Long-term Measurements: Towards a Longitudinal Understanding of Human-AI Interactions

2026/08/03 by Nicole Mitchell, Dhruv Agarwal, Maty Bohacek +2
Computer Science · #cs.AI

paper · pdf

arxiv created 2026/08/05 · arxiv updated 2026/08/06

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. This combination of features can introduce longitudinal risks---cognitive, developmental and socio-affective changes in humans---that might not surface during a short-term interaction, but can have lasting long-term effects on users. This forms the basis of a critical new mission for NLP: to pivot from static, short-term evaluations of text generations to long-term measurements of behavioral changes, towards a diachronic understanding of human-model interactions. In this work, we draw from measurements used in social science fields that are crucial to understand emergent phenomena in longitudinal data. We discuss how computational methods in the field of NLP need to be combined with such measurements, not only to understand long-term safety risks of human-model interactions, but to help steer model development towards positive rather than negative outcomes for users. This ability to model human behavioral shifts as a function of model interactions can facilitate online rather than post-hoc detection of problematic behaviors, and should be leveraged in alignment frameworks to mitigate long-term risks in users.

Citations

• The efficiency-gain illusion: People underestimate the rate of AI use and overestimate its benefits on simple tasks
• Sycophantic AI makes human interaction feel more effortful and less satisfying over time
• "AI Psychosis" in Context: How Conversation History Shapes LLM Responses to Delusional Beliefs
• Evaluating Language Models for Harmful Manipulation
• Biased AI writing assistants shift users’ attitudes on societal issues
• Sycophantic Chatbots Cause Delusional Spiraling, Even in Ideal Bayesians
• Large Language Models Polarize Ideologically but Moderate Affectively in Online Political Discourse
• Neural steering vectors reveal dose and exposure-dependent impacts of human-AI relationships
• AI deskilling is a structural problem
• A Longitudinal Randomized Control Study of Companion Chatbot Use: Anthropomorphism and Its Mediating Role on Social Impacts
• Integrating LLM in Agent-Based Social Simulation: Opportunities and Challenges
• Value Profiles for Encoding Human Variation
• LLM Generated Persona is a Promise with a Catch
• Dehumanizing Machines: Mitigating Anthropomorphic Behaviors in Text Generation Systems
• Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations
• Why human-AI relationships need socioaffective alignment
• Position: Evaluating Generative AI Systems Is a Social Science Measurement Challenge
• Steering AI-Driven Personalization of Scientific Text for General Audiences
• The Dark Side of AI Companionship: A Taxonomy of Harmful Algorithmic Behaviors in Human-AI Relationships
• How developments in natural language processing help us in understanding human behaviour
• Development and validation the Problematic ChatGPT Use Scale: a preliminary report
• Scaling Synthetic Data Creation with 1,000,000,000 Personas
• The effects of over-reliance on AI dialogue systems on students' cognitive abilities: a systematic review
• τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains
• STAR: SocioTechnical Approach to Red Teaming Language Models
• Believing Anthropomorphism: Examining the Role of Anthropomorphic Cues on Trust in Large Language Models
• WildChat: 1M ChatGPT Interaction Logs in the Wild
• Mechanistic Interpretability for AI Safety -- A Review
• SafetyPrompts: a Systematic Review of Open Datasets for Evaluating and Improving Large Language Model Safety
• Large Language Models for Data Annotation and Synthesis: A Survey
• R-Judge: Benchmarking Safety Risk Awareness for LLM Agents
• Are Large Language Models Temporally Grounded?
• Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMs
• The History and Risks of Reinforcement Learning and Human Feedback
• Large Language Models Help Humans Verify Truthfulness -- Except When They Are Convincingly Wrong
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model
• OpenAssistant Conversations -- Democratizing Large Language Model Alignment
• Artificial Influence: An Analysis Of AI-Driven Persuasion
• A Longitudinal Study of Self-Disclosure in Human–Chatbot Relationships
• Improving alignment of dialogue agents via targeted human judgements
• Training language models to follow instructions with human feedback
• Red Teaming Language Models with Language Models
• Ethical and social risks of harm from Language Models
• Fine-Tuning Language Models from Human Preferences
• Emotion Recognition in Conversation: Research Challenges, Datasets, and Recent Advances
• Analyzing Polarization in Social Media: Method and Application to Tweets on 21 Mass Shootings
• Concrete Problems in AI Safety
• Different types of well-being? A cross-cultural examination of hedonic and eudaimonic well-being.
• The WHO-5 Well-Being Index: A Systematic Review of the Literature
• New Well-being Measures: Short Scales to Assess Flourishing and Positive and Negative Feelings
• Performance of an Abbreviated Version of the Lubben Social Network Scale Among Three European Community-Dwelling Older Adult Populations
• Trust in Automation: Designing for Appropriate Reliance
• Extending the Cross-Cultural Validity of the Theory of Basic Human Values with a Different Method of Measurement
• UCLA Loneliness Scale (Version 3): Reliability, Validity, and Factor Structure
• Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology
• The Satisfaction With Life Scale
• Happiness is everything, or is it? Explorations on the meaning of psychological well-being.