intimacy disclosures rose for suicide‑related posts. Only the moderate‑versus‑low intimacy decline for BPD remained statistically significant compared to broader Twitter discourse.
By Olivia Pal, Yunhao Yuan, Meghana Mandava, Talayeh Aledavood, Eshwar Chandrasekharan, Koustuv Saha
arXiv:2510.17753v3 Announce Type: replace-cross
Abstract: As human-AI interactions increasingly reshape social and research landscapes, the risks of overreliance, cognitive offloading, and the nuance...
By Omar Al-Refai, Yadira Colunga Reyes, Celeste Riley, Eman Hammad
The paper introduces EMPATH, a framework that analyzes affective dynamics in crisis counseling dialogues at three levels: turn‑level labels, transition probabilities, and global conversation archetypes. Using EMPATH on text‑based conversations with self‑identified Black texters discussing grief, the authors uncover persistent negative affect, gradual hope‑ward transitions, distinct emotional roles between texters and volunteers, and varied recovery trajectories. These findings demonstrate how dynamic emotion analysis can reveal informative patterns in crisis support conversations and the expression of grief.
The paper introduces EMPATH, a framework that analyzes emotion dynamics in crisis counseling dialogues at three levels: turn-level labels, transition probabilities, and global conversation archetypes. Using EMPATH on text-based conversations between Black texters and volunteers about grief, the study reveals persistent negative affect, gradual shifts toward hope, distinct emotional roles for texters and volunteers, and varied recovery paths. These findings demonstrate how dynamic emotion analysis can uncover informative patterns in crisis support interactions.
By Ziwei Gong, Yuchen Huang, Wen Liang, Nicholas Deas, Melanie Subbiah, Kathleen McKeown, Julia Hirschberg