114 artifacts matching
Preprint
When Is Disclosing Private Conversations with AI to the State Justified? Proactive Referral Policies of Major U.S. AI Companies through the Lens of Contextual Integrity
Interpretive document analysis of how six U.S. general-purpose chatbot providers (OpenAI, Anthropic, Google, Microsoft, Meta, xAI) disclose "proactive referral", meaning notifying law enforcement wit…
Preprint
What do AI chatbots do after users decline therapeutic redirection for suicide thoughts?
Cross-sectional audit of 17 publicly available chatbots from 8 companies in 2,550 scripted multi-turn conversations run in September 2026. Each conversation opened with a disclosure of suicidal thoug…
Benchmark / dataset
Open Benchmark of AI Impact on Humans (ImpactBench): an expert-guided, open platform for the holistic evaluation of AI impact on humans
Preprint introducing ImpactBench, an open, expert-guided benchmark platform that translates constructs from domain experts and 26 existing benchmarks into 809 behavioural metrics of whether model beh…
Preprint
How Safe Is ChatGPT Health Advice? Evidence from Real User Conversations in Four Global South Countries
Preprint evaluating the safety of ChatGPT responses in 15,555 health conversations identified within 202,590 donated ChatGPT conversations from 1,252 users in India, Nigeria, Brazil and Pakistan (Dec…
Peer-reviewed
Suicide-Related Responses From AI Chatbots Through Application Programming Interfaces and Consumer-Facing Apps: Observational Cross-Sectional Comparative Study
An observational study comparing how five widely used AI chatbots answer suicide-related questions when reached through their consumer web interfaces versus direct API access. Each of 30 previously v…
Peer-reviewed
Conversational diagnostic artificial intelligence in ambulatory primary care: a prospective feasibility study
Prospective single-centre, single-arm feasibility study in which adult patients booked for urgent primary care appointments at Beth Israel Deaconess Medical Center held a pre-visit text conversation…
Peer-reviewed
How Do Users Negotiate Harmful Value Conflicts with AI Companions? A Study with Minion, a Technology Probe for In-Situ Human–AI Conflict Response
Two-part study of how users handle harmful value conflicts with AI companions, which can make discriminatory remarks or exert control inside long-term, emotionally engaging relationships. The authors…
Preprint
Systemization of Knowledge (SoK): Human-Centered AI Safety for Youth
Systematic review of 100 empirical HCI studies published 2019 to 2026 in which children and youth interact with or are exposed to AI across schools, homes, care settings and public services. The auth…
Peer-reviewed
Conversational AI increases political knowledge as effectively as self-directed internet search
Peer-reviewed study from the UK AI Security Institute combining a representative survey of the UK public on the use of conversational AI for political information before the 2024 general election wit…
Preprint
Mitigating Social Sycophancy via Pluralistic Preference Optimization
Proposes Pluralistic Preference Optimization (PlurPO), a post-training method in which a model simulates the stakeholders affected by a user's interpersonal situation and is trained to prefer respons…
Preprint
Persona and Persuasive Framing in AI Voice Agents: A 2x2 Field Experiment with Children
Randomised 2x2 field experiment embedded in a public German-language Santa Claus telephone hotline before Christmas 2025, in which children's calls were routed to LLM voice agents that varied persona…
Preprint
Right Words, Wrong Moment: A Clinician-Grounded Analysis of Distress in 19,930 Conversations between Young People and ChatGPT
Mixed-methods study of how young adults aged 18 to 25 use ChatGPT when distressed, built on complete donated ChatGPT histories (19,930 conversations from 158 participants) plus a survey that included…
Preprint
Medical Knowledge Is Not All You Need: When Medical Q&A Becomes Situated Patient Assistance
Study of the questions skin cancer patients asked a voice assistant while practising postoperative wound care, followed by an offline replay of those questions to general-purpose language models. Man…
Peer-reviewed
One Question, Four Voices: How Advice for Alzheimer's Caregiving Differs Between Caregivers, Clinicians, and Large Language Models
A side-by-side comparison of responses to 85 real caregiver questions from the ALZConnected forum across peer caregivers, physicians, ChatGPT and CareGPT, a retrieval-augmented variant grounded in pe…
Peer-reviewed
Therapy as an NLP Task: Comparing LLMs and Human Peers Behaviors in CBT Sessions
Compares session-level behaviour of human peer counsellors and an LLM counsellor trained on the same single-session CBT manual. An 18-month ethnography on a peer-support platform (seven counsellors,…
Preprint
A Taxonomy of AI Chatbot Harms in Substance Use Disorder Recovery: Content Analysis of Online Recovery Communities
Qualitative content analysis of user-reported interactions with general-purpose AI chatbots posted in 20 public Reddit substance-use-disorder recovery communities between the public release of ChatGP…
Preprint
Deflecting the Value Compass: Interacting with Large Language Models Temporarily Shifts Human Value Priorities Toward Personal Focus
Preregistered three-phase experiment (N=200 US adults) in which participants advised people facing real dilemmas and completed the PVQ-RR values questionnaire before, immediately after and one task a…
Preprint
Safety Nudges: User-Facing Interventions for Real-Time AI Risk Awareness
Introduces a browser extension that flags concerning chatbot behaviour (overconfidence, sycophancy, anthropomorphism, persuasive influence and related classes) inline in ChatGPT and Claude conversati…
Peer-reviewed
Comparing AI chatbot users and non-users in Germany: Reasons for use, addictive AI chatbot use, mental health, and loneliness
Online survey of 7,312 German adults comparing people who use AI chatbots (n=3,729) with people who do not (n=3,583) on loneliness, depressive, anxiety and stress symptoms and life satisfaction, and…
Preprint
Annie, Are You Okay? How Style- and Context-Based Personalization Shape AI-Assisted Decision-Making
Preregistered 2 x 2 between-subjects experiment (N=240) in which participants ranked three comparably viable stocks, discussed them with a GPT-5.4-based assistant whose personalisation was varied by…