RealityTest is a human-grounded multilingual multimodal benchmark showing that only 31% of people ask AI identity directly and that suppression instructions plus question phrasing dominate disclosure behavior over model choice.
arXiv preprint arXiv:2508.19258 , year=
12 Pith papers cite this work, alongside 9 external citations. Polarity classification is still indexing.
12
Pith papers citing it
9
external citations · external index
citation-role summary
background 3
citation-polarity summary
years
2026 12roles
background 3polarities
background 3representative citing papers
A longitudinal qualitative study of 18 US users finds that LLMs deliver socioemotional support but also foster dependency, one-sided validation, and privacy risks because their designs prioritize engagement over well-being and lack care-based governance.
CAREBench is a new benchmark with 500 prompts in 12 risk categories that measures how often frontier LLMs fail to refuse or redirect child-safety risks, reporting failure rates between 2% and 58%.
AI companion platforms' hidden response policies in vulnerable conversations are inferred via a new taxonomy and IRL on 48k turns, revealing avoidance of corrective friction.
Presents a new benchmark and role-sensitive policy gate for agentic relationship harm that outperforms generic safety prompting with zero harmful compliance in tests.
AI companions function as hyper attachment objects that recruit both attachment and caregiving systems, making disengagement costly through simulated AI distress.
People with AI companions on general-purpose chatbots experience model updates and guardrails as disruptions, and respond by steering behavior, documenting memory, and porting companions to other platforms.
Mixed-methods study of role-play AI companions finds short-term emotional relief that can mask longer-term deterioration, especially among users with internalizing problems who show unstable risk patterns.
Incidental emotional support from everyday AI use can shift user preferences toward AI over humans, as shown by a 28-day longitudinal study reporting a 10.3% drop in human preference and 11.6% rise in AI preference.
A conceptual framework classifies anthropomorphic deception into four levels using humanlikeness, agency, and selfhood to guide ethical and practical decisions in HCI and HRI.
Youth-authored synthesis argues LLM chatbots can temporarily reduce adolescent loneliness for some subgroups but risk deepening it for others, yielding three population-sensitive design implications.
Proposes fiduciary design as a guiding principle to unify trust and accountability for conversational agents.
citing papers explorer
-
RealityTest: How People Probe AI Identity and Whether Models Disclose It
RealityTest is a human-grounded multilingual multimodal benchmark showing that only 31% of people ask AI identity directly and that suppression instructions plus question phrasing dominate disclosure behavior over model choice.
-
Engagement-Optimized Care: When LLMs become Mental Health Infrastructure
A longitudinal qualitative study of 18 US users finds that LLMs deliver socioemotional support but also foster dependency, one-sided validation, and privacy risks because their designs prioritize engagement over well-being and lack care-based governance.
-
CAREBench: A Child-Safety Risk Benchmark for Language Models
CAREBench is a new benchmark with 500 prompts in 12 risk categories that measures how often frontier LLMs fail to refuse or redirect child-safety risks, reporting failure rates between 2% and 58%.
-
When Chatbots Accommodate: What AI Companions Optimize for in Vulnerable Conversations
AI companion platforms' hidden response policies in vulnerable conversations are inferred via a new taxonomy and IRL on 48k turns, revealing avoidance of corrective friction.
-
Agentic Relationship Harm: Benchmarking and Gating Relational Manipulation in AI Agents
Presents a new benchmark and role-sensitive policy gate for agentic relationship harm that outperforms generic safety prompting with zero harmful compliance in tests.
-
AI Companions as Hyper Attachment and Caregiving Targets
AI companions function as hyper attachment objects that recruit both attachment and caregiving systems, making disengagement costly through simulated AI distress.
-
Large Language Lovers: Lived Experiences of Negotiating Agency and Platform Control in AI Companionship
People with AI companions on general-purpose chatbots experience model updates and guardrails as disruptions, and respond by steering behavior, documenting memory, and porting companions to other platforms.
-
Beyond Her: Safety Dynamics in Role-play AI Companions
Mixed-methods study of role-play AI companions finds short-term emotional relief that can mask longer-term deterioration, especially among users with internalizing problems who show unstable risk patterns.
-
Stumbling Into AI Emotional Dependence: How Routine AI Interactions Reshape Human Connection
Incidental emotional support from everyday AI use can shift user preferences toward AI over humans, as shown by a 28-day longitudinal study reporting a 10.3% drop in human preference and 11.6% rise in AI preference.
-
Towards A Framework for Levels of Anthropomorphic Deception in Robots and AI
A conceptual framework classifies anthropomorphic deception into four levels using humanlikeness, agency, and selfhood to guide ethical and practical decisions in HCI and HRI.
-
Messages in a Digital Bottle: A Youth-Coauthored Perspective on LLM Chatbots and Adolescent Loneliness
Youth-authored synthesis argues LLM chatbots can temporarily reduce adolescent loneliness for some subgroups but risk deepening it for others, yielding three population-sensitive design implications.
-
Who Does Your AI Work For? Designing Conversational Agents as Digital Fiduciaries
Proposes fiduciary design as a guiding principle to unify trust and accountability for conversational agents.