# INTIMA: A Benchmark for Human-AI Companionship Behavior

arXiv: https://arxiv.org/abs/2508.09998 · HTML: https://mlsanity.com/paper/2508.09998/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 13 of its 33 references that are in the map.
- First posted on arXiv 4 Aug 2025.

## Builds on (most cited first, 13 of 13)

- [Large Language Models are not Fair Evaluators](https://mlsanity.com/paper/2305.17926/) (2023)
- [Bias and Fairness in Large Language Models: A Survey](https://mlsanity.com/paper/2309.00770/) (2023)
- [LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods](https://mlsanity.com/paper/2412.05579/) (2024)
- [SycEval: Evaluating LLM Sycophancy](https://mlsanity.com/paper/2502.08177/) (2025)
- [Systematic Evaluation of LLM-as-a-Judge in LLM Alignment Tasks: Explainable Metrics and Diverse Prompt Templates](https://mlsanity.com/paper/2408.13006/) (2024)
- [The Dark Side of AI Companionship: A Taxonomy of Harmful Algorithmic Behaviors in Human-AI Relationships](https://mlsanity.com/paper/2410.20130/) (2024)
- [Why human-AI relationships need socioaffective alignment](https://mlsanity.com/paper/2502.02528/) (2025)
- [DarkBench: Benchmarking Dark Patterns in Large Language Models](https://mlsanity.com/paper/2503.10728/) (2025)
- [Multi-turn Evaluation of Anthropomorphic Behaviours in Large Language Models](https://mlsanity.com/paper/2502.07077/) (2025)
- [Transparent AI Disclosure Obligations: Who, What, When, Where, Why, How](https://mlsanity.com/paper/2403.06823/) (2024)
- [Drama Engine: A Framework for Narrative Agents](https://mlsanity.com/paper/2408.11574/) (2024)
- [Can Large Language Models Be Good Companions? An LLM-Based Eyewear System with Conversational Common Ground](https://mlsanity.com/paper/2311.18251/) (2023)
- [How Do Users Negotiate Harmful Value Conflicts with AI Companions? A Study with Minion, a Technology Probe for In-Situ Human-AI Conflict Response](https://mlsanity.com/paper/2411.07042/) (2024)

## Built on by (most cited first, 8 of 8)

- [CompanionBench: A Theory-Anchored, Real-World-Grounded Benchmark for AI Emotional Companionship](https://mlsanity.com/paper/2608.02046/) (2026)
- [Using LLM-as-a-Judge/Jury to Advance Scalable, Clinically-Validated Safety Evaluations of Model Responses to Users Demonstrating Psychosis](https://mlsanity.com/paper/2604.02359/) (2026)
- [Frictionless Love: Associations Between AI Companion Roles and Behavioral Addiction](https://mlsanity.com/paper/2604.20011/) (2026)
- [Young children's anthropomorphism of an AI chatbot: Brain activation and the role of parent co-presence](https://mlsanity.com/paper/2512.02179/) (2025)
- [Not What, But How: A Framework for Auditing LLM Responses across Positioning, Generalization, Anthropomorphism, and Maxims](https://mlsanity.com/paper/2606.02493/) (2026)
- [Agentic Relationship Harm: Benchmarking and Gating Relational Manipulation in AI Agents](https://mlsanity.com/paper/2606.03271/) (2026)
- [Examining Human-Like Behaviors in LLMs: A Multi-Dimensional Analysis of Model Behaviors, User Factors, and System Prompts](https://mlsanity.com/paper/2606.18258/) (2026)
- [A Responsive Present, a Shared Past, a Social Other: Teens' Overreliance on Companion AI Chatbots](https://mlsanity.com/paper/2609.14843/) (2026)

## Abstract (arXiv)

AI companionship, where users develop emotional bonds with AI systems, has emerged as a significant pattern with positive but also concerning implications. We introduce Interactions and Machine Attachment Benchmark (INTIMA), a benchmark for evaluating companionship behaviors in language models. Drawing from psychological theories and user data, we develop a taxonomy of 31 behaviors across four categories and 368 targeted prompts. Responses to these prompts are evaluated as companionship-reinforcing, boundary-maintaining, or neutral. Applying INTIMA to Gemma-3, Phi-4, o3-mini, and Claude-4 reveals that companionship-reinforcing behaviors remain much more common across all models, though we observe marked differences between models. Different commercial providers prioritize different categories within the more sensitive parts of the benchmark, which is concerning since both appropriate boundary-setting and emotional support matter for user well-being. These findings highlight the need for more consistent approaches to handling emotionally charged interactions.