Long-term Measurements: Towards a Longitudinal Understanding of Human-AI Interactions
Not on the map. Why?
Abstract (arXiv)
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.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 46 of its 128 references that are in the map. First posted on arXiv 3 Aug 2026.
Builds on
46 of 128 references.
References · 46
- 2022Training language models to follow instructions with human feedbackcited by 15,712
- 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
- 2019Fine-Tuning Language Models from Human Preferencescited by 2,124
- 2016Concrete Problems in AI Safetycited by 1,582
- 2024τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domainscited by 1,280
- 2022Red Teaming Language Models with Language Modelscited by 983
- 2021Ethical and social risks of harm from Language Modelscited by 957
- 2023OpenAssistant Conversations -- Democratizing Large Language Model Alignmentcited by 712
- 2024WildChat: 1M ChatGPT Interaction Logs in the Wildcited by 598
- 2022Improving alignment of dialogue agents via targeted human judgementscited by 527
- 2024Mechanistic Interpretability for AI Safety -- A Reviewcited by 415
- 2024Scaling Synthetic Data Creation with 1,000,000,000 Personascited by 397
34 more
- 2024R-Judge: Benchmarking Safety Risk Awareness for LLM Agentscited by 268
- 2024Large Language Models for Data Annotation and Synthesis: A Surveycited by 201
- 2023Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMscited by 191
- 2019Emotion Recognition in Conversation: Research Challenges, Datasets, and Recent Advancescited by 118
- 2025Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversationscited by 88
- 2025LLM Generated Persona is a Promise with a Catchcited by 77
- 2024SafetyPrompts: a Systematic Review of Open Datasets for Evaluating and Improving Large Language Model Safetycited by 75
- 2019Analyzing Polarization in Social Media: Method and Application to Tweets on 21 Mass Shootingscited by 67
- 2020A Survey of Reinforcement Learning Algorithms for Dynamically Varying Environmentscited by 62
- 2024The Dark Side of AI Companionship: A Taxonomy of Harmful Algorithmic Behaviors in Human-AI Relationshipscited by 61
- 2025Position: Evaluating Generative AI Systems Is a Social Science Measurement Challengecited by 61
- 2024The opportunities and risks of large language models in mental healthcited by 54
- 2023Large Language Models Help Humans Verify Truthfulness -- Except When They Are Convincingly Wrongcited by 51
- 2025Why human-AI relationships need socioaffective alignmentcited by 47
- 2023The History and Risks of Reinforcement Learning and Human Feedbackcited by 42
- 2024Believing Anthropomorphism: Examining the Role of Anthropomorphic Cues on Trust in Large Language Modelscited by 36
- 2023Artificial Influence: An Analysis Of AI-Driven Persuasioncited by 31
- 2025Value Profiles for Encoding Human Variationcited by 29
- 2023Are Large Language Models Temporally Grounded?cited by 25
- 2026Sycophantic Chatbots Cause Delusional Spiraling, Even in Ideal Bayesianscited by 20
- 2025Neural steering vectors reveal dose and exposure-dependent impacts of human-AI relationshipscited by 17
- 2024STAR: SocioTechnical Approach to Red Teaming Language Modelscited by 13
- 2025Dehumanizing Machines: Mitigating Anthropomorphic Behaviors in Text Generation Systemscited by 11
- 2026Sycophantic AI makes human interaction feel more effortful and less satisfying over timecited by 11
- 2025A Longitudinal Randomized Control Study of Companion Chatbot Use: Anthropomorphism and Its Mediating Role on Social Impactscited by 10
- 2025From the Fluency Fallacy to the Micro-to-Macro Validity Gap: Opportunities and Pitfalls of LLMs in Social Simulationcited by 8
- 2024Steering AI-Driven Personalization of Scientific Text for General Audiencescited by 7
- 2026Evaluating Language Models for Harmful Manipulationcited by 6
- 2026Positive Alignment: Artificial Intelligence for Human Flourishingcited by 5
- 2026Investigating the Effects of LLM Use on Critical Thinking Under Time Constraints: Access Timing and Time Availabilitycited by 4
- 2026AI-Wrapped: Participatory, Privacy-Preserving Measurement of Longitudinal LLM Use In-the-Wildcited by 3
- 2026"AI Psychosis" in Context: How Conversation History Shapes LLM Responses to Delusional Beliefscited by 2
- 2026The efficiency-gain illusion: People underestimate the rate of AI use and overestimate its benefits on simple taskscited by 2
- 2026Detecting and Controlling Sycophancy with Cascading Linear Featurescited by 2
Built on by
Not cited yet.